Build credible impact evidence without inflated claims.
On this guide Browse sections
Impact measurement is where NGOs are most likely to overstate.
A credible organisation distinguishes activity, output, outcome and impact.
1. Four levels
Activity
What the NGO did.
Output
What was directly delivered.
Outcome
A change observed after the intervention.
Impact
Longer-term change attributable or plausibly connected to intervention, often requiring stronger evidence.
2. Start with the claim
Write the claim before choosing evidence.
Example:
Then ask:
- Who are the 120?
- How measured?
- What is adoption?
- What happened to others?
- Who verified?
- Is there comparison?
3. Avoid “lives impacted”
Unless defined.
Instead use:
- people trained;
- households served;
- children enrolled;
- farmers adopting practice;
- patients receiving service;
- institutions supported.
4. Denominator matters
“80 improved” is incomplete.
80 out of 100?
80 out of 500?
State denominator.
5. Attrition
If 200 enrolled but only 120 completed follow-up, say so.
Do not calculate the improvement percentage as if all 200 were measured.
6. Baseline and endline
Use the same definition/tool where possible.
Document:
- sample;
- tool;
- timing;
- evaluator;
- missing data;
- change.
7. Contribution vs attribution
Social change has many causes.
Use contribution language where appropriate:
Avoid:
unless the design supports it.
8. Triangulation
Use more than one evidence source where useful:
- programme records;
- assessment;
- interview;
- observation;
- partner data;
- external source.
9. Case stories
A case story can show:
- lived experience;
- context;
- programme pathway;
- change.
It does not prove project-wide impact.
10. Qualitative evidence
Useful methods:
- interview;
- focus group;
- observation;
- case note;
- stakeholder feedback.
Document how selected.
11. Quantitative evidence
Useful metrics:
- count;
- percentage;
- rate;
- change;
- average;
- completion;
- adoption.
Explain definition.
12. External evaluation
Consider when:
- project is large;
- donor requires;
- claim is high-stakes;
- methodology needs independence.
13. Limitations
Every serious impact report should state limitations.
Examples:
- small sample;
- no comparison group;
- self-reported;
- short follow-up;
- incomplete data;
- seasonal change;
- external programme influence.
14. Evidence register
| Claim | Indicator | Source | Period | Sample | Limitation | Evidence file |
|---|
15. Use evidence in proposals
Use previous project results only where relevant.
Do not apply an outcome from one district to another as guaranteed.
16. Use evidence in annual reports
Separate:
- reach;
- outputs;
- outcomes;
- learning.
17. Use evidence in donor reports
Compare:
- target;
- actual;
- variance;
- cause;
- action.
18. Do not cherry-pick
Report negative or mixed findings where material.
Credibility improves.
19. Data ethics
Avoid unnecessary disclosure of vulnerable beneficiaries.
Use consent/permission and anonymisation where appropriate.
20. Common impact mistakes
- attendance = impact;
- target = achievement;
- story = proof;
- no denominator;
- no limitation;
- no definition;
- no baseline;
- only success reported;
- causation overstated;
- sensitive identity exposed.
21. Build an evidence ladder
Different claims require different evidence.
Level 1 — Activity evidence
Attendance, service record, delivery record.
Level 2 — Output evidence
Completed training, asset built, service delivered.
Level 3 — Outcome evidence
Assessment, behaviour, adoption, school attendance, health/service outcome.
Level 4 — Stronger impact evidence
Longer-term follow-up, comparison, external evaluation or robust design.
Do not use Level 1 evidence for Level 4 claims.
22. Claim register
Maintain:
| Claim | Period | Population | Measure | Source | Limitation | Approved for use |
|---|
This improves consistency across:
- proposal;
- website;
- annual report;
- donor report;
- social media.
23. Evidence reuse
A verified result can be reused with the same context.
Do not change:
- period;
- geography;
- denominator
when reusing it.
24. Social-media claims
Social posts should follow the same evidence standard.
Avoid:
- “transformed thousands”;
- “100% successful”;
- “eradicated”;
unless evidence supports it.
25. Case-study method
Record:
- why selected;
- interview;
- consent;
- factual verification;
- before/after context;
- limitation.
Do not script an artificial success story.
26. Outcome harvesting
For complex projects, the NGO can identify significant observed changes and then work backward to verify contribution.
Document evidence and alternative explanations.
27. Cost-effectiveness
Possible measures:
- cost per participant;
- cost per completion;
- cost per outcome.
Use carefully because different beneficiary needs have different costs.
28. Benchmarking
Compare with:
- baseline;
- previous year;
- programme target;
- credible external benchmark.
Do not compare incomparable programmes.
29. Longitudinal evidence
Where useful, follow the same participants over time.
Track attrition.
30. Qualitative coding
For interviews/focus groups:
- define themes;
- document sample;
- record method;
- avoid selecting only favourable quotes.
31. Negative evidence
Important findings include:
- no change;
- unexpected outcome;
- dropout;
- harm/risk;
- unintended exclusion.
Use these to improve programme.
32. Board/governance use
Impact evidence should inform:
- strategy;
- programme continuation;
- budget;
- fundraising;
- staffing;
- risk.
33. Funding use
Use impact evidence in proposals only with:
- source;
- period;
- geography;
- context.
Do not promise the same result automatically in a new project.
34. Public evidence page
A good website impact page can present:
- reach;
- outputs;
- outcomes;
- reports;
- methodology;
- limitations.
This is more credible than large unsupported headline numbers.
Official sources to keep open
When you report impact to government or CSR funders, keep your public figures consistent with what you submit on these portals:
Links checked on 21 October 2026. Portals change; if a page has moved, search the department’s own site rather than a third-party copy.
Worked examples: making claims you can defend
These are hypothetical examples to show the method. They are not real organisations or cases.
For example, a trust in Punjab: its brochure says “10,000 lives impacted”, which adds up everyone ever contacted. Fix: report how many people completed the programme and what changed for them.
For example, a society in Kerala: it claims improved test scores but has no baseline. Fix: say what you observed and that no baseline was taken.
For example, a Section 8 company in Rajasthan: it credits its programme for a rise in school enrolment that also had a government push. Fix: describe your contribution rather than claiming full attribution.
For example, a society in Jharkhand: it uses one powerful case story as if it were typical. Fix: present it as one example and give the overall figures alongside.
For example, a trust in Telangana: it shares children’s photos with names. Fix: get consent and remove identifying details.
For example, a society in Arunachal Pradesh: many participants dropped out, but the report only covers those who finished. Fix: report the drop-out and its likely reasons.
Frequently asked questions
What is the difference between output, outcome and impact?
Outputs are what you deliver, such as training sessions held. Outcomes are changes for participants, such as new skills used. Impact is longer-term change that your work contributes to.
Can we say “lives impacted”?
Only if you define it and can show the evidence. It is clearer to state how many people took part and what changed for them.
How can a small NGO measure impact without a big budget?
Collect a simple baseline, track a few outcomes well, use participant feedback, and be honest about limitations. A clear method matters more than a large sample.
Should we hire an external evaluator?
It helps for larger or longer projects, or if the funder requires it. For small projects, a well-documented internal review is often acceptable.
What should we avoid in impact claims?
Unsourced statistics, claiming full credit where others contributed, ignoring drop-outs, and using one success story as if it were the norm.
Your next step
Turn your evidence into reports and annual reports funders can trust.
NGOGrant boundary
NGOGrant can help structure evidence and reporting. Claims must remain proportionate to the actual method and data.