Additional course links
By Sopact Academy · Updated September 12, 2026. Figures and participant examples in this lesson are fictional.
This lesson is for training, coaching, fellowship and service teams building intake from an agreed collection plan. Bring the measure-and-source plan from the related lesson, your application form and the proposed follow-up questions. Leave with a field list, a baseline-reuse decision and a check of each planned comparison.
Practical outcome: a short intake form that captures the required starting observations and enough context to plan support. A matched record makes comparison possible; it does not make a causal impact claim true.
Build the baseline in four steps
- Define the starting observations required for the outcomes.
- Check which application answers can be reused or need updating.
- Keep structured answers and narrative context linked to their sources.
- Test baseline and follow-up questions together before launch.
This lesson uses a fictional training cohort to connect intake, mid-program check-ins and exit. If your process includes selection, application review is a separate decision. Do not assume the answers used to select a participant are automatically suitable for measuring their starting situation.
You can design and test the form in a document or spreadsheet. A connected platform helps collect repeated submissions, retain source material and apply configured analysis. Review the form with the people who will complete it as well as the staff who will interpret the answers.
What makes an intake answer a usable baseline?
Picture reporting season. The exit survey says confidence averages 7.4. The funder asks the obvious question: up from what? To interpret that number as change, you need a compatible starting measure. Repeating an appropriate question and scale is one way to get it; a justified mapping from another instrument may also work. Registration details alone cannot supply a missing confidence baseline, and a dictionary cannot turn an incompatible measure into one. The program can still report what it observed at exit while stating what change cannot be assessed.
An intake form can capture necessary registration details without measuring a starting outcome. That is a design gap only when the team needs to assess change in that outcome later. Give each field a purpose: enrollment, service planning, outcome measurement, a reporting requirement or another explicit decision.
Registration identifies the enrollment; a baseline records the relevant starting condition. Some characteristics also change and need dated updates. Store a stable participant identifier separately from contact details and the program enrollment. Email addresses can change or be shared; use them for contact or carefully checked matching rather than assuming they are permanent unique identities.
Step 1 — Define the starting observation
For each planned outcome, define the minimum useful starting observations. A skill outcome may need an assessment rather than a confidence score. Employment quality may need employment status, hours, wage and stability. If an established instrument is appropriate, follow its guidance rather than replacing it with a convenient homemade scale. Name the later collection moment and what comparison you expect to make.
Paste this into any AI:
Here are the outcomes, population and collection plan: [paste them]. Propose baseline questions only where a starting observation is needed. For each, state the construct, wording, response scale, timing, follow-up question, source and missing-data treatment. Identify existing instruments that require expert review; do not invent validation. Flag questions that may be sensitive or affected by selection. Explain how each answer will inform a decision.
Keep comparisons interpretable. Preserve wording and scales where the same construct is measured, and document justified instrument changes. Choose open questions for the context you need; there is no universal two-question limit. Ask about goals and barriers when they support the program, but avoid requesting sensitive detail without a clear use and appropriate access arrangements.
Step 2 — Decide what to reuse from the application
Review the application before adding another form. Reuse an answer only if it was collected at the right time, remains current, measures the required construct and can be linked correctly. Some answers need confirmation after admission, when the purpose is support planning rather than selection. Keep both dated observations if circumstances changed.
Here are my current questions and baseline requirements: [paste both]. Mark each proposed field REUSE, VERIFY or ADD, with reasons covering timing, wording, scale, population and purpose. Identify duplicated questions, but do not remove operational or required fields without an owner checking them. Use a stable internal participant identifier plus an enrollment identifier; do not require an email address as the only way to join records.
A separate intake can confirm changed circumstances and collect information deliberately excluded from selection. If an applicant reported $14 per hour in July and $15 in September, the answers may both be correct. Keep the dates, identify which observation represents the program start, and ask for clarification only when needed.
Step 3 — Keep narrative answers alongside the measures
Keep the original answer when coding narrative text into categories. In a configured Sopact workflow, analysis can prepare themes and source references as submissions arrive. Treat suggested categories as aids for staff review, particularly when wording is ambiguous, translated or sensitive. Confirm any notification or routing behavior during setup; do not treat AI classification as an emergency-response service.
What that looks like for one arriving record:
| What arrived | Suggested interpretation to check |
|---|---|
| "My car finally died and the bus only runs twice a day; daycare runs $180 a week I don't have" | Transport one possible theme, childcare and financial as additional possible themes — each pinned to the participant's exact words |
| Confidence self-rating — 3.6 on 1–10 | Retain as this person’s starting self-rating with its question and date; repeat a suitable measure only where the comparison needs it |
| "Support my kids without working two jobs" | Possible goal about supporting the family with less pressure from multiple jobs; confirm the participant’s meaning before coding or assessing attainment |
| Whole record | Suggested staff review SUPPORT DISCUSSION — staff confirm transport and childcare needs before agreeing a referral |
The example contains several possible barriers: transport, childcare and affordability. A classification should retain the quote and allow more than one theme. Staff ask what support the person wants and record what was agreed. A barrier is not a prediction that the person will leave the program, and a tag does not mean a referral has happened.
Step 4 — Test the before-and-after pairs
Check the form and planned follow-up instruments together. This can be done manually if the reviewer has the questions and definitions. A software-assisted check is useful only when the relevant versions are available to it. Test identity matching, timing, wording, scales and missing values before collecting a full cohort.
- "List every baseline measure and the wave that re-asks it" — the pairs your change claims will stand on.
- "Flag any measure with no later wave" — check whether a later observation is needed for the intended comparison; retain fields with other legitimate uses.
- "Flag any later measure with no baseline" — identify where a starting observation is required; an exit-only measure may still answer a different question.
- "Confirm every record carries the persistent ID" — check matching along with timing, definitions and missing answers
| Baseline measure | Later wave | Same scale? | Verdict |
|---|---|---|---|
| Confidence, 1–10 | Mid and exit, same wording | Yes | Separate fictional matched-record example: 4.3 → 7.1 → 7.4; check question, timing and coverage |
| Hourly wage, $/hr | Six-month follow-up | Yes | Potential wage comparison: $9.96 → $25.11; check currency, gross/net basis, work context and observation dates |
| Public assistance, Y/N | Six-month follow-up | Yes | Comparable yes/no wording may describe status; retain eligibility and policy context before interpreting change |
| Skill self-rating, 1–5 | Exit asks skills as 0–100 | No | FLAGGED scale drift — no comparable pair |
A mismatch found before launch may be straightforward to correct. After collection, changing the scale label does not make different questions equivalent. Preserve the original observations and document whether a defensible comparison is possible. The figures in the example are fictional and do not establish improvement or causation; the audit checks comparability, not program effectiveness.
Common mistakes
Confusing registration with baseline. Keep registration fields that serve a real purpose, then add suitable starting observations where you need to examine change. Not every outcome needs a self-rating, and qualitative baseline information can be useful when collected systematically.
Changing the instrument without a plan. Record versions and assess comparability. A justified change may improve measurement, but do not conceal it or compare raw values as though nothing changed.
Reusing an old answer without checking it. Reduce needless repetition while allowing participants to update changed circumstances. Retain the observation date and distinguish a correction from a new observation.
Collecting narrative answers without a review owner. Decide who reads or checks the classifications, how quickly they should respond, and how an agreed action is recorded.
Relying on email alone. Use a stable participant ID and an enrollment ID. Keep duplicate resolution and correction procedures; an unlinked record may be recoverable, but should not be silently assigned to a guessed person.
What you have now
You now have fields with named purposes, dated starting observations, contextual questions that support decisions and a record-matching plan. Keep an audit of the proposed follow-up questions and assign a reviewer to unresolved mismatches. This supports a later comparison without promising that the comparison alone demonstrates impact.
The one thing to do this week
Test one outcome end to end this week. Ask two colleagues to interpret three de-identified sample records using your definitions. If they select different baseline dates or disagree about missing answers, fix those rules before adding more questions. Then test the form with a representative participant and record where the wording or navigation caused difficulty.
Intake form worksheet: specify each field before building it
This fictional training-program worksheet illustrates the design choices. Adapt the measures to your program; the confidence item is not a validated assessment of skill.
| Field | Purpose and question | Handling rule |
|---|---|---|
| Participant and enrollment IDs | Link this submission to the person and program episode | Assigned by the system or staff; check duplicates without exposing internal identifiers unnecessarily |
| Observation date | When does this answer describe the participant? | Distinguish the observation date from the date staff entered it |
| Current employment | Are you currently doing paid work? Yes / No / Prefer not to answer | Confirm the program’s employment definition and record missing or declined answers separately |
| Hourly wage, if applicable | What is your usual gross hourly rate in the specified currency? | State units; handle salaried or variable work explicitly rather than guessing a rate |
| Job-search confidence | How confident do you feel about carrying out the job-search tasks described by this program? | Use the same explained scale at follow-up; self-report is distinct from demonstrated competence |
| Support context | What could make participation difficult, and what support would you like to discuss? | Explain who reviews it and provide another way to discuss sensitive needs |
| Contact preference | How may the team contact you about this program? | Separate contact details from the stable participant ID; allow updates |
Make the form usable before asking people to complete it
The W3C forms tutorial recommends clear labels, instructions and feedback, and avoiding unnecessary questions. Show which fields are required in words, explain units and provide helpful error messages. Test keyboard navigation and mobile completion. Do not use a placeholder as the only field label.
Explain what the data will be used for, who can access it and how a person can correct an answer. Decide which sensitive fields can be optional and how staff handle requests for help. A required question with no suitable answer option can force inaccurate data.
Watch how baseline and follow-up records connect
4 minutes 13 seconds · Sopact demonstration using synthetic training records.
▶ Play: Training Program Data
Frequently asked questions
Can an application form serve as the intake baseline?
Sometimes. Check that the answer was collected before the relevant services, remains appropriate for the starting point and uses comparable wording and scales. Selection incentives and changed circumstances may justify confirmation at intake. Preserve dates rather than automatically replacing one answer with another.
Is an email address a reliable participant identifier?
It is useful contact information, but it may change or be shared. Use a stable internal participant identifier, with separate program enrollments and dated observations. Confirm matching and duplicates rather than treating the same email as proof of the same person.
Must every baseline measure be a number?
No. A structured description, observation or interview can document starting conditions when the method fits the question. Define how it will be collected and compared. A numeric self-rating should not be substituted for evidence of a different construct just because it is easy to chart.
What if the baseline is missing?
Mark it missing and assess whether existing dated records provide a suitable alternative. Do not infer a past answer from a later result. Report the limitations and the number of participants with comparable observations. Plan better collection for the next cycle.
Can we change questions after the program starts?
Yes when there is a reason, but save the original version and explain the change. Determine whether comparisons remain defensible. Adding a new question does not retroactively create a baseline for earlier participants.
Does a matched pre-and-post record prove impact?
It can describe observed change for the matched participants. Other influences, who responded and how outcomes were measured still matter. Use an evaluation design appropriate to any causal claim and report missing follow-up clearly.
Next in the series: How to Spot At-Risk Participants Mid-Program — the baseline you just captured becomes the reference line, and the mid-program wave reads every record against it while the team can still adjust support.