Build comparable before-and-after questions, match responses appropriately, and report change without overstating what it proves.
A pre and post survey asks about a relevant measure before and after an activity or period to examine change. A pre survey establishes the starting point; a post survey repeats comparable measures at a specified follow-up point.
When the same people answer both times and their responses can be linked appropriately, you can analyze individual change. When responses are unlinked, you can still compare the groups observed at each point, but you cannot say how a particular person changed. Neither design alone proves that the activity caused the change.
| Moment | Purpose | Example |
|---|---|---|
| Pre survey | Establish starting knowledge, confidence, experience or conditions. | Before a course, ask about confidence in a clearly defined task. |
| Immediate post survey | Examine change that could plausibly occur by the end. | Repeat the confidence question and assess the relevant skill. |
| Later follow-up | Examine application, persistence or a delayed outcome. | Ask whether the skill was used and what enabled or prevented use. |
Satisfaction with a course is different from learning. CDC recommends distinguishing learning from the transfer of learning and notes that a skills demonstration may be appropriate alongside knowledge assessment. Read the training evaluation guidance.
Use an appropriate validated measure when the decision requires one. The question examples below are adaptable wording, not validated scales. Pilot them with people like the intended respondents.
Choose the questions that fit the activity rather than using all 30. For a self-assessment, define a consistent response scale—for example, from “not yet able” to “able independently”—and explain it. Self-reported ability is not the same as observed performance.
Use the same reference period where possible. For a knowledge test, check that the item tests the intended learning rather than familiarity with the question. Some open questions are useful at both points; others should be post-only and clearly labeled.
| Field | Before | After |
|---|---|---|
| Linking code, where justified | Assign or collect a suitable stable code. | Use the same code through a secure process. |
| Context | Relevant group, date and activity stage. | Same definitions, updated date and stage. |
| Core measures | Ask the selected baseline questions. | Repeat comparable questions and scales. |
| Explanation | Ask about current barriers or expectations. | Ask about changes, barriers and application. |
| Post-only feedback | Not applicable. | Ask about usefulness and suggestions separately. |
Keep the identity lookup separate from the analysis dataset where appropriate. An email address can change and may not be necessary. Anonymous surveys can be legitimate; do not promise anonymity while retaining a link that identifies the respondent.
Validate the match first. Check duplicates, missing values, score direction and whether all items belong to the same measure. Then calculate each matched person’s change for measures where that calculation is meaningful.
| Illustrative participant | Pre score | Post score | Change |
|---|---|---|---|
| A | 2 | 4 | +2 |
| B | 3 | 3 | 0 |
| C | 4 | 3 | −1 |
| D | 2 | 3 | +1 |
The pre mean is 2.75 and the post mean is 3.25, a mean change of 0.5 among these four matched respondents. Two improved, one stayed the same and one declined. That distribution says more than the mean alone.
For ordinal scales, consider distributions and medians as well as any justified numerical summary. Choose statistical methods that fit the measure, design and sample; do not apply a paired test to unrelated groups. Report uncertainty where appropriate.
Report the pre and post groups separately and state that the comparison is unpaired. You can describe their results, but a difference may reflect changes in who responded as well as changes in the underlying population.
Do not delete unpaired records merely because they cannot enter a paired analysis. They may help describe coverage or nonresponse. Keep matched and unmatched results distinct, and explain the basis of each denominator.
Where feasible, compare the characteristics of people with and without follow-up. A high response rate is useful, but it does not automatically remove nonresponse bias.
A change after training is consistent with learning, but it may also reflect other experiences, changes in measurement or regression toward the mean. Stronger causal questions may require a comparison group or another suitable evaluation design.
A retrospective pre/post question asks people at the end to assess both their earlier and current state. It can be useful in some circumstances, but recall and changed understanding can affect answers. Label it accurately rather than presenting it as an independently collected baseline.
When the goal is behavior at work, add suitable follow-up rather than treating an immediate confidence increase as evidence of lasting application.
Keep the question version, respondent or group context, date and source with the result. Link relevant open responses, assessments or follow-up interviews only where the permissions and study design support it.
AI-assisted classification can help organize explanations—for example, lack of time, equipment or opportunity to practice. A reviewer should check the classification and source before presenting a pattern. Avoid inferring a cause from one comment or labeling a person as a risk without sufficient evidence.
Continue to survey report examples for reporting methods, denominators and limitations clearly.
Use these existing Academy guides for the practical next step. They are suggested companion readings; follow each guide’s course navigation for the full sequence.
Explore the Loop methodology for the ongoing cycle of collection, analysis and improvement.
Keep core change measures comparable, including wording and scales. Post-only questions about satisfaction or application can be added separately.
It can support a valid design, but unlinked responses cannot show individual change. Explain the design and protect any linking information.
Choose timing based on the expected change. Immediate learning and later workplace application require different observation points.
A survey can ask about perceptions or experience; an assessment may test knowledge or demonstrated performance. Use a measure suited to the intended claim.
Report the invited or eligible population where known, responses at each stage, matched sample, item-level missingness and relevant limitations.
It shows measured change in the observed sample. Attribution to the program requires consideration of other explanations and an appropriate evaluation design.
