How do you track the same participant over time?

Use a stable, appropriately protected identifier to connect dated observations, while keeping question versions, sources and follow-up status visible. Plan how you will recognize returning participants, handle uncertain matches and preserve useful records when staff, contact details or systems change.

Use this reference when your collection plan follows a person through repeated observations or several service episodes. Design a participant record and test what happens when someone misses a wave, changes email address or returns through another service.

This chapter focuses on the collection and record design. For statistical comparisons, use the separate lesson on analyzing longitudinal survey data. A well-organized record makes analysis possible; it does not remove the need to choose an appropriate method.

Define what the continuing record needs to answer

Start with the decision and time horizon. Are you checking whether a training outcome remains present after six months, following a member's experience across renewals or understanding a person's changing support needs? Those purposes require different timing, measures and access.

Write down the unit you follow. One person may have several enrollments, service episodes or memberships. If they return for a second program, retain the person identifier and create a separate episode identifier. Otherwise, the new intake may overwrite the first, or the same person may be counted as two people.

Distinguish a panel that follows the same people from a repeated survey of whoever participates each year. Both can be useful, but they answer different questions. A stable annual average among different respondents is not evidence that each person's experience remained stable.

The UK Data Service's longitudinal-data guide provides a practical introduction to inspecting and managing repeated observations. Keep the identifier and the observation period explicit before comparing records.

Agree access, continuity and responsibilities early

Before committing to a collection arrangement, establish who can access the records, add a question, export the information and continue the follow-up if a supplier or staff member changes. Record responsibilities for security, corrections and retention. These are operational questions as well as contractual ones.

“We own the data” is not enough detail. Ask whether an export includes stable IDs, dates, question definitions, attachments, relationships and review history. A spreadsheet of answers without those elements may be difficult to reuse. Conversely, working with an outside research provider does not automatically prevent a long study; the arrangement and permissions matter.

Define the purpose and the applicable arrangements for recontact, linkage and use of the information. A choice not to receive future invitations is not automatically the same as a request to erase all prior observations. Route changes to the responsible owner and apply the relevant policy and obligations.

Use the course lesson on access and sharing boundaries to separate operational contact details from analysis and report audiences. Do not publish a named list of people who missed a wave. Authorized follow-up staff may need that list; readers of an aggregate report usually do not.

Choose an identifier approach you can support

For a service relationship where the organization already knows the person and linkage is appropriate, an issued ID carried through a secure invitation can reduce the need to re-enter identifying details. Keep the mapping to contact information protected and separate where appropriate.

Issued IDs still need a recovery process. Invitations can be forwarded, links can be opened by someone else, contact details can change and duplicate records can be created. Test those cases instead of assuming a code guarantees correct identity.

Self-generated identification codes are another research approach. They can support matching where direct identifiers are not collected, but code consistency and collisions need evaluation. They should not be dismissed as always unusable or presented as automatically anonymous.

A 2025 cohort study of self-generated identification codes matched approximately 72.65% of its school-based sample using perfect and partial matching. That result demonstrates both potential and loss: it is evidence from one study, not a benchmark for your population or a recommendation to copy its code design.

Choose according to the purpose, participants' needs and the collection promise. Do not quietly convert an anonymous survey into an identifiable record to make analysis easier. If linkage is not intended or justified, analyze the data at an appropriate aggregate level.

Separate the person, the episode and the observation

Think in three linked parts. The person record identifies the continuing relationship. An episode describes a particular program, membership or service period. An observation records what was collected, when, from which source and under which version.

You may implement these parts in linked tables, an approved database or another suitable system. A wide spreadsheet can be workable for a small fixed study; a long-format table often helps with repeated observations and changing schedules. The essential requirement is a documented relationship, not a universal prohibition on files or columns.

Keep attachments and transcripts attached to the right observation or episode, not just somewhere under a person's name. A document's upload date can differ from the date of the event it describes. Record both when the distinction matters.

Worked example: one person, two program episodes

The following record is fictional. P-024 joins a training program in January, misses a July follow-up and returns for a different course the next year. The first program history remains available.

The record supports a +2 change from January to April on the stated measure and a score of 3 in October. It does not show what happened in July or prove continuous improvement between checks. The new skill measure belongs in its own series unless a defensible comparison is established.

A team reviewing this record can ask a focused question: what was reported after the first program, and what information is missing? It should not infer a successful July outcome simply because the person returned later.

Make timing part of the definition

Record actual dates as well as labels such as “six-month follow-up.” A response collected at month eight is not identical to one collected at month six. Establish an acceptable collection window and say how late responses are handled.

Decide whether analysis uses calendar time, time since intake, time since exit or another meaningful clock. Programs with rolling enrollment often need elapsed time. A common reporting quarter can still be useful, but it should not conceal different lengths of participation.

Keep scheduled waves distinct from unscheduled notes or service events. A note written during an urgent call is useful evidence, but it is not necessarily a completed follow-up survey. Avoid counting it as one merely to fill a gap.

Changes to cadence are possible. Record when they occurred and why, and adapt the comparison accordingly. Planning early reduces avoidable problems, but it is too strong to say that every decision becomes impossible to repair after the first wave.

Version questions without discarding the history

Maintain core measures where comparability is important, and document additions and revisions. New questions should appear as “not asked” in earlier waves. A blank caused by a skipped response is a different status.

Exact wording is useful but not the whole measurement context. Language, response options, collection mode and interpretation can also change. Conversely, a carefully evaluated revision may retain useful comparability. Record the decision rather than treating any edit as either harmless or fatal.

Use the question-versioning lesson to create a change log. Keep the original answers and definitions available. Do not overwrite historical values to make an updated form look consistent.

Review uncertain matches instead of forcing them

Define what counts as an exact match and what requires review. A returning record with the same authorized ID may be straightforward; a conflicting ID, duplicated invitation or ambiguous name requires more attention.

Keep the unmatched record with a status and source. Do not invent an ID or choose the nearest-looking person simply to raise the match rate. A false match can create a convincing but incorrect history.

When an approved review resolves a duplicate, retain the decision, reviewer and original references. Consider whether a mistaken merge can be reversed. Email addresses and names can help an authorized reconciliation process, but neither should be assumed permanent or unique.

Track match quality separately from response coverage. Receiving twenty surveys does not mean that twenty people have usable paired observations. Some may be duplicates, new entrants, incomplete responses or unresolved matches.

Report missing waves without guessing the outcome

Distinguish not eligible, not yet due, not asked, invited but no response, incomplete and unresolved linkage. Apply categories consistently and avoid collecting sensitive explanations merely to fill an administrative field.

Compare available baseline characteristics of respondents and nonrespondents where appropriate. Differences can reveal a concern, but they do not tell you the missing outcomes. People may stop responding for many reasons; the direction of resulting bias cannot be assumed.

For each comparison, state the usable base and why records were excluded. Follow the survey-attrition lesson for a fuller missingness review. Keep individual outreach lists restricted while reporting suitable aggregate coverage.

Test the workflow before expanding it

Use a fictional twenty-person cohort for a two-wave rehearsal. Create three nonresponses, one duplicate return, one changed contact address and one unresolved identifier. Add a question in the second wave and change a measure's response scale.

1. Check identity: valid returning IDs connect to their history; the ambiguous return remains flagged until review.
2. Check episodes: a new enrollment creates a new episode without deleting the first.
3. Check dates and versions: collection timing and changed measures remain visible.
4. Check missingness: nonresponse, not asked and unresolved linkage remain distinct.
5. Check access: each role sees the information appropriate to its task, including exported records.
6. Check continuity: another authorized colleague can interpret the record and retrieve the source without relying on the original designer.

Document expected and actual results. A vendor demonstration should use these cases in the proposed configuration, including export and correction. In a Sopact evaluation, ask how collection, documents and analysis stay associated with the person and episode. Verify the actual setup rather than assuming a feature label proves the workflow.

Your handover pack should contain the data dictionary, identifier rules, version log, follow-up schedule, access responsibilities and known unresolved records. Review it after meaningful changes. That small amount of documentation helps the record survive beyond one person's memory.

Watch: connect a participant journey from application onward

Watch the video · 4 minutes 13 seconds. This training-workflow demonstration uses illustrative data. It shows the connected-record idea; it is not evidence of a multi-year customer outcome. Browse more videos in the video library.

Frequently asked questions

What is longitudinal participant tracking?

It links dated observations about the same person over time, with the measures, sources and relevant context preserved. This differs from comparing surveys of different respondents each year.

Should participants create their own code?

That depends on the study and privacy design. Self-generated codes can work but may have substantial matching loss. Issued identifiers also need secure delivery, recovery and duplicate checks.

Does a stable ID make the data anonymous?

No. If the ID can be linked back to a person, the record remains linkable. Apply the appropriate access and use arrangements rather than calling it anonymous because the name is hidden.

Can we add new questions later?

Yes. Preserve the historical instrument, label new questions as not asked in earlier waves and assess whether revised measures remain comparable. Do not silently overwrite definitions.

What does a missing wave mean?

It means the expected observation is unavailable under the recorded status. It does not automatically mean the person lost the outcome, left the program or scored zero.

What if a participant returns to another program?

Where linkage is appropriate, keep the person identifier and create a distinct episode. Preserve the earlier history and avoid combining different measures into one trend without justification.

What should survive a system change?

Stable references, dates, instrument definitions, source links, relationships, relevant permissions and unresolved-record notes. Test a usable export and an authorized handover before relying on continuity.

When several people contribute observations

Continue to several people describing one person. That reference adds another distinction: the person being described and the person supplying the observation are not always the same.

Reviewed September 12, 2026. P-024 and the twenty-person rehearsal are fictional teaching examples.