Impact survey question examples by outcome level and framework — output, outcome and impact sets with scales and waves — vs satisfaction questions.
Impact survey questions ask whether a program changed the people it was meant to change — not whether they enjoyed it. They are written at three levels: outputs (what the participant received), outcomes (what changed for them), and impact (whether the change held, and whether it would have happened anyway). The same wording is asked at baseline, exit, and follow-up so the answers can be compared. A question that is only asked once cannot measure change.
Most surveys labelled impact surveys are satisfaction surveys with different headings. "How would you rate this program?" produces a number that goes up when people are happy and tells a funder nothing about whether anyone's situation improved. The examples on this page are organized so you can take them away and use them: output, outcome and impact questions with the scale and the wave to ask each one at.
Key takeaways
A question measures impact only if it is asked more than once, worded identically each time, and answered by an identifiable person. Change the wording between baseline and exit and the difference you observe is partly a difference in the question. Ask an anonymous cohort and you can compare two averages but never say that a participant moved.
Sopact calls that standard The Wave-Matched Question: identical wording, asked of the same person under one persistent Contact ID, at baseline, exit, and at least one follow-up. It is checkable — take any existing question bank and ask which of its items meet all three conditions. In most banks, fewer than half do.
The reason is a data-model one Sopact calls question-to-wave binding. The established survey tools were built for market research, where each survey is a standalone study of an anonymous sample, so they have no native concept of the same respondent appearing in a later survey. A record-centric tool holds one participant record that accumulates waves, so a question asked in March and again in September resolves to the same person automatically. Question wording and scale construction are covered on survey design; the identity mechanics on pre-and-post surveys.
The first era of survey tooling was market research: a questionnaire fielded once to a sample, analyzed, and closed. SurveyMonkey and Qualtrics come from this lineage and are very good at it. The unit of work is the survey, and a respondent is a row inside one survey.
The second era was the self-serve form builder — Google Forms, Typeform — which made fielding a survey free and instant, and made the standalone-survey assumption universal. Every wave becomes a new form with a new response sheet, and joining them is a spreadsheet task.
The third era treats the participant, not the survey, as the unit. The survey becomes one moment in a record that already exists and keeps going after the program ends. That is what makes a longitudinal question bank usable rather than aspirational.
The one test: pick a single participant and ask your tool to show their answer to the same question at baseline, exit, and follow-up, without an export. If that takes a spreadsheet, your question bank is longitudinal on paper only. What to do with the results once they link is on survey analysis.
Use the three sets below in order: output questions establish what the participant actually received, outcome questions measure what changed for them, and impact questions test whether the change held and whether the program caused it. Each row gives the wording, the scale, and the wave to ask it at. Substitute your own skill, role, or credential where a question is bracketed.
Ask the outcome set at every wave with identical wording — that is the set that carries your evidence. The output set can be asked once at exit. The impact set only works at follow-up, because it asks about persistence.
| Question | Scale or type | Ask at |
|---|---|---|
| How many sessions of the program did you attend? | Number | Exit |
| Did you complete the full program? | Yes / No / Partially | Exit |
| Which components of the program did you take part in? | Multi-select | Exit |
| Did you receive the materials you were told you would receive? | Yes / No | Exit |
| How many hours per week did you spend on program activities? | Number | Exit |
| Did you meet with a mentor or coach during the program? | Yes / No | Exit |
| What, if anything, prevented you from attending sessions? | Open text | Exit |
| Did you earn the credential the program prepares you for? | Yes / No / In progress | Exit, follow-up |
| Question | Scale or type | Ask at |
|---|---|---|
| I am confident I can perform [specific skill] on my own. | 1-5 agreement, named anchors | Baseline, exit, follow-up |
| I know what steps to take next to reach my goal. | 1-5 agreement | Baseline, exit, follow-up |
| I have someone I can go to when I get stuck. | 1-5 agreement | Baseline, exit, follow-up |
| How would you rate your current ability to [specific task]? | 1-5, named anchors | Baseline, exit, follow-up |
| Are you currently employed? | Yes / No / Employed part-time | Baseline, exit, follow-up |
| What is your current hourly wage or annual salary? | Number | Baseline, follow-up |
| Are you currently enrolled in further education or training? | Yes / No | Baseline, follow-up |
| In the last month, how often did you use [specific skill] outside the program? | Never / Once / Monthly / Weekly / Daily | Exit, follow-up |
| What changed for you as a result of this program? | Open text | Exit, follow-up |
| What has been hardest to apply since the program ended? | Open text | Follow-up |
| Question | Scale or type | Ask at |
|---|---|---|
| Are you still in the role or position you moved into? | Yes / No / Different role | Follow-up 6-12 months |
| Has your income changed since you finished the program? | Increased / Same / Decreased | Follow-up |
| How much of this change do you attribute to the program? | None / A little / Some / Most / All | Follow-up |
| What else was happening in your life that contributed to this change? | Open text | Follow-up |
| Would you have been able to make this change without the program? | Definitely not / Probably not / Maybe / Probably / Definitely | Follow-up |
| Has anyone else in your household been affected by this change? | Yes / No + open text | Follow-up |
| Are you currently relying on any public benefits you were relying on before? | Yes / No / Fewer | Baseline, follow-up |
| Looking back, what would have made the biggest difference to your outcome? | Open text | Follow-up |
Theory of change, logic model, IRIS+, the five dimensions of impact and SROI do not need different questions — they need the same questions mapped to different rows in their reporting structure. Write the bank once, map it once, and a new funder framework becomes a mapping exercise rather than a new survey.
One caution on IRIS+: if you are reporting an IRIS+ metric, word the question to that metric's published definition rather than paraphrasing it, or the number you report will not mean what the catalog says it means. The catalog is at IRIS+. The causal chain the questions evidence is on theory of change and logic model, and the wider frame on five dimensions of impact.
| Framework | What it asks you to evidence | Which question set serves it |
|---|---|---|
| Theory of change | That the causal chain held from activities to outcomes | Output set, then outcome set at three waves |
| Logic model | Inputs, activities, outputs, outcomes as distinct rows | Output set feeds outputs; outcome set feeds outcomes |
| IRIS+ | Standardized metrics comparable across organizations | Outcome set, worded to the IRIS+ metric definition |
| Five dimensions of impact | What, who, how much, contribution, risk | Impact set — contribution and counterfactual questions |
| SROI | Valued outcomes with deadweight and attribution | Impact set — attribution and counterfactual questions |
A satisfaction question asks how the program felt; an impact question asks what is different in the participant's life and whether the program caused it. Read the last column — the same topic produces a very different item depending on which one you are writing.
| Topic | Satisfaction question | Impact question (wave-matched) |
|---|---|---|
| Skills | How useful was the training? | I am confident I can perform [skill] on my own. (1-5, all three waves) |
| Employment | Did the program help your career? | Are you currently employed, and at what wage? (baseline and follow-up) |
| Support | Were the staff helpful? | I have someone I can go to when I get stuck. (1-5, all three waves) |
| Application | Would you recommend this program? | In the last month, how often did you use [skill] outside the program? |
| Attribution | Not asked | Would you have been able to make this change without the program? |
The satisfaction column is not worthless — it tells you whether people will come back, and it belongs in an exit survey. It just cannot answer a funder asking what changed. Every question in the right-hand column is a Wave-Matched Question: same wording, same person, more than one wave.
Comparable pre and post questions require six things: identical wording across waves, a named-anchor scale rather than bare numbers, one open prompt beside every closed item, a stable question order, at least three waves, and every response bound to the same participant ID. Break any one and the comparison weakens; break the last and it disappears.
Identical wording is the rule teams break most often, usually with good intentions — someone improves a question between waves and destroys the comparison. Keep a locked version of the instrument and treat wording changes the way you would treat changing a metric definition mid-year. Named anchors matter because "4" means nothing on its own; "I can do this without help" does, and it means the same thing to a different reader six months later.
The open prompt beside each rating is what makes the number defensible. A confidence score that drops four points is a mystery; the same score with fifty sentences explaining it is a finding. Writing those prompts well is covered on qualitative questions, and combining the two kinds of answer on mixed-method surveys.
Well-written questions still produce unusable evidence if wave three lands in a different spreadsheet than wave one. Collecting clean at the source, with every response bound to a participant at the moment it arrives, is what lets a question bank behave longitudinally. That is the premise of the Loop, Sopact's method for continuous impact intelligence: collect clean at the source, analyze the moment data arrives, improve while you can still act.
It is also what keeps the bank honest over time. One locked definition per question, one codebook for the open prompts, and the same instrument every wave — so a change between waves is a real change. That standard has its own chapter in reliability and reproducibility. Where these questions fit in the wider practice is on impact measurement, and audience-specific variants on beneficiary feedback survey and stakeholder survey questions.
One method, three moves that never stop
Then the cycle runs again, a little sharper each wave. Read the method: the Loop methodology →
The bank above is a starting point; the work is adapting it to your outcomes and checking that every item is wave-matched. Each prompt below pastes into Sopact Sense's Assistant, or reasons through with your team; the arrow above each links the Academy walkthrough that shows the expected output and the tips.
Academy walkthrough → Write outcome questions from your outcomes
From this program's intended outcomes, write impact survey questions at all three levels: [PASTE OUTCOMES + PARTICIPANT DESCRIPTION]. For each outcome give one output question, one to three outcome questions with a named-anchor 1-5 scale, and one impact question testing persistence or attribution. Mark which wave each is asked at. Reject any question that would be answered the same way by someone who learned nothing but enjoyed the program. Return a table: Outcome / Level / Question / Scale / Waves.
Academy walkthrough → Audit an existing bank for wave-matching
Audit this question bank against the wave-matched standard: [PASTE QUESTIONS]. For each item say whether it is asked at more than one wave, whether the wording is identical across waves, and whether the response can be tied to an identifiable participant. Mark each PASS or FAIL with the reason. Then list which failed items are worth fixing and which should be dropped. Return a table: Question / Multi-wave? / Identical wording? / Identifiable? / Verdict.
Academy walkthrough → Map your questions to a funder framework
Map these questions to the framework this funder reports against: [PASTE QUESTIONS + FRAMEWORK: theory of change / logic model / IRIS+ / five dimensions / SROI]. For each question give the framework row it evidences and flag any framework row with no question serving it. Where the framework has a published metric definition, flag any question whose wording does not match it. Return two tables: Question / Framework row, and Framework row / Gap.
Academy walkthrough → Pair every rating with an open prompt
For each closed question below, write the one open-ended prompt that should sit beside it: [PASTE CLOSED QUESTIONS]. The prompt should explain movement in that specific rating, be answerable in one or two sentences, and avoid leading the respondent toward a positive answer. Return a table: Closed question / Open prompt / What the pair lets you say.
Each walkthrough is short and practical: what to do, the prompt to run, the output to expect, and the tips that keep it reliable.
Watch: collecting the rating and the reason together at the source, so each wave of questions writes into one participant record.
An impact survey is an instrument that measures whether a program changed the people it served, asked at more than one point in time so change can be observed. It differs from a satisfaction survey by asking what is different in a participant's life rather than how the program felt. Sopact's standard for the items inside one is the Wave-Matched Question: identical wording, same participant ID, at least two waves.
An impact survey means a questionnaire designed to evidence change rather than record opinion. A complete one includes output questions (what the participant received), outcome questions (what changed for them, asked at baseline, exit and follow-up on an identical scale), impact questions (whether the change held and whether the program caused it), and at least one open-ended prompt beside each rating. Sopact binds every wave of it to one persistent Contact ID so the instrument behaves longitudinally.
A social impact survey is the same instrument applied to social rather than commercial outcomes — employment, education, health, housing, wellbeing — usually because a funder or an impact investor requires evidence of change in those domains. The question structure does not change; only the outcomes do. Sopact's guidance is to write the bank against your own theory of change first, then map it to whichever framework the funder reports in.
Outcome-level examples include: 'I am confident I can perform [specific skill] on my own' on a 1-5 named-anchor scale asked at every wave; 'Are you currently employed, and at what wage?' at baseline and follow-up; and 'In the last month, how often did you use [skill] outside the program?' at exit and follow-up. Impact-level examples include 'Would you have been able to make this change without the program?' The full bank on this page is organized by output, outcome and impact level with the scale and wave for each.
Impact evaluation questions are the subset that test persistence and attribution rather than immediate change: whether the participant is still in the role, whether income changed, how much of the change they attribute to the program, and what else in their life contributed. They are asked at follow-up, six to twelve months after exit, because that is the only point at which they can be answered. Sopact treats the attribution items as mandatory — a change you did not cause is not your impact.
A five-point agreement or ability scale with named anchors on every point, not just the ends. '4' carries no meaning on its own; 'I can do this without help' does, and it still means the same thing to a different reader a year later. Keep the same scale and the same direction across every wave. Scale construction in more depth is on Sopact's survey design page.
A satisfaction question asks how the program felt — useful, enjoyable, worth recommending. An impact question asks what is different in the participant's life and whether the program caused it. The test is simple: if someone who learned nothing but had a pleasant experience would answer the same way, it is a satisfaction question. Sopact keeps both in an exit survey but only the impact set is wave-matched.
Fewer than you want to. Keep the wave-matched core to roughly five to ten items so the same instrument can be fielded three or more times without exhausting participants, and add one open prompt per rating. Response rate matters more than question count, because attrition between waves does more damage to a longitudinal finding than a missing question does. Sopact's guidance is to cut anything that will not be asked again.
Yes, and you should. Theory of change, logic model, IRIS+, the five dimensions of impact and SROI mostly need the same evidence mapped into different reporting rows, not different questions. Write one bank, map it once per framework, and a new funder becomes a mapping exercise. Sopact holds the mapping alongside the definitions so a second funder does not trigger a second survey.
You can field the questions, but those tools come from market research where each survey is a standalone study of an anonymous sample, so there is no native concept of the same respondent appearing in a later wave. Every follow-up becomes a new form with a new sheet, and the join is spreadsheet work repeated each cycle. Sopact binds each response to a persistent Contact ID at collection, which is what makes a Wave-Matched Question possible at all.
Next: get the wording and scales right on survey design, or run the analysis these questions feed on how to analyze survey data.
