for we aim at what is honorable not only in the Lord's sight but also in the sight of man.2 Corinthians 8:21
Readers have learned to discount impact reports. They expect a large headline number, warm photos, a few testimonials, and very little about how any of it was measured. That expectation is the opportunity: a report that shows its working stands out immediately, and it is easier to write than a polished one because it only has to be true.
A trustworthy impact report does three things the usual version avoids. It keeps reach, delivery, use, return, and change on separate lines so the reader can see where the numbers narrow. It names the method behind every outcome claim. And it publishes limitations and failures at the same size as the wins. Those three choices do more for credibility than any amount of design.
Below is a structure you can copy section by section. It works for a nonprofit annual report, a ministry update, a social enterprise, or a founder's quarterly letter. It also works when the numbers are small — which is exactly when most organisations are tempted to round up.
1. Lead with the promise, not the headline number.
Open with one sentence describing who you set out to help and what change you hoped would follow. This is your theory of change in plain language. Without it, every number in the report floats free and the reader cannot judge whether a result is good. “We help first-time parents in east Hamilton feel confident feeding a newborn in the first month” gives a reader everything they need to interpret what comes next.
- Name the people, not a demographic slogan.
- Name the change you hoped for, written before the results came in.
- State the reporting window up front.
2. Show the funnel from reach to change.
Present your numbers in order, from widest to narrowest: how many were reached, how many received the complete thing you offered, how many used it, how many came back, and how many show evidence of the intended change. A funnel that narrows is normal and believable. A report that jumps straight from reach to “lives changed” invites the reader to wonder what happened in between.
If later numbers are higher than earlier ones, check that you are counting the same cohort over the same window before you publish. Mismatched populations are the most common reason impact reports quietly contradict themselves.
3. Name the method beside every outcome.
Every outcome claim needs a short note on how you know it: who was asked, when, what was asked, how many responded, and how the answer was judged. “23 of 61 participants (38%) answered a 30-day follow-up; 17 said they still use the budgeting method weekly” is an outcome with a method. “Participants transformed their finances” is not.
You do not need academic rigour. You need enough detail that a sceptical reader could understand, and in principle repeat, what you did. Self-report is fine if you call it self-report. A comparison group is better if you have one. Independent verification is better still. Match your language to the strength of the method.
4. Use stories to illustrate the numbers, not replace them.
Stories carry meaning numbers cannot, and a good report should include them. The problem arises when a single exceptional story stands in for the typical result. Say plainly where a story sits: “This is one of the four participants who reported the full change” is honest. Presenting the best case as representative is not, even when every word of the story is true.
Get consent, protect privacy, and avoid language that makes the people you serve props in your own success narrative.
5. Publish limitations and failures at full size.
Include a section — not a footnote — on what did not work, what the data cannot show, and what you would measure differently. Low response rates, short follow-up windows, results that did not hold, programmes you stopped: all of it belongs here. Readers trust a report more when it volunteers its weaknesses, because it signals that the strong claims were held to the same standard.
- Response rate and who did not respond.
- What the method cannot rule out.
- Results that were lower than hoped, and the likely reasons.
- Anything you stopped doing, and why.
6. End with the decision the evidence earned.
Close with what you will do next because of these results: continue, change something specific, or stop. This turns the report from a performance into a tool, and it gives the next report a baseline to be judged against. A reader who can see the decision can also see whether it was justified.
Date the report and keep old versions public. A record of honest reports, including the disappointing ones, builds more trust over time than any single impressive one.
7. A worked example: a small pilot without an impact claim.
The following example is hypothetical, not a SIX33 result. Imagine a founder testing a worksheet that helps first-time founders state a complete promise. Twelve people received the invitation, six received the whole worksheet, four completed it, and two returned with a revised sentence. No follow-up evidence was collected about whether their projects became more useful. The reach, delivery, use, and return counts are twelve, six, four, and two; evidenced change is not measured, not zero and not four.
A defensible finding is: “Four participants completed the worksheet during the pilot window; two returned a revised promise.” The method might be a dated delivery log, completion submissions, and a record of the replies. The limitations are equally important: the participants volunteered, the group is small, and finishing a worksheet does not establish a lasting change in their work. Avoid claiming that the pilot transformed four businesses.
The next decision could be to change one confusing prompt and observe another round. To investigate a later outcome, define it separately, agree on a follow-up window, and decide what evidence would support it. A record of use can justify improving the next delivery without being upgraded into verified impact.
- Reporting window: write the actual start and end dates.
- Keep unknown results unmeasured; use zero only when you measured and observed none.
- Finding: state what happened in the language your method supports.
- Decision: change one prompt, then test the revised version.
8. Move from the scorecard to a report, then review before sharing.
If your numbers are already in the SIX33 impact scorecard, open the proof pack builder and choose Review scorecard snapshot. It reads the scorecard in this browser and shows the project, promise, intended change, and five counts before changing anything. Blank counts remain unmeasured. The planned change is labelled as an intention, not inserted as a finding. Your original scorecard stays intact.
Back up an existing report first if you want to keep it. Choosing Replace report with snapshot creates a fresh report draft from the reviewed fields; it does not merge two reporting windows or add counts together. The scorecard's combined evidence notes appear in the method field so you can separate the method from the limitations. People served, dates, finding, decision, and next test remain blank for you to complete. Copying a field does not verify it.
Before sharing, read the report as someone outside the team. Can they tell who was counted, during which period, and how you know the result? Remove private participant details and unapproved quotes. Export Markdown for the readable report and JSON for a restorable working copy. Files are readable rather than encrypted; keep a private backup separate from the version you intend to share. The builder exports a file but does not publish it for you.
- Match every count to a definition, cohort, and reporting window.
- Put the evidence method beside the finding and the material limitations beside the claim.
- Check that the intended change is not presented as an observed outcome.
- Remove identifying details you do not intend to share.
- Name a continue, change, or stop decision and the next test.
- Keep a dated backup before publishing the reviewed report.
9. Run a completeness review without mistaking it for verification.
Before sharing a report, check whether a reader has enough context to inspect it. The SIX33 builder's Before you share panel checks eight areas: the project and promise, people served, reporting dates, signal formats and consistency, finding, method, limitations, and decision with next test. Each item opens the relevant field so you can correct a missing date or explain a result without searching the whole form.
A filled checklist is not an impact certificate. A plausible sentence can still be false, a quotation can still lack permission, and a carefully counted outcome can still have another cause. Read the actual claims, compare them with the underlying record, and check consent and privacy yourself. Text length helps the tool notice an empty method; it cannot judge whether the method is credible.
Keep measurement states separate during the review. A blank count means not measured. Zero means the defined measurement found none. An invalid or fractional count needs correction rather than silent rounding. If a later count exceeds an earlier one, inspect definitions, cohorts, and windows before treating the sequence as a funnel. The review preserves those warnings in exported Markdown and JSON, even if you choose to keep the report as an unfinished working draft.
- Fill missing context before sharing a public version.
- Investigate count warnings; do not delete them merely to make the report look complete.
- Check the original record, permissions, and identifying details yourself.
- Export unfinished work as a backup when needed; a saved draft is not a published or verified report.
The short versionLead with the promise, show the funnel narrow, name the method, publish the limits at full size, and end with a decision. Credibility comes from what you are willing to show, not from how large the headline is.
Direct answers
Questions about writing an impact report.
What should an impact report include?
A clear statement of who you serve and the change you hoped for, the reporting window, numbers from reach through delivery, use, return and evidenced change, the method behind each outcome, honest limitations and failures, a few clearly-labelled stories, and the decision the results lead to.
Does a completed report checklist verify an impact claim?
No. A completeness checklist can flag missing fields and obvious count inconsistencies, but cannot establish truth, permission, causation, or independently verified impact. Review the original evidence, permissions, limitations, and private details yourself before sharing.
How do I turn a scorecard into an impact report?
Keep each signal separate, add the reporting window and people counted, write the finding in the language your evidence supports, and publish its method and limitations. In SIX33, Review scorecard snapshot copies a reviewed local scorecard into the proof pack builder only after confirmation. It does not combine counts, invent findings, verify impact, or publish the report.
How long should an impact report be?
Long enough to show the funnel, the method and the limitations — often one to four pages. Length adds nothing if the method is missing. A short report with a clear method is more credible than a long one without it.
How do you make an impact report credible?
Separate reach from outcomes, state the method and response rate for every outcome claim, publish limitations and negative results prominently, label stories as illustrations rather than typical results, and date the report so it can be compared with later ones.
What is the difference between outputs and outcomes in an impact report?
Outputs describe what you did, such as sessions held or products delivered. Outcomes describe what changed for people as a result. Both belong in the report, but outputs should never be described in outcome language.
Can I publish an impact report with small or zero results?
Yes. A dated, honest report with small numbers is more useful and more trustworthy than an inflated one. It sets a baseline, shows your standards, and makes later growth believable.