Measure change at exit by matching each participant’s relevant starting and ending observations, checking comparability and reporting who has usable data. Keep completion counts alongside those results and ask participants what contributed to their experience. A before-and-after difference describes observed change; it does not by itself establish what the program caused.
By Sopact Academy · Updated September 12, 2026. All numerical examples are fictional.
This lesson helps training, coaching, fellowship and service teams turn their intake and check-in records into an exit report. Bring the baseline definitions and the support history from the mid-program lesson. Leave with an exit-question plan, a participant summary and a cohort calculation that shows its coverage and limitations.
Exercise outcome: a report that distinguishes delivery, measured change, participant explanations and unanswered questions. The figures below are fictional teaching examples, not customer results.
Build an exit report in four steps
- Select the baseline measures that should be repeated at exit.
- Ask what contributed to change without treating opinion as causal proof.
- Check a source-linked report for each participant.
- Summarize the cohort with coverage, distributions and limitations.
Exit is one point in the lifecycle, not necessarily the final outcome. A course may finish before a participant finds sustained employment. Schedule later observations where the outcome requires them, and record how participants prefer to be contacted. Do not make agreement to optional follow-up a condition of receiving the program.
You can calculate matched change in a spreadsheet when identifiers, dates and definitions are clear. Connected software can help maintain those relationships across repeated submissions and documents. Either way, staff must check the source data and interpretation before sharing results.
Completion is not change
A fictional report says 80 enrolled, 62 completed and 58 earned a credential. Those are useful delivery and attainment facts. They do not tell you how skills changed, whose circumstances improved or why. A person can gain something valuable without completing, and someone starting near the top of a scale may finish with little numerical change.
Link comparable observations using participant and enrollment identifiers. Keep the observation dates and the version of the measure. Investigate stable or declining scores in context rather than automatically grading the person or program as unsuccessful. Maintaining a condition can itself be the intended outcome in some services.
Step 1 — Match the measures that belong at exit
Build the exit plan from the collection plan, not from an unrelated questionnaire. Repeat measures whose expected change belongs at exit; leave later outcomes to their planned follow-up. Add relevant completion facts and questions about the participant’s experience. Avoid repeating every baseline field simply because it exists.
Try this with fictional or appropriately authorized material in your approved AI workspace:
Here are the intake questions, measurement definitions and planned follow-up schedule: [paste them]. Propose exit questions that fit the expected timing of each outcome. Return: baseline field, exit field, wording and scale comparison, eligible population, calculation, source and unresolved issue. Mark unmatched fields without inventing baseline values. Keep later follow-up outcomes separate. Suggest neutral questions about helpful, unhelpful and external influences; do not assign causal percentages from an opinion.
Comparability matters more than literal symmetry. Keep wording and scales stable where appropriate; document a justified change and assess whether a comparison remains defensible. A new exit-only measure can still be useful for a clearly stated question, but it cannot supply its own missing starting value. Record appropriate follow-up contact permissions and preferences.
Step 2 — Ask what contributed without overclaiming attribution
Ask what helped, what did not and what else influenced the result. For example: “What changed for you during this period? Which parts of the program contributed, if any? What other people or circumstances mattered?” Let people say they do not know, nothing changed or something became worse. Their explanation is evidence of their perspective, not an observed counterfactual.
Here is the program and outcome: [describe them]. Draft neutral exit questions about perceived change, useful support, unhelpful elements and other influences. Include room for no change, negative change and uncertainty. Summarize answers only with source references. Separate the participant’s stated explanation from an analyst’s interpretation. Do not label an answer strong causal evidence merely because it credits the program.
A participant who says a mentor helped them finish offers useful information about a possible mechanism. Thanking staff is a different kind of feedback. Neither answer proves what would have happened without the program. The World Bank’s guidance on selection bias explains why groups that differ for other reasons cannot simply be treated as comparable. Completers and non-completers may differ before the program and during it; their outcome gap is not automatically the effect of completion.
Step 3 — Check the participant exit summary
Prepare each participant summary from the relevant dated observations. Show the calculation, the source of each value and any missing or changed measure. Separate an extracted quotation, a calculated difference and a reviewer’s interpretation. In a configured Sopact workflow, linked records can support this preparation as responses arrive; the reviewer still verifies the result.
| Evidence | Fictional result | Interpretation |
|---|---|---|
| Confidence, 1–10 | Intake 4; mid 3; exit 7; +3 points intake-to-exit | Observed self-rating change; check comparability |
| Same skills assessment | 40 to 78; +38 points | Observed assessment change; do not equate score points with causal impact |
| Participant explanation | Credits mentor support with helping complete the course | Self-reported contribution; retain quotation and source |
| Review | Comparable pairs available; causal attribution not established | Check missing context before reporting |
In the fictional example, confidence changes from 4 at intake to 3 mid-program and 7 at exit: an exit-minus-intake difference of +3 points. An assessment moves from 40 to 78: +38 points. Whether those differences are meaningful depends on the instruments and context. The participant’s account of mentor support adds context, but does not establish that the mentor caused either difference.
Step 4 — Summarize the cohort and show coverage
Build the cohort result from a defined eligible group. For paired change, use records with comparable observations at both points, then report how many eligible participants are excluded and why. Averages are useful alongside the sample size, distribution and coverage; they do not need to be replaced with individual stories. Use de-identified reporting where individual disclosure is unnecessary.
- "Report cohort change built from each person's own before-and-after" — calculate comparable paired change and report coverage and variation.
- "Describe stable and declining scores with their context and coverage" — investigate instrument limits and program goals before interpreting them.
- "Compare completers and non-completers at follow-up" — describe the gap while explaining selection and missing-data limits.
- "Show participant explanations separately from calculated change" — do not turn a reported opinion into an attributable effect.
| Fictional cohort check | Result | Reporting implication |
|---|---|---|
| Eligible starters | 40 | Define eligibility before calculating |
| Comparable intake/exit pairs | 30 of 40; 75% coverage | 10 do not have usable pairs |
| Total paired confidence change | +60 points across 30 pairs | Mean paired change +2 points |
| Within the 30 pairs | 20 increased; 6 unchanged; 4 decreased | Show distribution alongside the mean; not automatic success/failure labels |
Keep later employment results separate from exit observations. A difference between completers and non-completers is descriptive unless the evaluation addresses relevant differences between those groups. Do not publish identifiable “non-mover” lists for funders as a substitute for explaining aggregate uncertainty or missingness.
Common mistakes
Reporting completion as impact. "62 completed" describes delivery. Keep the count with a clear definition, alongside the outcomes you intended to examine.
Ignoring instrument changes. Keep measure versions and explain whether comparisons remain defensible. Identical wording is useful but not sufficient if the population, timing or administration changed.
Treating attribution answers as a causal estimate. Ask about perceived contribution and other influences, but use an appropriate evaluation design for a causal claim.
Publishing an average without its coverage. Report the eligible group, matched count, missingness and variation. A paired mean can be useful; it should be reproducible from the underlying records.
Ending collection before the intended outcome. Plan later observations where needed and follow the agreed contact procedure. Declined or missing follow-up stays visible rather than being assumed positive or negative.
What you have now
You now have exit questions aligned to the relevant baseline, participant explanations separated from causal claims, and a cohort summary with coverage and limits. Keep completion and credentials as useful facts, but do not let them substitute for the outcome you intended to examine.
The one thing to do this week
Compare three sample participant records this week: one with a valid pair, one with missing baseline and one with a changed instrument. Have a colleague repeat the calculations and classification. Resolve disagreements before using the report for the whole cohort.
Who this is for
Program leads preparing outcome reports, evaluators joining intake and exit records, and teams that need to explain both observed changes and gaps in their evidence.
Exercise: write a defensible cohort finding
Using the fictional table, write: “Among 30 participants with comparable intake and exit confidence scores, mean change was +2 points. These 30 represent 75% of 40 eligible starters. Twenty scores increased, six were unchanged and four decreased. Results describe observed self-reported change and do not establish causation.” Add the missing-data reasons you actually know; do not invent them.
Compare the same people at both points when reporting paired change. Subtracting an intake mean for 40 people from an exit mean for 30 can mix changes in who responded with changes in scores. For more survey-design detail, use the pre-and-post survey guide.
For the reporting format, use How to Write an Impact Report and the report examples. Keep the chapter’s findings and limitations together when adapting them.
Frequently asked questions
Is completion an outcome?
It can be a relevant attainment or delivery measure, depending on the program. It does not automatically show change in skill, employment or wellbeing. State what completion means and report it alongside the outcomes the program aims to affect.
Does no measured improvement mean the program failed?
No. Consider the intended goal, starting point, instrument sensitivity, timing and missing data. Stability may be a desired result, or the measure may miss a relevant change. Investigate rather than automatically marking a person or program unsuccessful.
Can we report average change?
Yes. For paired change, calculate the difference for each comparable pair and summarize those differences. Include the matched count, eligible group, coverage and distribution. An average can be accurate while hiding substantial differences between participants.
Does participant attribution establish causation?
No. It records the participant’s account of contribution and other influences. That is valuable contextual evidence, but it does not observe the alternative outcome without the program or remove selection bias.
What should happen to missing pairs?
Keep the reason where known and report the number excluded. Do not fill missing starting values from later answers. Consider whether those with usable pairs differ from those without them before generalizing.
Can we compare completers and non-completers?
You can describe their observed outcomes with clear definitions and coverage. Do not assume the difference is caused by completion: the groups may differ in starting circumstances, participation and other influences.
Watch the training-record workflow
4 minutes 13 seconds · Sopact demonstration using synthetic training records.
▶ Play: Training Program Data
Related support-review practice: How to Use Mentor Notes to Review Participant Support — use dated notes to examine participant experiences and follow through on support actions.