ImpactMojoQualitative Analysis Software 101www.impactmojo.in

ImpactMojo 101 Series · Free Forever

Qualitative
Analysis
Software 101

NVivo, MAXQDA, ATLAS.ti, Taguette and QualCoder — from Transcript to Theme, with the Thinking Left to You

Research MethodsSouth Asia Focus100 SlidesFree Access

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Agenda

What We Cover

What the Software Does, and Does Not

Slides 3–10

Preparing the Data

Slides 11–19

The Coding Workflow

Slides 20–28

Taguette: The Free Start

Slides 29–36

QualCoder: Free and Full

Slides 37–44

NVivo, MAXQDA and ATLAS.ti

Slides 45–53

Queries, Matrices and Visuals

Slides 54–62

Coding in Teams

Slides 63–70

Mixed Methods

Slides 71–78

Rigour and Reporting

Slides 79–88

AI, Choosing a Tool, and Practice

Slides 89–99

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Section One

What the Software Does, and Does Not

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Definition

QDA software is a filing system for evidence, not an analyst

Qualitative data analysis software (QDAS, CAQDAS)

A program that stores texts, audio, video and images; lets the researcher attach codes to segments of them; keeps memos and attributes alongside; and retrieves, counts, cross-tabulates and displays what was coded. It does the bookkeeping of analysis. It does not do the analysis.

Forty interviews in three languages, coded by hand with highlighters and index cards, is possible and was done for decades. Software makes the retrieval instant ('every passage coded fear of the bank among women in Gaya'), the codebook consistent, the audit trail automatic, and the team's work mergeable. The interpretation, the codes' meanings, the themes and the claims come from the researcher, exactly as before.

What it stores: documents, transcripts, PDFs, audio and video with time-stamped coding, images, survey open-ends, field notes, photographs.
What it attaches: codes (a hierarchy), memos (your thinking), attributes (respondent age, district, arm), links between segments.
What it retrieves: all segments by code, by code and attribute, by co-occurrence, by proximity; word frequencies; matrices of codes by cases.
What it displays: code trees, matrices, co-occurrence tables, networks, comparison diagrams, word clouds (of little use).

The recurring failure in student work: a project with 240 codes and no argument. The software made the codes easy; nothing made the argument.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

The Field

The tools: five to know, and what each is

Tool	Maker	Cost (approximate, 2026)	Where it stands
Taguette	Rémi Rampin and contributors; open source	Free; runs locally or on a server; a hosted version at taguette.org	The simplest tool: highlight and tag text; exports everything. The right first tool and enough for many projects
QualCoder	Colin Curtain and contributors; open source, Python	Free; desktop	Full-featured: text, PDF, image, audio and video coding; attributes; queries; coder comparison; REFI-QDA export
NVivo	Lumivero (formerly QSR International)	Student licences roughly US$100 a year; institutional site licences common at Indian universities	The most widely taught; strong queries and matrices; transcription and AI add-ons
MAXQDA	VERBI Software, Berlin	Similar student pricing; educational licences	Strong on mixed methods and visual tools; a Stats module; team cloud
ATLAS.ti	ATLAS.ti Scientific Software Development, Berlin	Similar; a web version by subscription	Quotation-based; networks; the earliest AI coding features
Also: Dedoose, Quirkos, Delve	Various	Monthly subscriptions	Web-based, simpler; Dedoose is built for mixed-methods teams

Prices change; check the vendors. What does not change: the free tools now cover the coding workflow completely, the commercial three differ in interface and advanced features more than in what they let you conclude, and every tool can export a project to the REFI-QDA exchange format so a team need not agree on one.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Method First

The method is not in the software

Thematic analysis, framework analysis, grounded theory, content analysis, narrative analysis and discourse analysis all use codes and all can be done in any of these tools. The software does not know which you are doing, and a project that starts in NVivo without a stated method ends as an inductive thematic analysis by default, because that is what unstructured coding produces. Decide the method from the question; then use the software to do that method's steps.

Reflexive thematic analysis (Braun and Clarke, Qualitative Research in Psychology 2006, 3:77; updated 2019 and 2021): codes are the researcher's interpretive labels; themes are built, not found; no codebook agreement is sought.
Framework analysis (Gale and colleagues, BMC Medical Research Methodology 2013, 13:117): a matrix of cases by themes, charted; built for applied policy research with deadlines and teams.
Grounded theory (Charmaz, Constructing Grounded Theory, 2nd ed., Sage, 2014): open, focused and theoretical coding toward a theory; constant comparison; memos central.
Content analysis: counting categories in text; the one method where the software's counts are the result.

Saldaña's manual

The Coding Manual for Qualitative Researchers (4th ed., Sage, 2021) catalogues thirty-odd coding methods (descriptive, in vivo, process, values, versus, and so on) and says which suit which questions. It is method-neutral about software and the best companion to this course. Qualitative Methods 101 in this series is the course on the methods themselves; this one is on the tools that carry them.

Write the method and the coding approach in the protocol before opening the software. The tool then has a job.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

When

When software helps, and when a table in Word is enough

Situation	Software worth it?	Why
Six key-informant interviews, one analyst, a report in three weeks	Marginal	A framework matrix in Word or a spreadsheet does it; the software's setup costs a day
Forty interviews, two languages, two coders, a thesis	Yes	Retrieval, consistency, merging and the audit trail; hand methods fail at this scale
Two hundred open-ended survey responses	Yes, a simple tool	Taguette or the spreadsheet route (section 09); coding is fast, counting is the point
Focus groups with audio, needing time-stamped coding of who said what	Yes, a fuller tool	QualCoder, NVivo, MAXQDA and ATLAS.ti code audio directly
A multi-year programme with recurring qualitative monitoring	Yes, with a shared codebook	The codebook and the project persist across rounds and staff
Documents and policy texts, hundreds of pages	Yes	Search, auto-coding by structure, and document attributes
Photographs and video from participatory work	Yes, a fuller tool	Region coding of images; video time-stamps

The threshold is roughly a dozen sources or two coders. Below it, the software is a habit to build for later; above it, the alternative is a filing system that cannot be searched.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Vocabulary

Vocabulary: the same ideas under five names

Idea	NVivo	MAXQDA	ATLAS.ti	QualCoder	Taguette
A source file	File (formerly Source)	Document	Document	File	Document
A coded passage	Reference / coded content	Coded segment	Quotation	Coded segment	Highlight
A code	Code (formerly Node)	Code	Code	Code	Tag
A code hierarchy	Code tree; parent and child codes	Code system	Code groups; folders	Code categories	Nested by naming
Respondent characteristics	Case classification attributes	Document variables	Document groups; attributes	Attributes	—
A note to yourself	Memo; annotation	Memo	Memo; comment	Memo; journal	—
Cases by codes table	Matrix coding query; framework matrix	Code matrix browser; crosstab	Code-document table	Coding matrix	Export and count
Co-occurrence	Matrix coding query (code × code)	Code relations browser	Code co-occurrence table	Code co-occurrence	—
Project exchange	REFI-QDA export/import	Same	Same	Same	CSV / DOCX export

This course uses the plain words (source, segment, code, attribute, memo, matrix) and names the tool's term where a menu is described. A researcher who knows the ideas can learn any of the five in a week.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

South Asia

What is different about qualitative work here

Languages. Interviews in Hindi, Bangla, Tamil, Marathi, Odia or a mix, with code-switching within a sentence; transcription and translation decisions decide what can be coded and by whom. Section 02.
Scripts. Every tool here handles Unicode Indic text for display and search; word frequencies and auto-coding are built for English and behave unpredictably on inflected Indian languages. Test before relying on them.
Teams. Field researchers who did the interviews, an analyst who did not, and a supervisor who reads English: the workflow has to carry meaning across all three. Section 08.
Connectivity and cost. Desktop tools that run offline (QualCoder, NVivo, MAXQDA, ATLAS.ti desktop) versus web tools that need a connection (Taguette hosted, Dedoose, ATLAS.ti Web). District offices decide this.
Data protection. Interview recordings and transcripts are personal data under India's Digital Personal Data Protection Act 2023; where they are stored, and whether an AI feature sends them abroad, is a compliance question. Sections 02 and 11.

The commonest South Asian workflow, translating everything to English first and coding the English, is also the one that loses the most. Section 02 sets out the alternatives and their costs.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

This Course

How the course is arranged

PREPARE: consent, transcription, translation, file formats, attributes (02)

→

CODE: codebook, first cycle, second cycle, memos, saturation (03)

→

TOOLS: Taguette (04), QualCoder (05), the commercial three (06)

→

ANALYSE: queries, matrices, visuals (07)

→

TEAM: agreement, merging, versions (08)

→

INTEGRATE: with surveys and numbers (09)

→

REPORT: rigour, transparency, COREQ and SRQR (10)

→

DECIDE: AI features, tool choice, a worked project (11)

Sections 02, 03, 07 to 10 are about the work and apply whatever the tool. Sections 04 to 06 are about the tools, with the free ones first and in more detail because they are the ones most readers of this course can install today. Worked examples follow one illustrative study: 36 interviews with women in self-help groups in Bihar and West Bengal, in Hindi and Bangla, about savings and household decisions.

Qualitative Methods 101 covers interviewing, sampling and the methods; Research Ethics 101 covers consent. This course starts where the recordings exist.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Section Two

Preparing the Data

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Consent

Consent decides what the software may hold and where

Before a recording is imported, three questions from the consent form decide the workflow. Did the participant agree to recording, to transcription by a third party, and to storage of the transcript? Did they agree to quotation in publications, with what anonymisation? Did they agree to the data being processed by an external service (a transcription API, a cloud QDA tool, an AI assistant)? A 'no' to the last removes most cloud features from the project.

Store recordings and transcripts on encrypted, access-controlled storage; the QDA project file contains the transcripts and needs the same protection.
Pseudonymise at transcription (a participant ID, place names generalised to district) and keep the key separately.
Under India's DPDP Act 2023, participants are data principals with rights to know and to erasure; a research exemption exists for specified purposes but does not cover careless handling. Data Protection & the DPDP Act 101 in this series is the reference.
For cross-border processing (most AI and transcription services), check the consent wording and the institution's policy.

Data sharing later

Funders increasingly ask for qualitative data to be archived (the Qualitative Data Repository at Syracuse, UK Data Service). That is possible only if consent covered it and the transcripts are de-identified to a documented standard. Decide at the consent stage; it cannot be added afterwards.

Record in the project's first memo what each participant consented to. The software can carry it as an attribute (quotation permitted: yes/no) and filter on it.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Transcription

Transcription: what level, by whom, in what script

Level	What is captured	Cost	Suits
Intelligent verbatim	Every word, with fillers and false starts removed	About 4–6 hours per hour of audio by hand; minutes by machine plus checking	Thematic and framework analysis; most applied work
Full verbatim	Every word, filler, repetition and pause	6–8 hours per hour	Discourse and conversation analysis; where the how matters
Jefferson-style	Timing, overlap, intonation marked	10+ hours per hour	Conversation analysis only
Summary or notes	Content summarised by the interviewer	Under an hour	Rapid appraisal; not for coding

For Indian languages, machine transcription has become usable: ImpactMojo's own VaniScribe runs on Sarvam's Indic speech models and returns Hindi, Tamil, Bangla and other transcripts in native script for a few rupees a minute; open-weight Indic models (Bodhan AI's Indic-Transcribe, released September 2026) can run locally where data may not leave the machine. All machine output needs a human pass by someone who heard the interview; the error rate on rural speech, mixed languages and named places is not small.

Transcribe in the language spoken, in its script, with a speaker label and a timestamp every minute. Translation is a separate step and a separate decision.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Translation

Translation: three workflows and what each loses

A. Code in the source language

Transcripts stay in Hindi and Bangla; coders read those languages; codes and memos are in English or bilingual; only the quotations used in the report are translated, by the analyst, with the original beside them. Loses least. Requires coders in each language and a bilingual codebook.

B. Translate everything, code the English

One translated corpus; any coder can work; supervisors can read it. Loses idiom, register, the word the respondent chose, and introduces the translator's interpretation before the analysis starts (Temple and Young, Qualitative Research 2004, 4:161). The commonest workflow and the most lossy.

C. Parallel: source and translation side by side

Both versions imported; coding on the source with the translation visible (MAXQDA and NVivo can display paired documents; QualCoder via a second file). Keeps the original in reach; doubles the storage; the practical compromise.

Whatever the workflow, record who translated, whether the translation was checked, and how untranslatable terms were handled (kept in transliteration with a gloss: izzat (honour, standing)).
Keep the source-language words for the concepts that carry the analysis. A theme called 'shame' built from sharam and lajja in two languages is a theme that needs both words in the report.
Machine translation of transcripts (Sarvam's Mayura, Bodhan's Indic-Translate, or general models) is now good enough for a coder's first read and not for a quotation in print.
The report says which workflow was used. Reviewers ask.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Files

File formats, structure and what imports cleanly

Material	Format to import	Notes
Transcripts	.docx or .txt, UTF-8; one file per interview	Speaker labels on their own line ('I:' and 'R:', or names); a header block with ID, date, place, language; consistent formatting lets NVivo and MAXQDA auto-code speakers
Audio	.mp3, .wav, .m4a	QualCoder, NVivo, MAXQDA and ATLAS.ti code audio with time-stamps and can link a transcript to it; Taguette does not take audio
Video	.mp4	Large files; compress first; region and time coding
PDFs (policy documents, reports)	Text-based PDF, not scanned	Scanned PDFs need OCR first (Bodhan's IndicOCR or Tesseract for Indic scripts); ATLAS.ti and MAXQDA code PDFs in place
Images, photographs	.jpg, .png	Region coding in QualCoder, NVivo, MAXQDA, ATLAS.ti
Survey open-ends	A spreadsheet: one row per respondent, one column per question, plus attribute columns	Import as a dataset (NVivo), a survey (MAXQDA), or a table of documents; each cell becomes codeable
Field notes, diaries	.docx, .txt	Date each entry; the researcher's own notes are data too
Social media, web pages	Exported text; NCapture (NVivo) or copy	Terms of service and ethics apply

Name files by a convention (INT_BR_014_HI for interview, Bihar, participant 14, Hindi) and keep a master sheet of files with attributes. The attributes are imported once and used in every matrix.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Attributes

Attributes: the variables that make qualitative data cross-tabulable

Each interview carries characteristics of the participant and the setting: state, district, SHG member or not, years of membership, age band, education, language, interviewer, date. Entered as attributes (NVivo case classifications, MAXQDA document variables, ATLAS.ti document groups, QualCoder attributes), they let every later query ask 'how does this theme appear among members versus non-members' or 'in Bihar versus Bengal', which is the comparison a mixed-methods evaluation needs.

Prepare a spreadsheet with one row per file and one column per attribute, and import it; typing attributes one by one is where errors enter.
Use the same codes as the survey (the same district names, the same arm labels) so qualitative and quantitative results line up in section 09.
Include the interviewer as an attribute. Interviewer effects are real and a matrix by interviewer is a quality check.
Attributes are for facts known before analysis. Judgements made during analysis (this woman is 'highly empowered') are codes, not attributes; keep them apart.

Focus groups

A focus group is one file with many speakers. Attributes belong to the group (village, date, composition) and speakers are coded as speakers (NVivo and MAXQDA auto-code by speaker label), so that 'what did the older women say' is a query on speaker code × theme. Without speaker coding, a focus group is one undifferentiated voice, which is the commonest way focus-group data is wasted.

Ten attributes, a spreadsheet, an hour. The return is every comparison in section 07.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Project Setup

Setting up the project: folders, naming, the first memo

Element	Convention	Why
Project file	One per study; named with the study and date; on encrypted storage; backed up daily to a second place	Project files corrupt and laptops are stolen
Source folders	By type and site: Interviews/Bihar, Interviews/Bengal, FGDs, Documents, Field notes	Attributes do the fine sorting; folders do the coarse
File names	TYPE_SITE_ID_LANG (INT_WB_022_BN)	Sortable; language visible; ID matches the master sheet and the survey
Attribute sheet	One spreadsheet, imported; the master list	Section 02
Codebook	Started from the protocol's a priori codes, with definitions; versioned	Section 03
Memos	A project memo (decisions, dated), a methods memo, a memo per emerging theme, and coding memos as needed	The audit trail; section 10
Versions	Save-as with a date at each milestone; never overwrite the only copy	Recovery and the trail
Team	One master project; coders work on copies and merge (section 08), or use a server or cloud version	Merging is where projects break

The first memo, written before any coding, states the question, the method, the coding approach, the translation workflow, the team's roles and the consent constraints. It is the methods section's first draft and the document a new team member reads first.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Cleaning Text

Cleaning transcripts before import: what to fix and what to leave

Fix: speaker labels (consistent, on their own line); encoding (UTF-8, so Devanagari and Bangla render); identifying details (names, phone numbers, specific villages) replaced with placeholders in square brackets; headers with ID and attributes.
Fix: transcription errors found on the checking pass, with the audio timestamp noted where the correction is a judgement.
Leave: the respondent's grammar, dialect, mixed languages and hesitations at the transcription level chosen. Cleaning speech into standard prose is editing the data.
Leave: the interviewer's questions. They are context for every answer and evidence of leading.

A checking pass

Someone who was present, or who speaks the language natively, reads each transcript against the audio for ten minutes at three points and corrects; if the error rate is high, the whole transcript is rechecked. Record the check as an attribute (checked: yes/no, by whom). Machine transcripts of Bhojpuri-inflected Hindi from a noisy courtyard need the whole pass.

Do the cleaning in the word processor before import. Editing text inside a QDA project after coding has started can detach codes from their passages in some tools.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Data Checklist

Before coding: the data checklist

Check	Done
Consent for recording, transcription, storage, quotation and external processing recorded per participant
Recordings and transcripts on encrypted storage with a backup
Transcription level chosen and stated; language and script decided
Every machine transcript checked by a native speaker; the check recorded
Translation workflow chosen (A, B or C) and stated; translators named
Files named by convention; one file per interview or group
Speaker labels consistent; focus-group speakers identifiable
Pseudonymised; key stored separately
Attribute sheet complete and imported; codes match the survey's
Project folders, first memo and a priori codebook in place
Project file backed up and versioned

A week for a 36-interview study, most of it the transcript checking. It is the week that decides whether the coding can be trusted, and it is the week most often compressed.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Section Three

The Coding Workflow

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

What a Code Is

A code is a label with a definition, an example and a boundary

Code

A word or short phrase that assigns a summative, salient, essence-capturing attribute to a portion of data (Saldaña). In software, a named container to which segments are attached. In a codebook, a name, a definition, when to use it, when not to, and an example.

The difference between a project that yields findings and one that yields a code list is the definition. 'Fear' is a word; 'FEAR OF FORMAL FINANCE: expressions of anxiety, avoidance or distrust toward banks, loan officers or paperwork, including stories of others' bad experiences; not general worry about money (see MONEY WORRY)' is a code another coder can apply and a reader can check.

Codebook field	Example
Name	FEAR_FORMAL_FINANCE
Definition	Anxiety, avoidance or distrust toward banks, officers, forms
Use when	Respondent describes her own or others' fear or avoidance
Do not use when	Worry about money in general (MONEY_WORRY); anger at a specific officer (BANK_MISTREATMENT)
Example	'I never went inside; they would ask for papers I don't have' (INT_BR_014)
Source	A priori, from the theory of change; revised 2026-03-04 after five transcripts

Every tool has a field for a code's description. Fill it. A code without a definition is a guess about what you meant, made later, by you.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

The Cycle

First-cycle and second-cycle coding

READ: every transcript once, without coding, with notes

→

A PRIORI CODES: from the protocol and theory of change, defined

→

FIRST CYCLE: code five transcripts closely; add inductive codes as they arise; memo

→

REVISE: merge, split, define; version the codebook; recode the five

→

CODE ALL: the rest, with the revised codebook; new codes flagged and reviewed

→

SECOND CYCLE: group codes into categories and themes; pattern coding; matrices

→

WRITE: themes as claims with evidence; back to the data to check

First-cycle coding labels the data; second-cycle coding (Saldaña's pattern, focused, axial and theoretical coding) organises the labels into an account. The software serves both: the first through attaching codes to segments, the second through the code hierarchy, merging, and the matrices of section 07. The revision step after five transcripts is where most of the codebook's quality is made, and it is the step most often skipped.

Recoding the first five transcripts after the codebook is revised is not wasted work. Those five were coded with a codebook that no longer exists.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Hierarchies

Code hierarchies: how deep, how many, and the two failure modes

Codes nest: SAVINGS contains SAVINGS_PURPOSE, SAVINGS_BARRIER, SAVINGS_GROUP_RULES; each of those contains two or three children. Two or three levels is workable; five is not. A project for 36 interviews on a focused question ends with perhaps 40 to 80 codes after revision, organised under eight to fifteen parents. Two hundred codes is a project that labelled every sentence and has not yet begun to analyse.

Too many codes: every segment gets its own label; retrieval returns one segment per code; nothing accumulates. Merge by definition.
Too few: 'EMPOWERMENT' holds 300 segments and every retrieval is the whole dataset. Split by the distinctions the segments show.
Descriptive parents (topics: SAVINGS, DECISIONS, HUSBAND) and analytic children (what is said about them) is a workable first structure; themes emerge across the parents.
Coding at the parent as well as the child is a setting in NVivo and MAXQDA ('aggregate'); decide it once.

Themes are not codes

A theme is a claim about the data ('women use group savings as a shield against household demands, which reshapes who decides'), built across codes and cases. In the software it may be a parent code that collects the evidence, or a memo that argues it, or a set (NVivo) or a document group (ATLAS.ti). Braun and Clarke's insistence that themes are constructed, not found, is a warning against reading the code tree as the findings.

Print the code tree at each version. The sequence of trees is the history of the analysis and belongs in the appendix.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Segments

How much to code: the unit of analysis

A code attaches to a segment: a phrase, a sentence, a paragraph, a whole answer. Too small and the segment means nothing out of context when retrieved; too large and every retrieval is a page to reread. The working rule is the smallest unit that makes sense on its own when retrieved, which is usually a sentence or two with the question that prompted it. Decide it in the codebook and apply it consistently, because inter-coder agreement (section 08) depends on it.

Overlapping codes on one segment are normal and are what co-occurrence queries use.
Code the interviewer's question with the answer when the question shaped it.
In vivo codes (the respondent's own phrase as the code name: 'mera paisa', my money) keep the language in the analysis and are worth a parent of their own.
Uncodeable material (small talk, the tea break) is left uncoded; 'everything must be coded' is a superstition.

Auto-coding and its limits

NVivo and MAXQDA can auto-code by structure (every answer to question 4 gets code Q4; every speaker gets a speaker code) and by text search (every passage containing 'bank' gets BANK_MENTION). The first is reliable and saves hours. The second finds words, not meanings, misses synonyms and Indian-language variants, and should be treated as a way to find passages to read, never as coding. AI-assisted coding (section 11) is the second kind with better vocabulary and the same limit.

A human coder reading with a defined codebook does about eight to twelve transcript pages an hour. Budget from that.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Memos

Memos: where the analysis actually happens

A memo is a dated note in the project: about a code (what it is coming to mean, how it differs from its neighbour), about a case (what is distinctive about this woman's account), about a pattern (members in Bengal talk about the bank differently from Bihar), about the method (why the codebook changed), or about the researcher (what I expected and am seeing). Every tool has them and links them to codes, files and segments. Grounded theorists treat memo-writing as the analysis; every method benefits from it.

Write a memo whenever you make a decision, notice a pattern, or feel surprised. Surprise is data about your assumptions.
Date every entry; the sequence is the audit trail.
Theme memos become the findings' first drafts; method memos become the methods section; reflexive memos become the reflexivity statement.
Link memos to the segments that prompted them, so the evidence is one click away.

Illustrative theme memo, 2026-03-11

'Across the eleven Bihar members coded so far, SAVINGS_PURPOSE co-occurs with HUSBAND_DEMANDS in seven; the phrase is some version of the money being "in the group" and therefore not available. Non-members describe savings as at home and "taken". The group seems to function as a socially acceptable refusal. Check: Bengal members (not yet coded); whether the four Bihar exceptions are women whose husbands are migrants (attribute query). Possible theme: the group as shield. Related: Kabeer's control over resources.'

That memo is a finding, a test, a hypothesis and a citation, in one paragraph, written at transcript eleven.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Saturation

Saturation: what the software can show and what it cannot

Saturation is the point at which further data add no new codes or refine no themes. The software can show when new codes stopped appearing (a count of codes created per transcript, in order coded) and when a theme's evidence stopped growing. It cannot tell you whether you would have found something new had you interviewed a different kind of person. Guest, Bunce and Johnson (Field Methods 2006, 18:59) found most new codes within the first twelve interviews of a homogeneous sample; Braun and Clarke argue saturation is the wrong frame for reflexive thematic analysis at all.

Report the code-creation curve if you claim saturation; it is a matrix of transcript order by new codes, available in any tool.
Report which subgroups the claim covers: saturation among Bihar members says nothing about Bengal non-members.
'Data saturation was reached' without evidence is the qualitative equivalent of 'the model fit'.

The honest sentence: 'No new codes were created after the twenty-eighth of thirty-six interviews; the last eight added instances to existing codes. The sample was not designed to reach saturation among non-members in Bengal (n = 5), and findings for that group are indicative.'

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Worked Codebook

A worked codebook extract: the SHG study after revision

Code	Definition (abridged)	Origin	Segments	Files
SAVINGS	Parent: any talk of saving money, in group or otherwise	A priori	—	36
SAVINGS_PURPOSE	What the money is for: emergencies, school, ceremonies, business	A priori	94	34
SAVINGS_SHIELD	Savings described as protected from household demands by being 'in the group'	Inductive, transcript 6	41	22
SAVINGS_BARRIER	What prevents saving: income, husband, group rules, distance	A priori	67	31
DECISIONS	Parent: who decides what, and how that is negotiated	A priori	—	36
DECISION_SMALL	Daily purchases, food, children's small needs	A priori	58	33
DECISION_LARGE	Assets, loans, ceremonies, migration	A priori	72	35
DECISION_STRATEGY	How women get a say: timing, allies, group backing, withholding	Inductive, transcript 9	53	27
FORMAL_FINANCE	Parent: banks, officers, forms, phones	A priori	—	30
FEAR_FORMAL_FINANCE	Anxiety, avoidance, distrust	A priori, redefined	38	21
IN_VIVO	Respondents' own phrases kept as codes	Inductive	29	19

Illustrative. Eleven of the study's 46 codes, with their origin and their counts from the code-summary report every tool produces. Two inductive codes (SHIELD, STRATEGY) became the analytic centre of the study and neither was in the protocol. The counts are for orientation (a code in 22 of 36 files is widespread; one in 3 is not) and are not findings.

Version 3 of the codebook, dated. Versions 1 and 2 are in the appendix with what changed and why.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Coding Checklist

The coding workflow: the checklist

Practice	Done
Method and coding approach stated in the first memo
A priori codes defined from the protocol before coding
Every transcript read once before coding
First five transcripts coded closely; codebook revised; the five recoded
Every code has a definition, use and do-not-use rules, and an example
Codebook versioned with dates and reasons
Segment size decided and consistent
Memos written at decisions, patterns and surprises; dated; linked
Inductive codes flagged and reviewed by a second person
Second-cycle grouping done deliberately, with theme memos
Saturation, if claimed, evidenced by the code-creation curve
Auto-coding and text search used to find, never to code

Twelve rows. They apply in Taguette and in NVivo equally, and they are what a methods reviewer means by 'systematic'.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Section Four

Taguette: The Free Start

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

What It Is

Taguette: highlight, tag, export, and nothing else

Taguette is a free, open-source tool that runs in the browser: install it with one command on a laptop, or use the hosted version at taguette.org. A project holds documents; you select a passage and give it one or more tags; tags have descriptions; everything exports to Word, CSV or HTML. It handles text (.docx, .pdf, .txt, .html, .odt, .rtf, .epub) and nothing else: no audio, no video, no attributes, no memos beyond tag descriptions, no queries beyond 'show me everything with this tag'.

Right for: a first qualitative project; a small study of text; teaching coding without a licence; a team that needs to code together in a browser.
Not right for: audio coding, attribute comparisons, co-occurrence analysis, large corpora.
Everything you tag can be taken to a fuller tool later: the highlight export is a spreadsheet of segment, tag and document.

Install and run

On a machine with Python: pip install taguette, then taguette, and the browser opens at a local address; data stays on the machine. Windows and Mac installers exist. The hosted service at app.taguette.org needs an account and stores your documents on its server, which the consent form must allow; a self-hosted server (the same software on an institution's machine) gives the collaboration without the storage question.

Ten minutes from download to first tag. That is the argument for starting here.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Workflow

A Taguette project, step by step

NEW PROJECT: name it; add a description that is the first memo

→

ADD DOCUMENTS: one per interview; the file name carries the ID and language

→

CREATE TAGS: the a priori codebook, each with its definition in the description field

→

HIGHLIGHT: select text, choose tags (several at once), save

→

REVIEW: click a tag to see every highlight; merge or rename tags as the codebook evolves

→

EXPORT: codebook, highlights by tag, and the whole project, for the appendix and the next tool

Tags are flat, so a hierarchy is made by naming: SAVINGS.purpose, SAVINGS.shield, SAVINGS.barrier. Searching 'SAVINGS' then finds the family. Merging two tags (select both, merge) reassigns every highlight, which is how first-cycle codes become second-cycle ones. Tag descriptions are the only memo field, so the codebook's definitions live there and the analytic memos live in a separate document.

Highlights can overlap and a passage can carry several tags. The 'all highlights' view with its tag filter is the retrieval tool, and for a study of twenty documents it is enough.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Collaboration

Coding together in Taguette

On a server (hosted or self-hosted), a project owner adds collaborators by username with full or view-only access, and several people can tag the same documents at once; each highlight records who made it. That makes Taguette the simplest tool for a small team coding together, and for a supervisor who wants to see a student's coding as it happens. It does not compute agreement; section 08's spreadsheet method does that from the highlight export.

Agree the codebook in the tag descriptions first, so every coder reads the same definition.
For an agreement check, two coders tag the same three documents with distinct tag sets (A.SAVINGS, B.SAVINGS), export, and compare.
The local version is single-user; sharing a local project means sharing the .sqlite3 file, which is not simultaneous.

A self-hosted Taguette on an institutional server is the cheapest collaborative QDA environment that keeps data inside the institution. An IT department can set it up in an afternoon.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Exports

What comes out, and what to do with it

Export	Format	Use
Codebook	.csv, .docx, .xlsx, .html	The appendix table of tags and definitions; the codebook for a fuller tool
Highlights for one tag	.docx, .html, .csv	The evidence for a theme, ready to quote; a document to read through when writing
All highlights	.csv (document, tag, text)	A spreadsheet for counting tags by document, for the agreement check, and for pivot tables by attribute if you add the attribute columns by hand
A document with its highlights	.docx, .html	A marked-up transcript for a supervisor or a participant check
The whole project	.sqlite3	The backup; reimportable into another Taguette
REFI-QDA	Not supported	Move to a fuller tool via the CSV export and re-code, or via QualCoder's import of the CSV as a starting point

The all-highlights CSV is the most useful object Taguette produces: opened in a spreadsheet with the attribute sheet joined on document name, it supports every count and matrix in section 07, by pivot table. That is a serviceable analysis for a small study and a good way to learn what the fuller tools automate.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Limits

Where Taguette runs out, and the signs

Need	Taguette	Move to
Coding audio or video directly	No	QualCoder (free), or the commercial three
Attributes (district, arm) and matrices by them	By hand in the CSV export	QualCoder or the commercial three
Memos linked to segments and codes	Tag descriptions only	Any fuller tool
Co-occurrence and proximity queries	No	QualCoder (basic), commercial (full)
A real hierarchy with aggregation	Naming convention only	Any fuller tool
Inter-coder agreement statistics	By hand from the export	QualCoder, NVivo, MAXQDA, ATLAS.ti
Hundreds of documents	Slows; the flat tag list becomes unwieldy	Any fuller tool
REFI-QDA exchange	No	QualCoder and the commercial three

The sign is usually the third time you build the same pivot table by hand. A project that started in Taguette moves to QualCoder with a morning's re-coding from the highlight export, and the codebook comes with it.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Teaching

Taguette in the classroom: a three-session sequence

Session	Activity	Output
1	Install; import three anonymised transcripts; create six a priori tags from a supplied codebook; tag one transcript each	A tagged transcript; a first argument about a definition
2	Two students tag the same transcript with prefixed tags; export; compare in a spreadsheet; revise the codebook together	An agreement table; codebook version 2
3	Tag the rest; export highlights by tag; write one theme memo with three quotations	A one-page theme memo

Three sessions teach what coding is, why definitions matter, why coders disagree, and what a theme looks like when it is built from evidence. None of it requires a licence, an account (the local version) or a fast machine. Students who then meet NVivo find the ideas familiar and the menus merely larger.

The disagreement in session 2 is the lesson. Do not resolve it for them.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Worked Taguette

A worked mini-study in Taguette: twelve interviews, one question

Tag	Description (abridged)	Highlights	Documents
SAVINGS.purpose	What the savings are for	31	12
SAVINGS.shield	Savings protected from household demands by being in the group	14	8
SAVINGS.barrier	What prevents saving	22	11
DECISION.large	Assets, loans, ceremonies	24	12
DECISION.strategy	How a say is obtained	18	9
FINANCE.fear	Fear or avoidance of banks	13	7
INVIVO	Respondents' phrases	11	8

Illustrative: the Bihar members only, coded in Taguette by one analyst in two days. The SAVINGS.shield tag, created at the sixth transcript, appears in eight of twelve; the export of its fourteen highlights is the evidence for the theme memo on the earlier slide.

What was possible, what was not

Possible: the codebook, the coding, the retrieval by tag, the theme, a report with quotations, and an appendix with the codebook and counts. Not possible without the spreadsheet: 'does SHIELD co-occur with HUSBAND demands' (section 07) and 'is SHIELD more common among women with migrant husbands' (attributes). For twelve interviews the spreadsheet did both in an hour.

The study was then extended to 36 interviews in two languages, and moved to QualCoder. The next section.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Section Five

QualCoder: Free and Full

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

What It Is

QualCoder: most of NVivo, for nothing, on your own machine

QualCoder is a free, open-source desktop program written in Python by Colin Curtain and contributors. It codes text, PDFs, images, audio and video; has a code hierarchy with categories; memos on codes, files, cases, segments and a journal; cases and attributes for files and cases; text search and auto-coding; reports by code, file, case and attribute; code frequencies and a co-occurrence matrix; a coder comparison with agreement statistics; charts; and import and export in the REFI-QDA exchange format. It runs offline. The project is a folder with a database inside.

Right for: any project this course describes, up to the size where a team needs simultaneous access.
The interface is functional rather than polished; the documentation (a wiki and a manual) is thorough.
Audio and video need VLC installed; PDFs are read with their text layer.

Install

Windows and Mac installers from the project's GitHub releases; on Linux, pip install qualcoder. Recent versions add optional LLM-assisted features that call an external model with an API key you supply; they are off by default and section 11 says what to think about before turning them on.

A student with a five-year-old laptop, no budget and 36 interviews in two languages has everything they need here.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

The Windows

The QualCoder interface: files, codes, cases, reports

Menu or tab	What it does	Used for
Project	New, open, close; import survey; project memo; settings (coder name, fonts, backups)	Setup; set the coder name before coding
Manage → Files	Import text, PDF, images, AV; edit file memos; assign attributes; bulk import from a folder	Section 02's data in
Manage → Cases	Create cases (a participant across several files); attach files or portions; case attributes	A woman interviewed twice; a focus-group speaker
Manage → Attributes	Define attributes for files and cases; import from CSV	District, arm, language, interviewer
Coding → Code text / image / AV	The coding workspace: file on the left, code tree on the right, memos and annotations	Sections 03's work
Coding → Codebook	Export the code tree with memos	The appendix
Reports	Coding reports by code, file, case, attribute; code frequencies; coder comparison; code relations; co-occurrence matrix; text mining	Section 07 and 08
Action log; Journals	The record of what was done; free-text journals	The audit trail and the memos

The coding workspace does the day's work; the reports do the analysis; the journals hold the thinking. Set the coder name in settings before the first code: the comparison report depends on it.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Coding in QualCoder

Coding text: the moves you make a hundred times a day

Open a file in Code text; select a passage; click a code (or right-click for the menu); the passage colours with the code's colour.
Create a code in place (right-click the tree); give it a memo (the definition) immediately.
Categories are folders in the tree; drag codes into them; a code can be in one category. Colours help; assign them by parent.
Annotations (a note on a passage without a code) and segment memos (a note on a coded segment) are both available; use annotations for 'come back to this' and memos for 'this is why'.
Overlapping codes show as overlapping colours; the segment list at the bottom shows every code on the selection.
Auto-code: by exact text, by sentence containing a phrase, or by speaker for structured transcripts. Use it to find, then read.

Indic text

Devanagari, Bangla and Tamil display and select correctly; set the font in settings to one with the script's coverage (Noto Sans Devanagari, Noto Sans Bengali). Text search works on exact strings; there is no stemming, so search for the root and the common inflections separately. The parallel-translation workflow (section 02, C) is done by opening the source-language file for coding with the English file in a second window.

Backups: QualCoder can save a timestamped copy of the project on every open; turn it on.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

AV Coding

Coding audio and video: time-stamped segments

Coding → Code audio/video opens the file with a player; you mark a start and end time and attach a code, exactly as with text, and the segment is the time span. A transcript can be linked to the audio so that the two scroll together, and coding either one codes both if the timestamps line up. For focus groups, where who said what is half the analysis, and for interviews where tone matters, this is the feature that justifies leaving Taguette.

Transcripts with timestamps every minute (section 02) make the linking work; VaniScribe's output carries them.
Code the audio for delivery (hesitation, laughter, a change of voice when the husband enters) and the transcript for content; the two code sets join in the reports.
Video is heavy: compress to 720p, and keep the originals elsewhere.

A qualitative study that codes only the transcript has thrown away the recording's information about how things were said. For most applied studies that is an acceptable loss and should be a stated one.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Reports

QualCoder's reports: retrieval, counts and the coder comparison

Report	What it gives	Section 07 or 08 equivalent
Coding report	Every segment for chosen codes, filtered by file, case or attribute value; exported to HTML, ODT, CSV	Retrieval: 'SHIELD among Bengal members'
Code frequencies	Counts of segments per code, by coder	The code-summary table
Code relations	Which codes overlap or are adjacent, with counts	Co-occurrence
Co-occurrence matrix	Code by code counts of overlapping segments	The co-occurrence table
Case and attribute reports	Codes by case; codes by attribute value	The matrix of cases by codes
Coder comparison	Agreement between two coders on selected codes and files: percentage agreement and Cohen's kappa	Section 08
Text mining	Word frequencies with a stop list	Orientation only
Charts	Bar and pie charts of code counts, hierarchy sunbursts	Report figures, with care

Every report exports as a file. The coding report filtered by attribute is the one used most: it is how 'what did non-members in Bihar say about the bank' becomes a document to read and quote from.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

REFI-QDA

REFI-QDA: moving a project between tools without losing the coding

The REFI-QDA project exchange standard (2019, from the Rotterdam Exchange Format Initiative) defines a .qdpx file that carries sources, codes, coded segments, memos, cases and attributes between tools. QualCoder, NVivo, MAXQDA, ATLAS.ti, Dedoose, Quirkos and Transana read and write it. A student can code in QualCoder and hand the project to a supervisor who uses NVivo; a team can move from a commercial licence to a free tool when the grant ends; an archive can accept the project in a documented format.

Not everything transfers perfectly: tool-specific objects (maps, queries, some memo links) are dropped; check the import and keep the original.
Export a .qdpx at every milestone as the archival copy, alongside the tool's own project file.
The codebook alone can be exchanged as REFI-QDC (.qdc) for sharing definitions across projects.

The standard is the reason the choice of tool matters less than it did. Choose for the work and the budget; the coding is portable.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Worked QualCoder

The SHG study in QualCoder: 36 interviews, two languages, two coders

Step	QualCoder	Result
Import	36 source-language transcripts and 36 English translations; 12 audio files linked; attributes from CSV (state, member, language, migrant husband, interviewer)	Project of 72 files, 36 cases
Codebook	Version 3 imported as a tree with memos	46 codes in 9 categories
Coding	Coder A (Hindi) and coder B (Bangla) on their language; each also codes four of the other's English translations for comparison	Coded in three weeks
Agreement	Coder comparison on the eight shared files, twelve key codes	Kappa 0.61 to 0.84; SHIELD 0.58 before, 0.79 after redefinition
Retrieval	Coding reports by attribute: SHIELD by state, by migrant husband	SHIELD in 15/18 with migrant husbands, 7/18 without
Co-occurrence	SHIELD × HUSBAND_DEMANDS matrix	Overlap in 19 segments across 14 files
Memos	Theme memos for SHIELD, STRATEGY, FEAR; a methods journal	The findings' first drafts
Export	Codebook, coding reports for the appendix, .qdpx for the archive	Report and deposit

Illustrative. The attribute retrieval is the moment the study found its central pattern: the group-as-shield theme was concentrated among women whose husbands migrated, whose remittances arrived as lumps that relatives claimed. The number (15 of 18) is a pointer, not a statistic; the fifteen accounts are the finding.

Everything in the table is free software on two laptops. The commercial tools (next) would have done it with more polish and no different conclusion.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Section Six

NVivo, MAXQDA and ATLAS.ti

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Side by Side

The commercial three: the same job, three designs

	NVivo	MAXQDA	ATLAS.ti
Maker; since	Lumivero (QSR International); 1999	VERBI, Berlin; 1989 (as winMAX)	ATLAS.ti GmbH, Berlin; 1993
Organising idea	Files, codes and cases in a database; queries as the analytic engine	Four windows: documents, codes, browser, retrieved segments; everything visible at once	Quotations as first-class objects; codes, memos and networks link to them
Strengths	Queries (matrix coding, coding comparison, crosstab); framework matrices; survey and social-media import; the most-taught	Mixed methods (crosstabs, joint displays, typology tables); visual tools (MAXMaps); paraphrasing and summary grids; a built-in Stats module	Networks; the query tool's operators (co-occurrence, proximity, embedding); PDF coding; the earliest and most developed AI features
Platforms	Windows and Mac (features differ); Collaboration Cloud	Windows and Mac; TeamCloud	Windows and Mac; a full Web version
Transcription	NVivo Transcription (paid, per hour); English-centred	Via partner services; a transcription mode with foot-pedal support	Via the Web version and partners
AI	AI Assistant: summaries, code suggestions	AI Assist: summaries, suggested codes, chat with data	Intentional AI coding, AI summaries, conversational AI
Learning curve	Steeper; many objects and windows	Gentlest of the three	Middle; the quotation model takes a day to click
Typical licence	Student around US$100 a year; institutional site licences common in India	Similar; educational licences; a free Reader	Similar; Web by subscription

Prices and features change yearly; verify with the vendor. The differences that matter for a project are the organising idea (which shapes how you think), the query engine (section 07), and which one your team and supervisor already use.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

NVivo

NVivo: files, codes, cases and the query menu

Data: Files (transcripts, PDFs, audio, video, pictures, datasets from surveys, web captures via NCapture) in folders. Codes in a tree, with Sentiment and Relationship codes as special kinds. Cases (a participant, a site) with classifications and attributes. Notes: memos and annotations. Queries: text search, word frequency, coding, matrix coding (codes by codes or by attributes), coding comparison (agreement between users), compound, crosstab. Explore: maps, charts, hierarchy charts, framework matrices. Everything is a saved object that can be rerun.

Auto-coding by paragraph style and by speaker is strong; a transcript with consistent speaker labels codes itself into speaker cases in seconds.
Framework matrices (cases by themes, with summaries typed into cells) are the framework method built in.
The coding comparison query gives kappa per code per file, which is section 08's table.

Cautions

The Mac version has lagged the Windows version in features for years; check before buying. Project files grow large with media and corrupt occasionally; the backup habit is not optional. The sentiment auto-coding is trained on English and should not be used on Indian-language or translated text as evidence. The word-frequency query's 'stemmed words' and 'synonyms' options are English-only.

NVivo is the tool most Indian universities license and teach. If the institution has it, learn it; the ideas transfer.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

MAXQDA

MAXQDA: four windows and the mixed-methods menu

The screen is four panes: the Document System (files in groups, with variables), the Code System (the tree, with colours and frequencies), the Document Browser (the file being coded, with coding stripes in the margin), and Retrieved Segments (every segment for the activated codes in the activated documents). 'Activation' is the central move: activate documents and codes and the retrieval updates. Memos attach to anything and show as icons in the margin. The Analysis menu holds the Code Matrix Browser, the Code Relations Browser, the Summary Grid and paraphrasing; the Mixed Methods menu holds crosstabs, joint displays and typology tables; MAXMaps draws concept maps from the project's objects.

Document variables are attributes and can be imported from a spreadsheet and exported to one, which is the bridge to section 09.
The Summary Grid (a summary per document per code, written by you) is the framework method's charting step.
MAXDictio adds word frequency and dictionary-based content analysis; the Stats module runs descriptive and inferential statistics on variables without leaving the program.

MAXQDA's gentleness is the reason many methods teachers prefer it: a student sees the documents, the codes and the retrieved evidence at once, and the idea of coding as retrieval is on the screen.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

ATLAS.ti

ATLAS.ti: quotations, networks and the query tool

In ATLAS.ti a selected passage becomes a quotation, an object with an ID that exists whether or not it is coded, can carry a comment, and can be linked to other quotations (supports, contradicts, explains). Codes attach to quotations; code groups and document groups organise them; memos are documents in their own right. Networks draw the links between codes, quotations and memos as a graph. The query tool combines codes with Boolean, semantic (within a hierarchy) and proximity operators (co-occurs, follows, within), which is the most expressive retrieval language of the three.

The quotation model suits work where the passages themselves are analysed and linked (discourse, narrative), less so where codes are counted.
Co-occurrence table, code-document table and Sankey diagrams are one click each.
The Web version runs in a browser with most desktop features and is the easiest route for a distributed team.

ATLAS.ti was first to build AI coding into the product (2023) and its 'Intentional AI coding' codes a document from a stated intent. Section 11 is about whether to use it and what to disclose.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Which One

Choosing among the three: a decision table

If	Lean toward	Because
Your institution licenses one	That one	Support, training, and a supervisor who can open your project
Your team is on Mac	MAXQDA or ATLAS.ti	Feature parity across platforms
The study is mixed methods with a survey	MAXQDA	Variables, crosstabs, joint displays and the Stats module in one place
The study is framework analysis for a policy client	NVivo (framework matrices) or MAXQDA (Summary Grid)	The charting step is built in
The analysis links passages to each other and draws conceptual maps	ATLAS.ti	Quotations and networks
Large document corpora, PDFs, policy texts	ATLAS.ti or MAXQDA	PDF handling and document-level coding
A distributed team with no server	ATLAS.ti Web, or MAXQDA TeamCloud, or NVivo Collaboration Cloud	Simultaneous access; check data-residency terms
No budget	QualCoder, or Taguette	Sections 04 and 05; the coding is the same
A thesis to be examined	Whatever the examiners can open	REFI-QDA export covers the rest

None of the three will make a weak analysis strong or a strong one weak. The advice for a first project is the tool closest to hand; for a team, the one the team agrees on; for a career, learn two.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Licensing

Licences, trials and what to check before paying

All three offer 14- to 30-day free trials with full features; time a trial to a coding sprint, not to an idle month.
Student licences require proof of enrolment and expire; a thesis that runs a year past the licence needs a plan (export to REFI-QDA and continue in QualCoder is one).
Institutional site licences at Indian universities often cover NVivo; ask the library before buying anything.
Subscription versus perpetual: the vendors have moved to subscriptions; a perpetual licence, where offered, is worth it for a multi-year programme.
Cloud and web versions store project data on the vendor's servers, usually outside India; check the consent form and the institution's data policy, and the vendor's data-processing terms, before uploading transcripts.
Free readers (MAXQDA Reader; NVivo's read-only export) let a supervisor or examiner open a project without a licence.

The question to ask a vendor in writing: where is my data stored, who can access it, is it used to train models, and how do I delete it. The answers decide whether the cloud version is usable for interview data at all.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Transcription Tools

Transcription inside and outside the tools, for Indian languages

Route	Languages	Cost	Notes
NVivo Transcription	English-centred; some other languages; weak on Indian languages	Per audio hour, paid	Convenient for English interviews; not the route for Hindi or Bangla
VaniScribe (ImpactMojo)	Hindi, Tamil, Bangla, Marathi, Telugu and others via Sarvam's models; native script or transliteration	A few rupees a minute with your own key	Built for this course's readers; output has timestamps and speaker turns
Bodhan AI Indic-Transcribe (open weights)	The scheduled languages	Free to run locally; hosted API by the minute	Data stays on the machine if run locally; needs a capable computer or a notebook
Whisper and its Indic fine-tunes	Hindi and major languages, variable quality on others	Free locally	Hallucinates in silence; check every transcript
Human transcription (in-house or agency)	Any	Rs 300–1,500 per audio hour depending on level and language	Still the standard for full verbatim and for dialects; quality varies with the agency
Interviewer's own transcription	Any	Their time: 4–8 hours per hour	Slow, and the best first analysis there is

The workflow most South Asian teams now use: machine transcription in the source language, a native-speaker checking pass against the audio, then import. Whatever the route, the transcript is the coder's data and the audio is the truth; keep both.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Migration

Moving between tools, and what to check after

EXPORT a REFI-QDA .qdpx from the source tool at a clean milestone

→

IMPORT into the destination; open a coded file and check the segments match

→

COUNT: code frequencies before and after must agree

→

CHECK memos, attributes, cases; note what was dropped

→

RE-LINK media if audio and transcripts were separated

→

ARCHIVE both project files with the date and the versions used

The exchange works well for text coding, codes, memos and attributes, and less well for tool-specific objects. A count of segments per code before and after is the test; a mismatch of more than a handful means something was lost and the original is the record.

Migrate at a milestone, never mid-coding. Two coders on two tools with one exchange between them is a workable team; two coders swapping the project back and forth weekly is not.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Section Seven

Queries, Matrices and Visuals

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Retrieval

Retrieval is the analysis: reading what was coded, together

The point of coding is to be able to read every passage about one thing at once, across all the interviews, and to notice what they share and where they differ. A coding report or retrieval (all segments for SHIELD) is the primary analytic object: read it as a document, annotate it, write the memo from it. Filtering it by attribute (SHIELD among Bengal members) is the comparison; reading the two filtered retrievals side by side is the finding's evidence.

Retrieve with context: a segment with the question before it and a sentence after. Every tool has a context setting; use it.
Read the whole retrieval before counting anything. The exceptions are usually more informative than the pattern.
Export the retrieval and write in its margins; the writing is the second-cycle coding.
A retrieval that is too long to read (300 segments) is a code that needs splitting.

In each tool

NVivo: open a code, or a Coding query with attribute criteria. MAXQDA: activate documents and codes; Retrieved Segments updates; activate by variable for the attribute filter. ATLAS.ti: the quotation manager filtered by code and document group; or the query tool. QualCoder: Reports → Coding report with file, case or attribute filters. Taguette: click the tag; filter by hand in the export.

The first hour of analysis is reading one retrieval. Everything below is a way of choosing which retrieval to read next.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Matrices

The matrix: codes by attributes, and what its cells mean

Code	Bihar member (n = 12)	Bihar non-member (7)	Bengal member (12)	Bengal non-member (5)
SAVINGS_SHIELD	10 files, 24 seg.	1, 1	8, 14	0, 0
SAVINGS_BARRIER	9, 17	7, 22	10, 19	5, 9
DECISION_STRATEGY	9, 21	3, 5	10, 24	2, 3
FEAR_FORMAL_FINANCE	4, 6	6, 13	5, 8	5, 11
HUSBAND_DEMANDS	11, 26	6, 12	9, 18	4, 8

Illustrative. NVivo's matrix coding query (codes by attribute values), MAXQDA's crosstab, ATLAS.ti's code-document table by document group, QualCoder's attribute report. Each cell: number of files containing the code, and number of segments.

Reading the cells

Files matter more than segments: 10 of 12 Bihar members raised SHIELD; one of seven non-members did. That is a pattern worth a theme. Segment counts say how much a respondent talked, which reflects the interviewer and the respondent's style as much as the topic. FEAR is more common among non-members in both states, which is a second pattern. The cells are pointers into retrievals; the finding is in what the retrievals say.

Never report a matrix cell as a proportion of a population. The sample was purposive; the numbers describe the sample and locate the evidence.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Co-occurrence

Co-occurrence and proximity: what is said together

A co-occurrence table counts the segments (or files) where two codes overlap. SHIELD × HUSBAND_DEMANDS co-occur in 19 segments across 14 files: women describe the shield in the same breath as the demands it shields against. Reading those 19 segments is how the theme's mechanism is stated. Proximity queries (ATLAS.ti's 'follows', NVivo's 'near') find codes within a set distance, which catches cause-and-effect talk where the two ideas are in adjacent sentences.

Co-occurrence depends on segment size (section 03): long segments co-occur with everything. Consistency makes the table meaningful.
A code that co-occurs with nothing is either a topic on its own or a code nobody used precisely.
Heat-map versions (MAXQDA's Code Relations Browser, ATLAS.ti's co-occurrence table with coefficients) show the structure of the codebook at a glance and are useful for revising it.
The co-occurrence coefficient (ATLAS.ti's c-coefficient) is a normalised count, not a correlation; treat it as a ranking.

Co-occurrence is the closest QDA software comes to finding something for you. It finds where to look. What is there is still yours to read.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Text Queries

Word frequency, text search and their limits in Indian languages

Word-frequency queries list the most common words; text search finds every passage with a term and can code them. Both are built for English: stemming, stop words and synonyms are English lists, and Hindi's postpositions, Bangla's inflections and Tamil's agglutination defeat them. A word frequency on a Hindi corpus returns hai, ka, ki, ko, mein and tells you nothing. On the English translations it tells you about the translator's vocabulary.

Use text search to find passages by a known term and its variants (samooh, samuh, group, SHG, dal), then read and code by hand.
Build a project stop-word list for the language; NVivo and MAXQDA allow custom lists.
Word clouds are decoration. A frequency table of a few chosen terms by group can be evidence of emphasis; say what it is.
For real text mining of Indian-language corpora, the tools are outside QDA software (Python with Indic NLP libraries), and the question is usually different.

The commonest misuse: a word cloud on the title slide of a presentation about 36 interviews. It is not analysis and the audience knows.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Framework Matrix

The framework matrix: charting cases by themes

Case	Savings purpose	Shield	Decision strategy	Formal finance
BR_003 (member, migrant husband)	School fees; a buffalo	Strong: 'the money is the group's until I take it'; husband's family cannot ask	Times requests for when remittance arrives; group president as ally	Never entered the bank; group leader deposits
BR_007 (member, husband present)	Emergencies	Weak: husband knows the schedule and asks	Concedes small, holds large	Fear of forms; went once with the group
BR_011 (non-member)	'Nothing to save'	—	None described; husband decides	Distrust from a relative's experience
WB_015 (member, migrant husband)	Daughter's marriage; a phone	Strong; explicit comparison with pre-membership	Uses group meeting days as cover for market visits	Has an account; uses it with the group

Illustrative, four of 36 rows. Cells are the analyst's summaries of what that case said under that theme, with a phrase quoted, written from the retrievals. NVivo's framework matrix and MAXQDA's Summary Grid hold it inside the project, linked to the segments; a spreadsheet does it outside.

Why it works

Reading down a column shows the range of a theme across cases; reading across a row shows one woman's account as a whole; sorting rows by an attribute (migrant husband) shows whether the column changes with it. Gale and colleagues (2013) built the method for applied health research with teams and deadlines, and it is the most transparent qualitative analysis a policy reader can be shown.

Charting 36 cases by eight themes is about three days. It is the analysis, not preparation for it.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Visuals

Visuals that earn their place, and those that do not

Visual	Tool	Use when	Avoid when
Code tree (the codebook as a diagram)	All	Showing the structure of the analysis in the methods	It is decoration for a list that reads better as a table
Matrix heat map (codes by attributes)	NVivo, MAXQDA, ATLAS.ti	Showing where a theme concentrates, as a pointer	The reader will read cell shading as proportions of a population
Co-occurrence network	ATLAS.ti networks; MAXMaps; NVivo maps	Showing how themes relate, with the links labelled and argued	It is a hairball of every code linked to every other
Concept map from memos	MAXMaps, ATLAS.ti networks	The theory of change that emerged; a figure for the discussion	It restates the code tree
Comparison diagram (two cases)	NVivo	Showing what two contrasting cases share and do not	Rarely; it is a two-row framework matrix
Hierarchy chart, sunburst	NVivo, QualCoder	Showing coverage of codes as a diagnostic during coding	In a report; it is a pie chart of segment counts
Word cloud	All	Never in a report	Always
Joint display (qual and quant side by side)	MAXQDA; by hand	Mixed methods, section 09	—

The best qualitative figure in most reports is a table of quotations by theme and case, or a framework matrix. A diagram is justified when the relationships between themes are the finding and the diagram is drawn from an argument, not from the software's auto-layout.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

From Query to Claim

From a query to a claim: the chain of evidence

PATTERN noticed in a memo or a matrix cell

→

RETRIEVAL read in full, with context; exceptions listed

→

COMPARISON across attributes; the counterexamples explained or admitted

→

CLAIM written as a sentence with its scope ('among members with migrant husbands')

→

EVIDENCE: three quotations that show the range, not three that agree

→

CHECK: a search for disconfirming cases; a second reader

The software makes the first three steps fast and leaves the last three where they were. A claim that arrived by this chain can be traced back through the project: the memo, the query, the retrieval, the segments. That traceability is what section 10 means by an audit trail, and the project file is it.

Illustrative claim: 'Among members whose husbands migrate, group savings were described as a protected fund that relatives could not claim; three of the eighteen described the opposite, a husband who demanded the group's money, and all three were in groups without a written withdrawal rule.' Scope, pattern, exceptions, and a mechanism to check.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Query Checklist

Queries and matrices: the checklist

Practice	Done
Every retrieval read in full before any count is reported
Matrices report files (cases) first, segments second, and the sample size per column
No matrix cell presented as a population proportion
Co-occurrence read as a pointer; the overlapping segments read
Text search and word frequency used to find, not to code or to conclude
Framework matrix or equivalent charted for the main themes
Every claim traceable: memo, query, retrieval, segments
Disconfirming cases searched for and reported
Visuals limited to those that carry an argument

Nine rows. The third is the one a quantitative reader will test you on, and the seventh is the one a qualitative reviewer will.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Section Eight

Coding in Teams

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Why Teams

Team coding: for capacity, for languages, and for challenge

Thirty-six interviews in two languages are more than one person can code in the time an evaluation allows, and more languages than one person usually reads. A team also brings a second reading: the field researcher who heard the interview, the analyst who did not, the supervisor who knows the literature. The software's job is to let them code separately and bring the work together; the method's job is to decide what 'together' means.

One codebook, owned by one person, versioned; every coder works from the current version.
Coding meetings weekly: new inductive codes proposed and decided; definitions revised; disagreements discussed on the segments.
Each coder's work identified (coder name in the software) so that comparison and attribution are possible.
A merging protocol (next slides) agreed before anyone codes.

The field researcher as coder

The person who did the interview codes it best and worst: best because they heard the tone and know the context, worst because they know what they meant to ask. A common arrangement: the interviewer codes their own interviews first cycle, a second person codes a sample, and the analyst does second-cycle grouping across all. The interviewer's memos are data of their own.

Budget the coding meetings. An hour a week for the coding period, with the segments on the screen.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Agreement

Inter-coder agreement: when to measure it, and what the number means

Two coders code the same transcripts independently with the same codebook; the software compares segment by segment and reports agreement, usually as percentage agreement and Cohen's kappa (chance-corrected) per code, sometimes Krippendorff's alpha. Kappa above 0.8 is conventionally 'strong', 0.6 to 0.8 'moderate to substantial'. O'Connor and Joffe (International Journal of Qualitative Methods 2020, 19) review the debate and give practical guidance; Campbell and colleagues (Sociological Methods & Research 2013, 42:294) show how much segment boundaries ('unitisation') drive the number.

Measure it when the method is codebook-based (content analysis, framework, template analysis) and the codes are meant to be applied consistently. Report it per code, not as one average.
Do not measure it for reflexive thematic analysis, where codes are the researcher's interpretation and agreement is not the goal (Braun and Clarke are explicit); describe the collaborative process instead.
Low kappa on a code is a finding about the definition, and the fix is in the codebook, then a recode of the shared sample.
Agreement on a sample of files (10–20%) is standard; agreement on everything is a different design.

'Inter-coder reliability was established (κ = 0.82)' as one number for forty codes is not informative. Per code, with the range, and what was done about the low ones, is.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Worked Agreement

A worked agreement table, and what changed after it

Code	% agreement	κ round 1	Action	κ round 2
SAVINGS_PURPOSE	94	0.84	None	—
SAVINGS_SHIELD	86	0.58	Redefined: requires explicit reference to the money being inaccessible to others; added two examples and a do-not-use	0.79
SAVINGS_BARRIER	91	0.76	None	—
DECISION_STRATEGY	83	0.52	Split into STRATEGY_TIMING, STRATEGY_ALLY, STRATEGY_WITHHOLD	0.71, 0.77, 0.74
FEAR_FORMAL_FINANCE	92	0.74	Boundary with BANK_MISTREATMENT clarified	0.81
HUSBAND_DEMANDS	90	0.71	None	—

Illustrative; eight shared files, two coders, QualCoder's coder comparison (NVivo's coding comparison query and MAXQDA's intercoder agreement give the same table). Round 2 was on four new shared files after the codebook revision.

What the table did

Two codes were unclear and became clearer, one by redefinition and one by splitting; the splitting turned out to be analytically useful, because STRATEGY_WITHHOLD became part of the shield theme. The table goes in the appendix with the two codebook versions. Percentage agreement is high everywhere because most text is uncoded and both coders agree it is; kappa corrects for that.

Agreement measured once at the start and never again is a ritual. Measured at the start, acted on, and re-measured, it is a method.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Merging

Merging coders' work: the three arrangements

Arrangement	How	Tools	Risks
Divide the files	Each coder codes different files in a copy of the master; the master imports each copy (NVivo import project; MAXQDA merge; ATLAS.ti merge; QualCoder import coding)	All four	Codebook drift between coders; new codes created in two places under different names; merge conflicts on files coded twice
Simultaneous on a shared project	A server or cloud project everyone opens (NVivo Collaboration Cloud, MAXQDA TeamCloud, ATLAS.ti Web, hosted Taguette)	Cloud versions; Taguette	Data residency and consent; subscription cost; connectivity
Sequential	One project file passed from coder to coder with a lock (only one person edits at a time)	Any	Slow; someone always has it; the lock is broken by email

The first is the usual arrangement for desktop tools. Its risk is managed by the codebook owner: coders may not create codes, only propose them in a memo; the owner adds them to the master; coders reimport the codebook weekly. Merging then goes cleanly, and the coder comparison reports work because the same codes have the same identities.

Test the merge with two small copies before the real one. A failed merge of three weeks' coding is recoverable only from the copies, so keep them.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Across Languages

Teams across languages: keeping one analysis from two corpora

The Hindi coder and the Bangla coder each code their own language with a shared, bilingual codebook whose definitions carry the source-language terms that anchor each code. Agreement is checked on the English translations of a shared sample, which measures whether the codebook travels, not whether the two coders read Hindi and Bangla the same way. Theme memos are written in English with quotations in both languages. The analyst reads across both coders' retrievals.

Weekly meetings compare how a code looks in each language; a code that means something different in Bangla is a finding, and may need splitting by language.
The in vivo parent holds each language's key terms; the report glosses them.
Where a supervisor reads only English, the translation workflow (section 02, C) with the source visible lets the coder show them the original when a quotation is disputed.

A theme found only in one language's corpus is either a real difference between the states or an artefact of the coder. Section 10's reflexivity statement says which you think, and why.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Versioning

Versions, backups and the project log

What	When	How
Project file backup	Daily during coding; automatic where the tool offers it	A dated copy on a second drive or institutional storage; not only the laptop
Milestone version	After the codebook revision; after the merge; before second-cycle coding; at submission	Save-as with a date; keep them all
Codebook version	At every revision	Export the codebook with date and a change note; keep the sequence for the appendix
REFI-QDA export	At milestones	The tool-independent archive
Project log	Continuous	A dated memo of decisions: what changed, why, who decided; NVivo's event log and QualCoder's action log supplement it
Coder copies	Until the merge is verified	Each coder's project before merging, kept

A qualitative project with a log, versioned codebooks and milestone files can be audited by an examiner in an afternoon; one with a single project file cannot be audited at all. The difference is habit, not software.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Team Checklist

Coding in teams: the checklist

Practice	Done
One codebook owner; coders propose, owner decides
Coder names set in the software before coding
Merging arrangement agreed and tested before coding starts
Shared sample (10–20% of files) coded independently by two coders
Agreement computed per code, acted on, and re-measured; reported in the appendix
Weekly coding meeting with segments on screen; decisions logged
Bilingual codebook with source-language anchors where languages differ
Backups daily; milestone versions kept; coder copies kept until the merge is verified
Cloud use consistent with consent and data policy

Nine rows, and the one that fails most often is the third: teams merge for the first time on the day it matters.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Section Nine

Mixed Methods

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Integration

Integration: where the qualitative and the survey meet

Most development evaluations collect both a survey and interviews and report them in separate chapters that never touch. Integration is the work of making them answer each other: the survey says the programme raised decision-making by 0.9 points; the interviews say how, for whom, and what 0.9 points means in a household. Fetters, Curry and Creswell (Health Services Research 2013, 48:2134) set out the ways: at the design (one strand informs the other's sampling or instrument), the methods (connecting, building, merging), and the interpretation (joint displays, narrative weaving).

Connecting: interviewees sampled from the survey, so attributes are survey variables and the qualitative cases sit inside the quantitative sample.
Building: interview findings become survey items in the next round (the shield theme becomes a question about who can access group savings).
Merging: the two sets of results brought together in a joint display and interpreted as one.

The software's contribution is attributes. If the 36 interviewees are also survey respondents with their survey values as attributes, every matrix in section 07 is a mixed-methods table.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Joint Displays

The joint display: one table, both kinds of evidence

Survey result (n = 640)	Qualitative finding (36 interviews)	Integrated reading
Members score 0.9 points higher on decision-making (95% CI 0.4 to 1.3)	Members describe group savings as a protected fund; decisions over large purchases are timed to its availability	The effect is concentrated in large-purchase decisions and works through control over a lump sum
Effect larger where husband is a migrant (interaction, p = .03)	SHIELD theme present in 15 of 18 migrant-husband cases, 7 of 18 others; remittance lumps attract relatives' claims, the group's fund does not	The mechanism is protection from claims, which matters most where lump sums arrive
No effect on 'own income' (2 points, CI −4 to 8)	Few members describe new income; several describe the group as changing control, not earnings	The programme changes control over money more than the amount; the survey's income item was the wrong outcome for the mechanism
Non-members report higher fear of banks (62% vs 41%)	Non-members' fear draws on relatives' stories; members' fear is reduced by going 'with the group'	Exposure through the group reduces avoidance; a bank-linkage component would build on it

Illustrative. Guetterman, Fetters and Creswell (Annals of Family Medicine 2015, 13:554) describe the joint display; MAXQDA's Mixed Methods menu builds several kinds from variables and codes; the rest of the tools leave it to a table in Word, which is fine. The third column is the integration and it is the paragraph a policy reader remembers.

The third row is the useful kind: the strands disagree, and the disagreement explains the survey's null.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Open-Ended Items

Survey open-ends: coding hundreds of short answers

A survey's 'why did you stop attending?' with 480 free-text answers is qualitative data at scale. Import the spreadsheet as a dataset (NVivo), a survey import (MAXQDA), a document table (ATLAS.ti) or a survey (QualCoder's Project → Import survey), so each respondent is a case with attributes and each answer a codeable cell. Auto-code by question; then code the answers with a short codebook (ten to fifteen codes); then export codes by case as a binary matrix back to the survey file.

Answers are short; segments are whole answers; multiple codes per answer are normal.
Here counts are legitimate results: the sample is the survey's sample, and '31% cited distance' is a proportion with a denominator. Report it with the n and the multiple-response note.
Two coders on a 10% sample with kappa, as in section 08; short answers code fast and disagree often.
The exported code matrix (case × code, 0/1) joins the survey data by ID and becomes variables for the quantitative analysis: 'cited distance' as a predictor of re-enrolment.

Sandelowski, Voils and Knafl (Journal of Mixed Methods Research 2009, 3:208) on 'quantitizing': what it is legitimate to count and what the count then means. Open-ends are the clearest case where it is.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Sampling

Sampling interviewees from the survey: attributes by design

When the qualitative sample is drawn from the survey (a nested or connected design), the interviewee's survey record is the attribute sheet: her arm, her baseline and endline scores, her household characteristics. Purposive selection can then be principled: maximum variation on the outcome (high gainers, no change, decliners), extreme cases, or stratified by the moderator the survey found (migrant husband). The interviews then explain the survey's numbers with the numbers attached.

Import the survey extract as the attribute file; the join key is the participant ID, which consent must have allowed to be linked.
A matrix of themes by outcome group (gainers, no change, decliners) is the mixed-methods analysis of mechanism.
Deviant cases (a member who declined) are selected on purpose and are the most informative interviews in the study.

Illustrative

Twelve members who gained more than 2 points, twelve who did not change, twelve non-members, balanced across states and migrant-husband status. The SHIELD theme appears in eleven of the twelve gainers and four of the twelve no-change members; the four were in groups without a withdrawal rule. That is a mechanism hypothesis the survey could not have produced and the next survey can test.

The interview guide changes when the survey results are known. Sequential designs plan for that; parallel ones cannot.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Quantitising

Counting qualitative codes: when it is evidence and when it is not

Count	Legitimate?	Because
'31% of 480 survey respondents cited distance'	Yes, as a proportion	The sample is the survey's; every respondent answered the question; the denominator is known
'15 of 18 interviewees with migrant husbands described a shield'	Yes, as a description of the purposive sample and a pointer	The denominator is the interview sample, which was chosen, not drawn
'42% of women in the district feel the group protects their savings'	No	A purposive sample of 36 cannot support a population proportion
'The theme of fear was mentioned 94 times'	Rarely	Segment counts reflect talkativeness and segment size; say files, not segments, if anything
'Members were more likely than non-members to mention X (p = .04, Fisher's exact on 36)'	Not as a test	The interviews were not a random sample and the codes were not pre-specified as outcomes; report the pattern and its scope
'Coded open-ends as variables, then regressed re-enrolment on them (n = 480)'	Yes	Quantitised survey data with a defined sample and pre-specified codes

The rule is the denominator. Where the denominator is a defined sample that answered a defined question, a count is a statistic. Where it is a purposive set of conversations, a count locates evidence and describes the set, and the sentence should make that plain.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Mixed Tools

The tools for mixed methods: what each does natively

Need	MAXQDA	NVivo	ATLAS.ti	QualCoder	Outside the tool
Import survey variables as attributes	Yes (variables)	Yes (classifications)	Yes (document groups; attributes)	Yes (attributes from CSV)	—
Import open-ends as codeable cells	Survey import	Dataset import	Survey import	Import survey	—
Crosstab codes by variables	Crosstab; Code Matrix Browser	Matrix coding query	Code-document table	Attribute report	Pivot table on the export
Joint display	Mixed Methods menu builds several	By hand	By hand	By hand	A table in Word
Export code matrix to statistics	Yes; the Stats module runs it in place	Yes (export to Excel/SPSS)	Yes	Yes (CSV)	jamovi or R
Typology from codes and variables	Typology table; Similarity analysis	By hand	By hand	By hand	Cluster analysis in R

MAXQDA is the tool built with mixed methods in mind and the one to choose when integration is the study's centre. Every other tool gets there with an export and a spreadsheet, which is a morning's work.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Mixed Checklist

Mixed methods: the checklist

Practice	Done
Integration point stated in the design: connecting, building, merging, or several
Interviewees linked to survey records where consent allows; attributes imported
Purposive qualitative sampling justified from the survey (variation, extremes, moderators)
Matrices of themes by survey-defined groups produced and read
A joint display drafted, with the integrated reading in the third column
Disagreements between strands reported and interpreted, not suppressed
Counts labelled by their denominator: proportion, description, or pointer
Open-ends coded with a short codebook, agreement checked, and exported as variables
The report weaves the strands rather than stacking chapters

Nine rows. Mixed Methods 101 in this series covers the designs; this list is what the software has to support to make them work.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Section Ten

Rigour and Reporting

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Trustworthiness

Trustworthiness: the four criteria and what the software contributes to each

Criterion (Lincoln and Guba 1985)	Question	Practices	What the project file shows
Credibility	Is the account believable to those who lived it and to careful readers?	Prolonged engagement; triangulation across sources; member checking; negative case analysis; peer debriefing	The retrievals behind every claim; the disconfirming cases coded and memoed
Transferability	Could a reader judge whether this applies elsewhere?	Thick description of setting and sample; purposive sampling stated	Attributes; the sampling memo; the case descriptions
Dependability	Was the process consistent and could it be followed?	Audit trail; codebook versions; team procedures; agreement checks	The log, the codebook versions, the comparison reports, the milestone files
Confirmability	Are the findings grounded in the data rather than the researcher's preferences?	Reflexivity; the chain from data to claim; a second reader	Reflexive memos; the query and retrieval behind each theme; a second coder's work

The project file is the audit trail if it was kept as sections 03 and 08 describe. A reviewer or examiner given the file, the codebook versions and the log can follow every step, which is what 'dependability' meant before software and what the software makes routine.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Audit Trail

The audit trail: what to keep, and what an examiner asks for

The raw data: recordings and checked transcripts, in the languages spoken, with the checking recorded.
The project file at milestones, and the REFI-QDA export.
Every codebook version with dates and reasons.
The log of decisions: sampling, transcription, translation, coding, merging, theme building.
The memos: methods, theme, case, reflexive.
The agreement reports, with what was done about low agreement.
The retrievals and matrices behind each reported theme.
The quotation register: every quotation in the report, its source ID, its original language, and the consent status.

What examiners actually ask

'Show me how you got from the interviews to theme three.' The answer is a memo dated before the theme was named, a matrix, a retrieval, and three cases including one that did not fit. 'How did you handle the Bangla interviews?' The translation workflow memo and the bilingual codebook. 'Who coded what?' The coder names and the comparison table. An examiner who gets these answers from the project itself stops asking.

Nothing in the list is extra work if the project was run as this course describes. It is the project.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Reflexivity

Reflexivity: the researcher in the analysis, written down

Who did the interviews, in what language, with what relationship to the programme and the participants; what the analyst expected to find; how the team's positions (caste, class, gender, urban, employed by the implementer) shaped what was asked and what was heard. A reflexivity statement is not a confession; it is the information a reader needs to weigh the account. The reflexive memos written during coding are its source, and the software keeps them dated.

Write the first reflexive memo before the first interview: expectations, worries, what would surprise you.
Write one when something does surprise you, and one when a theme starts to feel settled.
Name the constraints: the evaluation was commissioned by the implementer; the interviewers were its field staff; participants knew that.
In the report: a paragraph in the methods, and a sentence where a finding could plausibly reflect the position rather than the data.

Illustrative: 'The Bihar interviews were conducted by the programme's own community mobilisers, known to the participants; accounts of the group's benefits may be more positive than they would be to a stranger, and the three most critical accounts came from the Bengal interviews conducted by an independent researcher.' That sentence changes how the findings are read, and it belongs in the report.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Checks

Member checking, negative cases, and peer debriefing: how each is done

Member checking

Taking findings back to participants: a summary read aloud in a group meeting, or the framework row for one woman shown to her. Birt and colleagues (Qualitative Health Research 2016, 26:1802) distinguish checking transcripts (low value) from checking synthesised findings (higher, and harder). Record what participants said in a memo; a disagreement is data.

Negative case analysis

Search the retrievals for cases that contradict the theme; code them as such; explain them or revise the theme. The three women whose husbands claimed the group's money are the shield theme's negative cases and its boundary.

Peer debriefing and second reading

A colleague outside the team reads the theme memos and a sample of retrievals and asks whether the claims follow. In the software: give them the project or a REFI-QDA export and the codebook; ask them to code two transcripts blind and compare. Their questions are logged and answered in the memo.

None of these is a guarantee. Each is a check a careful reader would expect, and each is a paragraph in the methods.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Thick Description

Transferability: the description that lets a reader judge

A qualitative finding travels only as far as a reader can see the setting it came from. Thick description means enough about the villages, the groups, the households, the programme's delivery and the interviews themselves for a reader in Odisha or Sindh to judge whether the shield mechanism would apply there. The attributes, the sampling memo and the case descriptions in the project are its raw material; a paragraph per site and a table of cases with their attributes are its form in the report.

Describe the sample by the attributes that the findings turned on (migration status, group rules, distance to a bank), not only the demographic ones.
Describe the programme as delivered in these sites, which is rarely the programme as designed.
Describe the interviews: where, who was present, how long, in what language, whether the husband came in.
State what the sample cannot speak to, by group and by setting.

Illustrative: 'Bihar groups met fortnightly in a member's courtyard with a written withdrawal rule in eleven of twelve; Bengal groups met monthly at the panchayat office and four of twelve had no written rule. The three cases in which a husband took the group's money were all in groups without one.' That paragraph is the transferability argument and it came from three attributes.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

COREQ and SRQR

Reporting standards: COREQ and SRQR, and how the software answers their items

COREQ domain (Tong et al., Int J Qual Health Care 2007, 19:349; 32 items)	Items the software helps answer
Research team and reflexivity (items 1–8)	Coder names; reflexive memos; interviewer attribute
Study design (9–23): theory, sampling, setting, data collection	Sampling memo; attributes; transcription and translation memos; saturation evidence
Analysis and findings (24–32): coders, codebook, derivation of themes, software, participant checking, quotations, consistency, clarity of themes	Codebook versions; agreement table; the chain from retrieval to claim; software and version; quotation register

SRQR (O'Brien et al., Academic Medicine 2014, 89:1245; 21 items)	Note
Title, abstract, problem, purpose	As any paper
Approach and paradigm; researcher characteristics and reflexivity	Method memo; reflexivity statement
Context; sampling; ethics; data collection methods and instruments; units of study; data processing; data analysis; trustworthiness techniques	The memos and log, in order
Synthesis and interpretation; links to empirical data	Themes with quotations traced to sources
Integration with prior work; limitations; conflicts of interest; funding	The discussion

COREQ is written for interviews and focus groups in health; SRQR is general. Health journals ask for COREQ; most others accept either. Fill the checklist against the draft; every unanswerable item is a gap.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Quotations

Quotations in the report: choosing, presenting and protecting

Choose quotations that show the range of a theme, including its edges, not three that say the same thing well.
Introduce each: what it shows; then the quotation; then who (pseudonym or ID, attributes that matter: 'member, Bihar, husband a migrant'); then a sentence on how typical it was.
Give the original where the wording matters, with the translation; mark the translator's choices ('izzat' kept, glossed).
Keep the interviewer's question where it shaped the answer.
Never alter meaning by cutting; mark cuts with ellipses and insertions with brackets.
Check consent for quotation per participant (the attribute from section 02); a woman who agreed to interview but not to quotation is paraphrased.

Anonymisation in small places

In a village of 400 households, 'the SHG president whose husband works in Surat' identifies one woman. Generalise attributes in quotations to the level that protects ('a group office-bearer'), change the district name where the story is distinctive, and let the quotation register hold the real source for the audit. The software's ID and the report's pseudonym are linked only in the register, which is stored with the consent forms.

A quotation register is a spreadsheet with one row per quotation used: report page, pseudonym, source file, segment, language, consent status. It takes an hour and it is what a data-protection query is answered from.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Methods Paragraph

The methods paragraph, assembled from the project

Illustrative

'We conducted 36 semi-structured interviews (median 52 minutes) with women in Bihar (n = 19) and West Bengal (n = 17), purposively sampled from the evaluation survey to vary by SHG membership, change in decision-making score and husband's migration status (Appendix B). Interviews were conducted in Hindi or Bangla by two field researchers (one employed by the implementing organisation) and one independent researcher, audio-recorded with consent, machine-transcribed in the source language (VaniScribe, Sarvam Saaras v3) and checked in full against the audio by a native speaker. Analysis followed the framework method (Gale et al. 2013) in QualCoder 3.6 with source-language transcripts and parallel English translations. An a priori codebook of 31 codes from the programme's theory of change was revised after the first five transcripts and again after an agreement check on eight transcripts coded independently by two coders (Cohen's κ 0.52 to 0.84 by code before revision, 0.71 to 0.84 after; Appendix C); the final codebook of 46 codes in nine categories is in Appendix D with its three versions. Themes were built through weekly team meetings, memos and a framework matrix of 36 cases by eight themes; negative cases were sought for each theme. Findings were presented to two participant groups, whose responses are recorded in Appendix E. The project file (REFI-QDA export) and codebook are deposited at [repository], with transcripts withheld under the consent terms.'

Every number and every claim in the paragraph is a file, a memo or a report in the project. That is what 'systematic' means and it is what COREQ and SRQR ask for.

One paragraph, five appendices, and a project file. Most qualitative sections in evaluation reports have the paragraph without the rest, which is why they are not believed.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Archiving

Archiving qualitative data: what can be deposited and where

Object	Deposit?	Where	Conditions
Codebook with versions	Yes, always	OSF, Zenodo, the institution's repository, with the paper	None
Project file (REFI-QDA) with coding and memos, transcripts removed	Usually	As above	Memos may contain identifying detail; review them
De-identified transcripts	Where consent allows	Qualitative Data Repository (Syracuse), UK Data Service, ICPSR; with access controls	Consent for archiving; de-identification to a documented standard; often restricted access
Audio	Rarely	Restricted repositories only	Voices identify; consent rarely covers it
Quotation register	No	With the consent forms, institutional storage	The link between pseudonyms and people
Interview guide, consent forms (blank), sampling memo	Yes	With the paper	None

Depositing the codebook and the coded project without transcripts lets a reader see the analysis's structure while protecting participants, and it is more than most qualitative papers offer. Funders increasingly require a data availability statement for qualitative work; this table is the answer to it.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Section Eleven

AI, Choosing a Tool, and Practice

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

AI Features

The AI features: what they do, tool by tool

Tool	Feature	What it does	Where the data go
ATLAS.ti	Intentional AI coding; AI summaries; conversational AI	Codes a document against a stated intent; summarises documents or code groups; answers questions about the project	To OpenAI via Azure, under ATLAS.ti's terms; opt-in per project
MAXQDA	AI Assist	Summarises segments and documents; suggests codes and subcodes; chats with a document; paraphrases	To OpenAI under VERBI's terms; opt-in; can be disabled institution-wide
NVivo	AI Assistant	Summarises files and coded content; suggests codes; explains	To Lumivero's provider; opt-in
QualCoder	Optional AI assistance	Semantic search and code suggestions via an API key the user supplies (OpenAI or compatible)	To whichever provider the key belongs to; off by default
Taguette	None	—	—
Outside the tools	General LLMs (ChatGPT, Claude, Gemini), Indic models	Anything, on pasted text	To the provider; usually outside India; terms vary on training use

Each feature sends transcript text to an external model. That is the first fact about every one of them, and it decides whether it can be used at all under the consent given (section 02) and the institution's data policy.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

AI Risks

Five risks, and the practices that contain them

Risk	What happens	Practice
Data leaves the machine	Transcripts of vulnerable women's accounts go to a server abroad, under terms the participants did not see	No AI features without consent that names external processing; prefer local models (Bodhan's open weights on a lab machine) where consent is narrow; strip identifiers first regardless
Hallucinated summaries	A summary states something no participant said; a suggested code has no segment	Every AI summary is checked against the retrieval; every AI code is read before it is kept; the model's output is a draft, never evidence
Homogenisation	The model codes in the register of English-language psychology; Indian-language nuance and in vivo phrasing disappear	Code in the source language by hand for the analytic codes; use AI, if at all, for structural and descriptive first-pass codes
Weak Indian-language performance	Hindi and Bangla transcripts summarised badly; code-switching mishandled	Test on three transcripts with a bilingual reader before any use; prefer models built for Indian languages
Undisclosed use	Reviewers and readers assume human coding; the paper's claims about method are false	Disclose the tool, version, what it was used for, on what, and how its output was validated, in the methods (COPE and most journals now require it)

A defensible use, illustrative: AI summaries of each transcript as a reading aid, on de-identified English translations, with consent covering external processing, checked by the interviewer, and disclosed. An indefensible one: AI coding of Hindi transcripts as the analysis, reported as thematic analysis.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

AI Validation

If you use AI coding: validate it like a second coder, and report the table

Treat the model as a coder with unknown reliability. Give it the codebook; have it code the same eight files a human coded; run the coder comparison; report kappa per code. Read the disagreements. Where the model is reliable (often descriptive and topic codes), it may be used for a first pass that a human reviews; where it is not (interpretive codes, in vivo, anything requiring the context of the interview), it is not used. The validation table goes in the appendix beside the human one.

Code	Human-human κ	Human-model κ	Decision
SAVINGS_PURPOSE	0.84	0.77	Model first pass, human review
SAVINGS_SHIELD	0.79	0.41	Human only
DECISION_STRATEGY_TIMING	0.71	0.38	Human only
FEAR_FORMAL_FINANCE	0.81	0.69	Model first pass, human review

Illustrative. The pattern is typical: the model finds topics and misses interpretation, especially the codes that became the study's findings.
Cost the review time honestly. Reviewing a model's coding of a transcript takes half the time of coding it, not a tenth.
Re-validate when the codebook changes or the model version does.
State in the methods exactly which codes were model-assisted and how.

The question a referee will ask: 'What did the model code that a human would not have, and what did it miss?' Have the answer from the comparison, not from impression.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Disclosure

Disclosing AI and software use: the sentences

Software

'Coding and retrieval were carried out in QualCoder 3.6 (Curtain, 2025); the project was exported in REFI-QDA format for archiving.' Name the tool and version; the version changes defaults and features.

Transcription

'Transcripts were produced by automatic speech recognition (VaniScribe on Sarvam Saaras v3) in Hindi and Bangla and checked in full against the audio by native-speaking members of the team; participants consented to third-party transcription.'

AI assistance

'MAXQDA's AI Assist (OpenAI GPT-4o via VERBI, May 2026) was used to generate document summaries as reading aids on de-identified English translations, under consent covering external processing; summaries were not used as data and all coding was by the authors. No AI tool was used to generate or edit the text of this report.'

Three sentences, one per use, each naming the tool, the version, the data it saw, the consent basis, and the limit of its role. Where no AI was used, one sentence says so; journals now ask, and silence reads as concealment. Academic Writing 101, section 11, covers the publisher policies.

Write the disclosure before the analysis. If it cannot be written honestly, the use should not happen.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Choosing

Choosing a tool: the decision table

Situation	Choose	Because
First project; under twenty text sources; one analyst	Taguette	Ten minutes to start; everything exports; nothing to unlearn
Any project with audio, attributes or a second coder; no budget	QualCoder	The full workflow, free, offline, REFI-QDA
Institution licenses NVivo, MAXQDA or ATLAS.ti	That one	Support, supervisors, and site-licensed collaboration
Mixed methods with a survey at the centre	MAXQDA	Variables, crosstabs, joint displays, Stats
Framework analysis for a policy client with a deadline	NVivo or MAXQDA	Framework matrix and Summary Grid
Discourse or narrative work linking passages	ATLAS.ti	Quotations and networks
Distributed team, no server, cloud permitted by consent	ATLAS.ti Web, MAXQDA TeamCloud or NVivo Collaboration Cloud	Simultaneous access
Distributed team, data must stay in the institution	Self-hosted Taguette, or QualCoder with the divide-and-merge arrangement	No external storage
Teaching a methods class with no budget	Taguette, then QualCoder	Section 04's sequence
Long programme, staff turnover, recurring rounds	Any REFI-QDA tool, with the codebook as the persistent asset	The tool may change; the codebook and the archive should not

The last row is the general principle. Invest in the codebook, the memos and the archive; the software is where they live this year.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Worked Project

The full worked project: from recordings to a deposited archive

Week	Work	Tool	Section
1	Consent review; transcription in source languages; native-speaker checking; pseudonymisation; attribute sheet	VaniScribe; Word; a spreadsheet	02
2	Project setup; a priori codebook with definitions; first memo; five transcripts coded; codebook revised; five recoded	QualCoder	03
3–5	Coding by two coders in their languages; weekly meetings; inductive codes proposed and adopted; agreement on eight files; revision; second agreement check	QualCoder; coder comparison	03, 08
6	Merge; retrievals read; matrices by state, membership and migration; co-occurrence; theme memos	QualCoder reports	07
7	Framework matrix charted (36 × 8); negative cases; joint display with survey results; two member-checking meetings	QualCoder export; Word; the survey's jamovi file	07, 09, 10
8	Findings written with quotations from the register; methods paragraph; appendices (codebook versions, agreement, sampling, member checking, code-creation curve); REFI-QDA export deposited	Word; OSF	10

Illustrative, for the 36-interview SHG study. Eight weeks for a two-coder team, of which the first is data preparation and the last is writing. The tools cost nothing; the transcription cost a few thousand rupees; the checking cost a week of a native speaker's time and was the best-spent week.

What the archive holds

The codebook in three versions; the REFI-QDA project without transcripts; the agreement tables; the sampling and translation memos; the interview guide in three languages; the framework matrix; the joint display. Enough for an examiner to follow every step, and nothing that identifies a participant.

The same project in NVivo or MAXQDA would differ in the menus and in nothing that matters to the findings.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Pitfalls

The dozen errors QDA software makes easy

Error	Section	Fix in one line
Translating everything to English and coding the translation, without saying so	02	Choose a workflow; state it; keep the source
Machine transcripts uncorrected	02	A native-speaker check against the audio, recorded
Codes without definitions	03	Fill the memo field for every code
Two hundred codes and no theme	03	Merge by definition; write theme memos
Auto-coding or text search reported as coding	03, 07	Use to find; code by reading
Matrix cells reported as population proportions	07	Files first, denominator stated, purposive sample named
Word clouds as analysis	07	Delete them
Teams merging for the first time on the deadline	08	Test the merge with copies in week 2
One kappa for forty codes	08	Per code, with the action taken
Qualitative and survey chapters that never meet	09	A joint display with an integrated column
Transcripts uploaded to an AI feature without consent for external processing	11	Check consent; strip identifiers; prefer local models; disclose
No audit trail beyond the project file	10	Codebook versions, log, memos, milestone files, REFI-QDA export

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Checklist

Before you submit: the qualitative software checklist

Data and process	Done
Consent covers recording, transcription, storage, quotation and any external processing
Transcripts in the source language, checked, pseudonymised
Translation workflow stated
Attributes imported; sample described
Codebook with definitions, versioned, in the appendix
Agreement checked per code where the method calls for it; reported
Memos and log kept; milestone files and REFI-QDA export saved

Analysis and reporting	Done
Every theme traceable to retrievals, matrices and cases
Negative cases sought and reported
Counts labelled by denominator; no purposive-sample proportions
Framework matrix or equivalent for the main themes
Joint display where a survey exists
Quotations chosen for range, introduced, sourced, consented, registered
COREQ or SRQR checklist completed
Software, versions and any AI use disclosed; codebook and project deposited

Fifteen lines. The translation workflow and the AI disclosure are the two most often missing from South Asian qualitative reports.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Reading List

Where to go next

Resource	What it covers	Notes
Saldaña, The Coding Manual for Qualitative Researchers (4th ed., Sage, 2021)	Coding methods, first and second cycle, memos	The companion to section 03
Miles, Huberman and Saldaña, Qualitative Data Analysis: A Methods Sourcebook (4th ed., Sage, 2020)	Displays, matrices, drawing and verifying conclusions	Section 07
Gale et al., BMC Med Res Methodol 2013, 13:117	The framework method, step by step	Sections 03 and 07
Braun and Clarke, Thematic Analysis: A Practical Guide (Sage, 2021)	Reflexive thematic analysis and what it is not	Why agreement is not always the goal
O'Connor and Joffe, Int J Qual Methods 2020, 19	Intercoder reliability: debates and guidelines	Section 08
Fetters, Curry and Creswell, Health Serv Res 2013, 48:2134; Guetterman et al. 2015	Integration and joint displays	Section 09
Tong et al. 2007 (COREQ); O'Brien et al. 2014 (SRQR)	Reporting standards	Section 10
Silver and Lewins, Using Software in Qualitative Research: A Step-by-Step Guide (2nd ed., Sage, 2014); the CAQDAS Networking Project (University of Surrey)	Tool-independent software strategy; comparative reviews of every package	The reference for sections 04 to 06
QualCoder wiki and manual; Taguette documentation; vendor manuals	The tools	Free
ImpactMojo: Qualitative Methods 101, Mixed Methods 101, Research Ethics 101, Data Protection & the DPDP Act 101; VaniScribe	The methods, the ethics, the law, and the transcription tool	impactmojo.in

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Summary

What to remember

The software files, retrieves and counts. The method, the codes' meanings and the claims are yours.
Transcribe in the language spoken; check against the audio; decide the translation workflow and say what it lost.
A code is a definition. A codebook is versioned. A memo is where the analysis happens.
Free tools cover the whole workflow: Taguette to start, QualCoder for everything else; REFI-QDA makes the coding portable.
Retrievals are read, matrices are pointers, and no purposive count is a proportion of anyone but the sample.

Teams share one codebook, test the merge early, and measure agreement per code where the method calls for it.
Attributes from the survey make the qualitative data cross-tabulable, and a joint display makes the strands one study.
The project file, the codebook versions, the log and the memos are the audit trail; deposit what consent allows.
AI features send transcripts abroad, hallucinate, and miss interpretation; validate like a second coder, or do not use them; disclose either way.
Report to COREQ or SRQR, with a quotation register behind every quotation.

The best QDA project is one an examiner can walk through from any theme back to the recording in ten minutes. Build for that reader.

ImpactMojoQualitative Analysis Software 101www.impactmojo.in

Qualitative Analysis Software 101 · Complete

Now go code it,
and keep the thinking.

The software holds the evidence and the trail; the interpretation stays with the researcher who heard the interviews. Explore the rest of the ImpactMojo 101 Series, free forever.

More 101 Courses Explore ImpactMojo Dataverse

CC BY-NC-ND 4.0·Free Forever·ImpactMojo 101 Series