Customer Interview Synthesis

Paste your interview notes or transcripts. You get the analysis: recurring themes with the count and the verbatim evidence, the jobs to be done underneath them, contradictions rather than a smoothed average, a ranked opportunity list, and an honest note on what the sample cannot support.

A Markdown synthesis containing: (1) a one-paragraph plain answer to the research question; (2) themes ordered by support, each with a count out of total participants, two or three verbatim quotes, and the decision implication; (3) jobs to be done as outcomes; (4) a 'Where participants disagreed' section that preserves contradictions; (5) 'What surprised me' against the stated hypothesis; (6) a ranked opportunity list with its evidence; (7) 'What this sample cannot tell you'. No percentages anywhere.

# Synthesis — why teams abandon onboarding (n=6)

## The answer
The data import is not the problem. One participant of six mentioned the import as a blocker, and it was a missing file rather than a hard step (P1); one explicitly said the import worked (P3). What actually stops people is that onboarding asks them to make commitments — to a colleague, to IT, to their team — before they have evidence the product does the thing they came for. Five of six abandoned at a moment of social or organisational exposure, not at a moment of technical difficulty.

## Themes

### 1. Onboarding demands a commitment before it delivers proof (4 of 6)
> P5: "I did the import, then it asked me to invite my team, and I was not ready to tell my team I was trialling something."
> P1: "I would have had to ask our ops person."
> P4: "I needed our IT team to approve it and that takes three weeks."

In each case the blocking step required involving another person before the participant had anything to show them. P6 is the same pattern in a different form: they completed setup hoping value would emerge, and it did not, so there was never a moment worth telling anyone about.

**Implication:** the sequence is wrong, not the steps. Something demonstrable has to happen before the first step that involves a colleague.

### 2. Nobody could tell what the product would do for them (2 of 6, and arguably 3)
> P6: "I could not work out if it did the one thing I needed. I did the setup hoping it would become obvious and it did not."
> P3: "I could not tell what would happen to the data once it was in."

P6 is about product value and P3 is about data behaviour, but both describe completing steps without a model of what they were building toward.

**Implication:** onboarding is a sequence of tasks with no stated destination. Participants could follow it and still not know what finishing would give them.

### 3. Data visibility is an unanswered blocker (1 of 6, weighted up)
> P3: "Nobody could tell me if it was going to be visible to the whole workspace."

Only one participant raised it, so treat the count honestly. It is listed as a theme because P3 both completed the import and then stopped anyway — an abandonment caused entirely by an unanswered question, which is the cheapest category of loss to fix.

### 4. Some sign-ups were never going to activate (1 of 6)
> P2: "Honestly I signed up to see the pricing after the trial. I was never going to set it up that day."

**Implication:** part of your abandonment metric is not a funnel problem. Before optimising, separate intent-to-evaluate from intent-to-browse, or you will chase a number that includes people you never had.

## Jobs to be done
- **Find out quickly whether this solves my problem, without telling anyone I am looking.** The strongest job in the set. Every commitment step works directly against it.
- **Avoid exposing my team or my data to something I have not evaluated.** P3 and P5, from different angles.
- **Get to a demonstrable result I can take to the people whose approval I need.** P1 and P4 both needed something to justify the ask, and had nothing.

## Where participants disagreed
- **On the import itself.** P1 was blocked by it, P3 and P5 completed it without comment. This is not noise; it is a segmentation signal. Participants with the file to hand were unaffected, which points at a data-readiness difference between accounts rather than a usability difference in the step.
- **On whether stopping was a decision.** P4 and P1 describe drift — losing the thread, forgetting. P2, P3 and P5 describe a decision. These require different interventions: re-engagement for the first group, an answered question for the second. Averaging them into one 'abandonment' metric hides both.

## What surprised me
**The hypothesis was wrong, and confidently so.** You went in believing step 4, the data import, was the cause. Of six participants, one was blocked by it and that was a missing file rather than the step; one volunteered that it worked fine; three never mentioned it. Had the interview guide asked about the import directly, this synthesis would probably have confirmed your hypothesis and sent you to fix the wrong thing.

The second surprise is that the real pattern — commitment before proof — is invisible in funnel data. Steps 4 and 5 look like a usability drop-off. They are a social one.

## Opportunities, ranked

**1. Move the team-invite step behind the first demonstrable result.** Directly supported by P5, and consistent with P1 and P4. It is a sequencing change rather than a build, so it is the cheapest item on this list and the best supported.

**2. Add a sample-data path so a user can see the product working without their own file.** Supported by P1 directly and by P6 indirectly. Removes the dependency on both the file and the colleague who has it.

**3. Answer the data-visibility question inside the import step.** One participant, but a complete loss caused by an unanswered sentence. Cost to fix is close to zero.

**4. State the destination at the start of onboarding.** P6's abandonment is a direct consequence of its absence. Weaker evidence than the first three and a larger design change.

**5. Separate browse-intent sign-ups in the metric.** P2 only, and it is measurement hygiene rather than an improvement — but until it is done, every number about this funnel is contaminated.

## What this sample cannot tell you
- **How common each pattern is.** Six participants, all of whom answered an interview request. Six of six could be a real pattern or an artefact of who agrees to talk. Everything here is a hypothesis with a count attached, and the counts are counts, not rates.
- **Anything about people who completed onboarding.** You interviewed only abandoners, so you cannot know whether completers faced the same commitment steps and pushed through, which is the single most useful comparison available and it is missing.
- **Whether the sequencing change would work.** These interviews explain the abandonment; they do not validate the fix. Ship opportunity 1 behind a flag and measure it.
- **Anything about accounts that never reached step 3.** Different population, probably different reasons, not in this sample.