What is a baseline survey?
A baseline survey measures relevant conditions at a starting point so later evidence can be interpreted against them. It may describe people, organizations, locations or another defined population. In a program evaluation, collect it before the relevant activity begins whenever the design calls for a pre-program measure.
A baseline is not automatically a survey: existing records, observations or assessments may provide suitable starting information. A survey is useful when the questions require information respondents can meaningfully supply. The IFRC’s Baseline Basics guide provides further guidance on planning a baseline study.
The most important design question is what you intend to compare later. Following the same participants is one option. Comparing repeated samples of a defined population is another. Those designs support different conclusions and require different data arrangements.
Decide what the baseline needs to support
Start with a short statement: “We need a starting measure of X for Y population, before Z, so that we can review…” Complete the sentence with the actual decision.
Scroll horizontally to see all columns →
| Purpose | Useful starting evidence | Later comparison |
|---|---|---|
| Training evaluation | Relevant knowledge, skill or current practice | Learning and later application, using suitable measures |
| Member-network planning | Member organizations’ current activity and needs | Changes under shared definitions and reporting periods |
| Customer experience | Experience of a defined interaction or customer group | Comparable later feedback with response mix disclosed |
| Program participation | Relevant starting circumstances and goals | Follow-up evidence appropriate to the intended outcomes |
A baseline should not become a request for every potentially interesting field. Collect information that supports the planned decision, interpretation or appropriate follow-up. Explain its purpose and use.
Choose between individual follow-up and repeated group measurement
Follow the same units over time
A panel design follows the same people or other units. It can show within-unit change when records are matched appropriately. Plan identifiers, contact arrangements where needed, timing and treatment of missing follow-up before the first collection.
The unit could be a member organization, school or supplier rather than a person. Use the least intrusive record structure that serves the purpose.
Measure the same defined population at different times
Repeated cross-sectional surveys can describe group-level change even when respondents differ. Keep the population definition and measurement approach comparable, and examine changes in the sample composition.
Do not describe the difference between two anonymous group averages as the change experienced by each person. Equally, do not dismiss anonymous baselines as worthless simply because individual matching is unavailable.
What questions belong in a baseline survey?
Choose measures that can answer the evaluation question. Where a suitable established instrument exists, consider its intended use and requirements before modifying it. The examples below illustrate question purposes; they are not a validated scale.
Scroll horizontally to see all columns →
| Question purpose | Illustrative wording | Design consideration |
|---|---|---|
| Current practice | In the past four weeks, how often have you used this method? | Use a clear reference period and relevant response options |
| Starting confidence | How confident are you completing this task without assistance? | Confidence is not the same as demonstrated skill |
| Relevant experience | Have you previously completed training on this topic? | Ask only if it helps interpret or plan the program |
| Barrier or context | What currently makes this task difficult? | Allow a useful account without assuming everyone has a barrier |
| Goal | What would you most like to be able to do after the program? | Keep the person’s goal distinct from a standardized outcome measure |
You do not need an open-text question after every rating. Ask for narrative where it adds useful context and the team has a plan to review it. A shorter, purposeful survey can be more usable than a long instrument with repeated explanations.
For response formats, see survey question types. Use survey logic when questions apply only to particular groups.
Set the timing and follow-up schedule
Record when the baseline was collected relative to the activity. If participants enter at different times, an entry baseline may be more relevant than one calendar date for everyone. Keep cohort and timing context so later comparisons are understandable.
Plan follow-up around when the expected change could reasonably be observed. A course-end questionnaire may assess immediate learning or experience; sustained practice may require a later measure. Do not add waves simply because the software allows them.
Explain follow-up expectations and permitted contact arrangements. Keep current contact information separate from the outcome measure, with access appropriate to its purpose. A stable identifier does not guarantee someone will respond later.
Define who is included
Specify the eligible population, how people or organizations will be invited and whether you seek a census or a sample. A large response count does not by itself make the result representative.
Track invitations, responses and relevant exclusions. Consider whether language, access or the collection method makes some groups less likely to participate. Report the limitations that matter for the intended use.
Sample-size requirements depend on the estimate, desired precision, design and planned analysis. Do not use a generic target as proof that a small subgroup or a complex evaluation is adequately supported.
A worked baseline and follow-up example
A fictional program has 100 participants with a baseline measure. Seventy provide a comparable follow-up. Among those 70 matched participants, the mean score changes from 40 at baseline to 55 at follow-up: an observed increase of 15 points.
The baseline mean for all 100 was 45. Comparing 45 with 55 would mix different groups and produce a 10-point difference. Both calculations can be computed, but they answer different questions. The matched comparison should use the baseline values of the people represented at follow-up.
- State that the 15-point change concerns 70 matched participants.
- Report that 30 baseline participants lack a comparable follow-up.
- Examine whether those missing differ on relevant baseline characteristics.
- Check that the instrument, scoring and timing support comparison.
- Keep the observed change distinct from a claim that the program caused it.
A before-and-after comparison alone does not rule out other influences. Stronger causal claims need an appropriate evaluation design and analysis. The pre-and-post survey guide develops this measurement problem.
Check the first wave before collecting the next
Test wording with intended respondents and inspect a pilot dataset. Confirm valid ranges, missing-value treatment, routing and relevant record matching. A required field prevents a blank submission; it does not ensure a meaningful answer.
Review possible duplicates and conflicting values. Keep the original response and a record of material corrections. Do not automatically turn missing, not applicable and zero into the same value.
Inspect narrative evidence for context and diversity of experience. A vivid comment can help explain a question to investigate, but it should not stand in for the whole group’s experience.
Use shared measures without forcing identical surveys
Across a network or several sites, agree the small common set of measures needed for comparison. Local teams may collect additional questions relevant to their work.
Document the definition, unit, reference period, scale, eligible population and owner for each shared measure. Collect stable registration details once where appropriate, then update fields that change. This reduces repeated collection while retaining current context.
If a definition or question changes, record the version and assess comparability. A four-point response scale and a five-point scale should not silently appear as one uninterrupted series. Keep genuinely different local measures separate.
What if the program has already started?
Do not label a post-start measure as a true pre-program baseline. Look for relevant existing records, dated assessments or other suitable evidence of the earlier situation. Check their population, timing and definitions.
Retrospective questions may provide useful accounts of an earlier state, but recall can be inaccurate and later experience can affect how people describe the past. Label that source clearly and narrow the claims it supports.
You can still establish a reference point now for future monitoring. Record the exposure that has already occurred and explain what the new measure can and cannot show. Missing historical evidence is a limitation to manage, not a reason to fabricate a starting value.
What to test in baseline survey software
Test collection, follow-up and analysis together. For a panel, check record matching and missing waves. For repeated group surveys, check population definitions and reporting by relevant segments. For either, inspect exports, corrections, permissions and changes to the instrument.
Many survey and data-collection platforms support recurring studies. Compare the actual configuration and maintenance rather than assuming other tools cannot connect waves. Sopact’s focus is self-managed collection, analysis and governance of continuing evidence; evaluate those requirements with a representative workflow.
Start with a small pilot containing a complete pair of responses, a missing follow-up, a changed contact detail and a corrected value. Ask the team to produce one result with its denominator and limitations.
Frequently asked questions
Must a baseline survey follow the same people later?
No. A panel can measure individual change, while repeated samples can describe group-level change. Choose the design that fits the question and explain its limits.
When should the baseline be collected?
Before the relevant activity when a pre-program measure is needed. For rolling enrollment, record timing relative to each participant’s entry. Label a late measurement accurately.
Does a baseline prove impact?
No. It provides starting information. A causal impact claim requires an appropriate design and analysis that considers other explanations for change.
What if follow-up responses are missing?
Report the gap, define the analysis population and examine relevant differences between respondents and nonrespondents where possible. Do not assume those missing had no change.
Can existing data serve as a baseline?
Yes, if its timing, definitions, coverage and quality fit the question. Document limitations and avoid combining incompatible measures.
Do we still need data review?
Yes. Entry validation helps with specific checks but does not eliminate inaccurate answers, duplicate records, mismatched waves or interpretation errors.
