What is a social impact report?
A social impact report is a document that tells its readers what changed for people because of a program or organization, for whom, compared with what, and how confident anyone can be, with the evidence and the gaps shown beside each finding. It covers difficulties and misses as well as progress.
The people in question may be program participants, employees, suppliers or a community. Activity counts still belong in the report, because they show what was delivered, but sessions run or money spent do not by themselves show a change in anyone's life.
THE SHORT VERSION
- Test every finding against five questions: for whom, compared with what, with what evidence, with what confidence, and now what.
- Show each number with its denominator and definition, and hold back any number counted differently instead of adding it in.
- Draft with AI from selected sources if you like, let taste change the words for each reader, and never let it change the numbers.
How is a social impact report different from a social impact assessment or a sustainability report?
A social impact report communicates findings about effects on people; a social impact assessment is a process for anticipating and managing those effects; a sustainability report covers people alongside the economy and environment. They can share evidence, but their purposes differ.
The International Association for Impact Assessment describes social impact assessment as analyzing, monitoring and managing intended and unintended consequences, positive and negative. Assessment often starts before a project begins and shapes its design.
GRI Standards support reporting on impacts on the economy, environment and people; a general social impact report is not automatically standards-compliant. A grant report is narrower again, answering one award's commitments.
What five questions should a social impact report answer?
For whom, compared with what, with what evidence, with what confidence, and now what. Ask them of every important finding; a finding that fails one of them is not ready to publish.
| Question | What to show | What to avoid |
|---|---|---|
| For whom? | Population, place, period, differences within the group | A result that hides who was left out |
| Compared with what? | Baseline, target, earlier cohort or outside number on the same definition | Comparing unlike groups or measures |
| With what evidence? | Sources, definitions, the records behind each claim | A statistic nobody can recount |
| With what confidence? | Coverage, missing data, other explanations | Observation presented as proof |
| Now what? | The decision, an owner, a review date | A promise with nobody accountable |
Not every finding has a fair comparison. When it does not, say so; a weak comparison does not become useful because the template has a column for it.
What should a social impact report include?
Eight parts: summary, context, methods, delivery and outcomes, differences between groups, people's experience, limits, and learning with action. The order puts the finding first and the method within reach.
| Part | What it holds | Common failure |
|---|---|---|
| Summary | Scope, main findings, the biggest limit | Headline stronger than the evidence |
| Context | The issue, the people, the intended change | No theory of change behind it |
| Methods | Sources, definitions, dates, analysis | Buried where nobody reads it |
| Delivery and outcomes | What was done, then what changed | Outputs presented as outcomes |
| Differences | Groups or sites with different results | Small groups made identifiable |
| People's experience | Accounts that explain a number, with permission | One story standing in for everyone |
| Limits | Missing data, uncertainty, framework mapping | Missing data left unmentioned |
| Learning and action | What changes, who owns it, when | Actions unconnected to findings |
For a copyable outline, use the impact report template; for the wider practice across funders, see impact reporting. The free guide How to Write an Impact Report and the report examples library show finished reports.
What does a social impact report look like for a workforce program?
Picture a workforce fund, fictional here, that backs four job-training partners and publishes one social impact report a year on what changed for trainees. Its headline finding, and how the five questions test it, shows the whole method in one line.
Every partner reports the same five agreed measures: enrolment of unique people, completion, job placement inside 90 days of exit, retention at the 12-month mark and hourly starting wage per track. Its draft headline reads: "100 trainees were placed in a job within 90 days of finishing training."
| Question | Answer for the fund's headline · fictional |
|---|---|
| For whom? | Trainees at partners A, B and D, who placed 42, 31 and 27 |
| Compared with what? | Earlier cohorts, or a county figure only if it uses the same 90-day definition |
| With what evidence? | Partner placement records, checked against the agreement |
| With what confidence? | Partner C's 55 used a six-month window and is held; 12-month retention not yet available |
| Now what? | Ask C for the 90-day count; add track to intake forms so wage can be split |
The tested headline is longer and more credible: "100 trainees at three of our four partners were placed within 90 days; the fourth partner's count is being confirmed on the same definition." Auroville International USA builds its funder report the same way, from about 80 project partners' quarterly reports. The chapter Roll up and benchmark portfolio results shows how the total is built.
The walkthrough below builds a report for a job-training and placement program from its theory of change, IRIS+ metrics and the SDGs. Watch how the example keeps the program’s outcome definitions intact as it moves into each framework’s labels.
How do you show denominators and comparisons honestly?
Put each number beside the population it counts, keep numbers counted differently apart, and compare only with figures that use the same definition. Most misleading social reports mislead by denominator, not by arithmetic.
Adding Partner C's 55 would give 155 placements, a derived total that is wrong under either window. Showing 100 with its three partners, and 55 held with its reason, gives the reader both facts without blending them. The same rule applies over time: if last year counted placement at six months and this year at 90 days, mark the break instead of drawing one trend line.
Break results down by group or site when the data supports it. For small groups, report counts or combine categories so nobody is identifiable, and treat a difference between sites as a question, not proof one team did better.
How do you write a social impact report with AI?
Give the AI only the sources it may use, your data dictionary and the agreed metrics, ask for a source beside every number, and require it to flag anything it cannot find. Readers cannot tell an estimate from a count unless the report says which is which.
The slide below shows why a list of expected metrics matters. In a demo workforce dataset, a first draft reported two of five metrics and flagged the other three, because the placements survey had not been selected; once it was added, the draft was complete.
PROMPT · PASTE INTO CLAUDE, CHATGPT OR YOUR AI TOOL
Draft the findings section of our [YEAR] social impact report for [READER]. Use only: [SELECTED SOURCES], our data dictionary and our agreed metrics. For each finding, answer: for whom, compared with what, with what evidence, with what confidence, and now what. Rules: 1. Show every number with its denominator, definition, period and source. 2. Never add numbers that use different definitions; list them separately with the reason. 3. If a number is not in the selected sources, write "not in our data" and name the metric. Do not estimate. 4. Do not claim the program caused a change unless a comparison in the sources supports it. 5. Do not invent quotations.
In Sopact Sense, the AI Assistant stays locked until you choose which surveys it may use, field selection keeps names and emails out, and every line of an answer links to a record you can open. The chapter Write each funder's report with AI, then check it gives the checklist to run before anything is published.
Accuracy alone does not get a report read, and the video below is about that gap. Watch for the context a reader needs around the numbers: who they are, what they will decide, and which words they trust.
How do you write one social impact report for several readers?
Keep one set of evidence and change only the words, order and format for each reader, guided by a short taste file per reader and one for your own voice. Taste changes the words, never the numbers.
A board, a funder and a community read differently. The taste example in the course describes an officer facing a stack of 40 reports before a board meeting, who wants "participants" rather than "beneficiaries" and distrusts any miss hidden in an average. Your own taste file says why you exist, what you learned, misses included, and the claims you will never make.
Before publishing, ask the AI to read the draft as that reader and say which sentence loses them. Capture each funder's taste, and your own shows how to build the files. Choose the format by the reader's task: a PDF fixes one period, a dashboard supports exploring, and a web page is easiest to navigate; date each so one figure never appears as two answers.
What can a social impact report not claim?
It can report what was observed and what adds up honestly; without a comparison group it cannot claim the program caused the change. The fund's 100 placements describe what happened to trainees; they do not show the jobs would not have come anyway.
Training delivered is an output; a job held is an outcome, and the report should never blur the two. Trainees who drop out of follow-up surveys rarely resemble those who stay, and self-reported pay is approximate. One participant's account shows what the change felt like, not how common it was; people decide what a finding means and check AI output against the records.
Start with one finding in your next social impact report
Choose the one finding your readers care most about and take it through every step before you write anything else.
- Write the finding as a single sentence with its number and denominator.
- Answer the five questions for it, marking any you cannot answer yet.
- List the sources behind it and open each record that produced the number.
- Set aside any figure counted on a different definition, with its reason.
- Draft it with the prompt above, then rewrite the wording for your main reader's taste.
Once that one finding holds up, you have the pattern for every other finding in the report, plus a list of what is still missing.
Frequently asked questions
What is social impact reporting?
It is the ongoing work of explaining to funders, boards and communities what changed for people, with the evidence, the limits and the decisions that follow. The social impact report is the document one cycle produces. It can cover program participants, employees, customers, suppliers or communities, and it should include negative effects and misses as well as progress.
What goes in a social impact report?
A summary, context, methods, delivery and outcomes kept apart, differences between groups, people's experience used with permission, limits, and learning with action. Every important finding should say for whom, compared with what, with what evidence, with what confidence, and what happens next. Adapt the depth to your readers and to any reporting standard that applies.
How do I write a social impact report without a baseline?
Say that no suitable baseline exists, describe the current findings and their scope, and plan how you will measure change next time. A first report can set the starting position. Do not build a before-and-after claim from an unsuitable comparison; an earlier cohort or an outside figure only helps when it uses the same definition.
Does a social impact report need individual participant IDs?
Not for every claim, but outcomes over time do. Retention at 12 months needs the exit record and the follow-up to belong to the same person, which requires one ID carried from the first form. Aggregated or anonymous evidence can still support group findings if the report says which kind each claim rests on.
How long should a social impact report be?
There is no standard length. Many readers want a two-page front section with the main findings and a table, then methods and detail behind it. Answer the five questions for each important finding, and keep essential limits visible even in the shortest board summary.
Can AI write a social impact report?
AI can draft it from selected sources, your data dictionary and the agreed metrics, and adapt the wording for each reader. It should flag missing numbers instead of estimating, and it should not decide what a result means or invent quotations. Check every figure against its record before publishing; drafting gets faster, but the judgment stays with your team.
