At a Glance
Teams often finish interviews with notes full of good quotations and no clear path from them to a recommendation. This page follows one small, fictional set of interviews through four steps so each claim can be traced back to the words that support it and the words that do not. For the analytic choices behind the steps, see choosing a qualitative analytic tradition.
At a Glance
| Claim (Fictional) | Memos | Codes | Supporting Excerpts | Disconfirming or Limiting |
|---|---|---|---|---|
| C1: A clash between the 6 p.m. start and work or care is common in these interviews, but does not explain every departure | M1 | TIM, CARE | E1, E3, E6 | E2 (stayed despite the clash), E5 (left with no clash) |
| C2: Support at the session may offset a timing clash | M2 | SUPP, TIM | E2 | One case; no other interview speaks to it |
| C3: A second route to leaving is difficulty with the material and embarrassment | M3 | DIFF | E5 | One case; how common is unknown |
| C4: Moving the start later may create a travel-safety problem | M4 | SAFE, TIM | E4, E8 | Two cases; no one asked about a specific later time |
Worked Case
Worked Case: Evening Learning Circles
Everything in this case is invented for illustration. The program, the people and the quotations are fictional. They are not participant evidence, and nothing here describes a real program or country.
A community organization runs evening learning circles for adults, starting at 6 p.m. Staff see that some participants stop coming in the first weeks. In six weeks, the organization must decide how to change next term's schedule. The evaluation question: why do some participants leave in the first month, and what could the program change?
A purposive set of six participants was interviewed (P1 to P6): four who left and two who stayed. The analyst reads each interview and gives each meaningful passage an ID. Eight passages matter for this case.
| ID | Participant | Status | Excerpt (Fictional) |
|---|---|---|---|
| E1 | P1 | Left in week 3 | "The circle starts at 6. My shift at the market ends at 6. By the time I arrive they are on the second exercise and I feel behind." |
| E2 | P2 | Stayed | "I come straight from work too and I'm usually late, but the facilitator sits with me and catches me up. That is why I stay." |
| E3 | P3 | Left in week 2 | "My sister watches the children until 5:30. After that I have to take them with me, and the room is not for children." |
| E4 | P4 | Stayed | "I have no children at home now. For me the only problem is the walk back in the dark." |
| E5 | P5 | Left in week 4 | "Honestly the timing was fine. I left because the reading level was too hard and I did not want others to see." |
| E6 | P6 | Left in week 2 | "I left after the first week because of the baby. A neighbor said a circle in the next village starts at 6:30 and I thought about it." |
| E7 | P3 | Left in week 2 | "If there was somewhere for the children I would come back tomorrow." |
| E8 | P1 | Left in week 3 | "A later start might help, but then the walk home is after 8 and that frightens me." |
Step 1: Read and Code
The analyst reads all six interviews before coding, then applies a codebook (version 0.1) with two deductive codes taken from the program's own assumptions: TIM (session timing conflict) and CARE (care responsibilities). See deductive vs inductive coding.
Three passages raise something the two codes do not capture. E4 and E8 mention the walk home in the dark; E5 describes difficulty with the material. The analyst does not force them into TIM or CARE. Codebook version 0.2 adds SAFE (travel safety after dark), DIFF (content difficulty and embarrassment) and SUPP (facilitator support, prompted by E2). Each new code records the date and the excerpt that prompted it. The earlier transcripts are then reread for passages the first version had no place for.
| Excerpt | Codes Applied |
|---|---|
| E1 | TIM |
| E2 | TIM, SUPP |
| E3 | CARE, TIM |
| E4 | SAFE |
| E5 | DIFF (the speaker says timing was fine; recorded as a case where TIM is absent) |
| E6 | CARE, TIM |
| E7 | CARE |
| E8 | TIM, SAFE |
The codebook definitions are in qualitative codebook. A code label is not a finding. TIM means only that the passage speaks about the start time.
Step 2: Write Memos
Memos record reasoning between the excerpts and the claims. See analytic memos.
- M1, after E1 to E3: "Both work and care end-times fall close to the 6 p.m. start. A shared window, roughly 5:30 to 6? E1 emphasizes feeling behind, not only being late. Not yet a claim."
- M2, after E2: "P2 has the same clash and stays. The facilitator catches him up. Does support matter more than the timing? One case only."
- M3, after E5: "P5 left though timing was fine. This is a second route. P4 stayed with no timing problem at all, only safety. The timing story does not cover both."
- M4, after E8: "A later start is the obvious fix for E1, E3 and E6, but E4 and E8 raise the walk home after dark. A later start could solve one problem and cause another."
M1 also carries a reflexive note: the analyst expected a scheduling problem because the program's own staff had said so. That expectation may have steered early attention. See positionality and reflexivity.
Claims and Limits
Step 3: Search for Cases That Do Not Fit
Before writing claims, the analyst states the working one: "People leave because the 6 p.m. start conflicts with work or care." Then asks what would contradict it. Two kinds of case would: someone who left without a timing problem, and someone with a timing problem who stayed.
E5 is the first kind. E2 is the second. Neither can be dropped as an exception. Both change the claim, as the negative case analysis page explains. E4 is a different kind of surprise: a participant who stayed and raised a barrier nobody planned for. It matters for the recommendation, not for the original claim.
Step 4: Build Claims with a Trail
Each claim is written with the excerpts, codes and memos that support it, and the material that limits it, as in the At a Glance table. Reading C1 in full: "In these six interviews, four people (P1, P2, P3, P6) described a clash between the 6 p.m. start and work or care, and three of them left. The clash does not account for every departure: P5 left with no such clash, and P2 stayed despite it."
The counts describe these six interviews. They do not estimate how many participants in the program share the experience. The sample was chosen to include leavers and stayers, which is not a random draw.
What the Findings Can and Cannot Say
They can say that timing and care clashes appear in several accounts, that at least one other route to leaving exists, that support at the session may matter, and that a later start brings a new concern. These are plausible leads for the schedule decision.
They cannot say how common each route is, whether fixing timing would reduce leaving, or that these reasons apply in other programs. Six interviews cannot support a rate or a cause. The approach was thematic and open to revision, so no saturation claim is made here; see qualitative saturation.
A recommendation consistent with the evidence: do not announce that a later start will fix leaving. Test several low-cost options in the next term: a catch-up segment or buddy at the start of each session, a way for caregivers to bring children, and a later start paired with a safe-travel arrangement. Ask everyone who leaves one short question about the main reason, so the organization learns how common each route is. The decision stays with the organization.
Using It
Checking the Trail with AI
An AI tool can draft codes or summaries from transcripts. Treat what it returns as a suggestion to trace. For each suggested theme, ask which excerpt IDs it relies on, open them, and check whether the words say what the suggestion says. Ask it to list passages that contradict each theme, and read those yourself. The page on checking AI-assisted qualitative coding runs this check on the same eight excerpts.
Common Mistakes
Quotations without a trail. A striking quote with no ID cannot be checked.
Counting as if sampling were random. "Four of six" describes six interviews.
Dropping the case that spoils the pattern. E2 and E5 changed the claim, and the claim improved.
Turning a code into a finding. TIM labels a passage. It does not explain anything.
Letting memos disappear. If the reasoning was never written, it cannot be reviewed.
Using one tradition's checks on another. Agreement scores and saturation do not apply to every approach. See choosing a qualitative analytic tradition.