Every programme counts its outputs.
The chain breaks one link later.
The results chain, made clickable. Six large Indian programmes drawn as the same five links, with each link marked by how it is actually measured: a public system updated daily, a national survey every few years, a researcher who happened to go there, or nothing.
The break is not gradual and it is not random. In all six it falls between outputs and outcomes, which is exactly where attribution begins.
The chain is the logical framework, built by Practical Concepts Incorporated for USAID around 1969–70 and given its standard definitions by the OECD Development Assistance Committee in 2002. ImpactMojo added the measurement layer and the Indian programmes. Full credits ↓
Six programmes, one shape
Each bar is a link in the chain. Its height is how close the measurement sits to the programme and how often it arrives: a public system reporting every transaction stands tall, a national survey on a five-year cycle is lower, an independent study lower again, and nothing is a stub. The vertical rule marks the step.
What the shape is not
It is not a claim that these programmes are badly run. MGNREGA’s management information system is among the most granular administrative datasets any government publishes anywhere, and the Poshan Tracker reports daily. The measurement that exists is often excellent.
It is not a claim that the missing measurement would be easy. An outcome takes longer to appear than a budget cycle, costs more to collect than an output, and belongs to a population rather than to a transaction. The reason outputs are counted is partly that they can be.
The claim is narrower. The counting stops one link before the thing the programme is defended with in public, in all six cases, and it stops there for a structural reason rather than six separate accidental ones: an output is something the programme controls and can be held to, and an outcome is something it can only influence. What gets built is what someone can be answerable for.
What each height means
A ranking of how close a measurement sits and how often it arrives, not a score of quality. A well-run system can report the wrong thing, and a single careful study can be right.
Pick a programme, then a link
The instrument behind each link, what it actually reports, and the point at which the chain stops carrying weight.
Whose framework this is
The framework
Practical Concepts Incorporated, led by Leon J. Rosenberg, built the logical framework for USAID around 1969–70. Its four-by-four matrix, and the chain of inputs, outputs, purpose and goal inside it, are the ancestor of every logic model and results chain since.
Designed to make a project’s assumptions writable down, which is still what it is best at
The standard definitions
The OECD Development Assistance Committee, Glossary of Key Terms in Evaluation and Results Based Management (2002). The distinctions between output, outcome and impact used on this page are the DAC’s, and the reason they can be compared across ministries at all.
Revised 2010; the 2002 definitions are the ones in general use
Theory of change
Carol Weiss, whose 1995 essay argued that an evaluation should test the theory a programme rests on rather than only its results, and gave the phrase its currency. The chain is the skeleton; the theory is the claim that each link causes the next.
The question this page borrows from her: which link do you actually have evidence for?
What ImpactMojo added
The measurement layer: for each of the thirty links across six programmes, the instrument that reports it, what that instrument actually says, and where the chain stops carrying weight. The four states and their heights are ours, and so is the reading of where each chain breaks.
ImpactMojo, 2026 · content CC BY-NC-ND 4.0 · code MIT
Sources drawn on
MGNREGA MIS, Ministry of Rural Development · Swachh Bharat Mission (Gramin) MIS · SQUAT survey, r.i.c.e. (2014) and Coffey and Spears, Where India Goes (2017) · Annual Status of Education Report, Pratham (2018, 2022) · UDISE+ · National Family Health Survey rounds 2 to 5, via the DHS StatCompiler · Poshan Tracker · National Health Accounts Estimates for India · PM-KISAN portal · Imbert and Papp, AEJ: Applied (2015) · Muralidharan, Niehaus and Sukhtankar, AER (2016).
Using this in a workshop
Take the group’s own programme and build its five links on a whiteboard, then mark each one with which of the four states it is in. The useful moment is not the drawing; it is the argument about whether a number someone offers is an output or an outcome, because that argument is usually the first time anyone in the room has had to say out loud what the programme is claiming to cause.
Then ask the harder question. For the link where your chain breaks, what would it cost to measure, who would have to agree to be answerable for it, and what would happen to the programme if the answer came back badly? The third part is generally why the measurement does not exist.
Read alongside Who Counts, which holds the measurement gaps this page runs into, and Resources, Agency, Achievements, which is about the opposite failure: an outcome indicator that moves without the thing it stands for moving.
If you reproduce the framework, credit the logical framework and the DAC. If you reproduce the measurement layer or the four states, credit ImpactMojo, and argue with the heights, which are ours.