Three detailed interviews may provide better language for an advertising brief than a long internal discussion. They do not automatically reveal what the market usually needs. When research for ChatGPT ads is limited, the team should be precise about what it has learned: a situation, a condition or a possible question the offer can answer.
This is an editorial method for using a small sample without hiding its limitations. It does not prescribe a magic minimum number. The amount of research needed depends on the question, the variation in situations and the strength of the conclusion you want to draw. The examples are fictional, not reported research results.
Map situations instead of reporting only a count
Suppose a business sells training in a workplace tool and has interviewed three former participants. All work in small teams and selected the course themselves. That is a clearly defined sample, but it does not automatically cover new employees sent on training or large groups learning together.
For each interview, record the task the person wanted to accomplish, their relevant experience and how the purchase was decided. Select characteristics that might change the offer’s relevance. A long list of background details is unhelpful if it does not illuminate the current decision.
This map exposes gaps better than a number alone. Three people in the same situation provide different evidence from three people handling very different tasks. Both samples can be useful, but they support different conclusions.
Separate occurrence, conditions and prevalence
| Statement | What it requires | What a small sample can contribute |
|---|---|---|
| This need occurs | A clear, relevant observation | A described example to investigate |
| This need matters in this situation | Relevant details and examined boundaries | A bounded draft brief |
| This need is common in the market | Evidence supporting that generalisation | Usually an open question rather than an answer |
Beginning at the first level is legitimate. The problem is drifting to the third without further evidence. An ad can describe a real use case without calling it the most common or important need for everyone.
Use wording such as “in these interviews” in the internal summary. Do not attach it as a decorative caveat after a strong market claim. The conclusion itself must match the scope of the investigation.
Do not turn two out of three into a market finding
Suppose two of the three fictional course participants needed to prepare a report before a recurring meeting. “Two of the three interviewees described the reporting task” is a description of the material. It does not show that two-thirds of potential customers share that need. The sample already consists of people who chose the course, and they may have been recruited through the same contact.
Check whether the similarity is meaningful. One person may need to build a report for the first time, while the other wants to check an existing template. The same word can conceal different learning needs. Compare the task and prerequisites before combining the answers.
If the interviews and the course’s actual prerequisites support that scope, a useful brief note could say: “Reporting before recurring meetings appears in the material. This draft concerns participants who already know the tool’s basics. Group training and beginners have not yet been investigated.” That statement gives the writer something concrete to use and identifies claims the draft cannot support.
What do two of three interviews establish?
- Observed
Two participants mention reporting before meetings.
Checkpoint - Examine the difference
Building a new report or checking a template?
Checkpoint - Not established
The proportion of the market with that need.
Reconsider
Deliberately look for a different case
If the first participants all describe the same benefit, choose the next interview to examine a meaningful boundary. If the first participants already know the basics, that might mean speaking with someone new to the tool or someone who chose to learn independently. Ask whether the same offer still addresses a concrete task.
This does not turn a few extra participants into a representative survey. It exposes the brief’s idea to a relevant challenge. Noncustomer interviews can provide particularly useful counterexamples when the first conversations only involved satisfied customers.
Decide what a differing answer would change. If the course does not suit the beginner, the brief may need clearer prerequisites. If every possible answer would still lead to the same headline, the question probably has too little influence on the decision.
Improve depth before collecting more voices
A short comment that a course was good can provide weaker evidence than a detailed account of the work a participant needed to perform. Check whether the existing interviews contain concrete events, alternatives and conditions.
Use the customer interview questions to fill specific gaps. A follow-up about an unclear decision may be more useful than another general endorsement. Avoid repeatedly pressing the same person until they confirm the team’s preferred explanation.
Distinguish independent observations from repetition. The same account appearing in an interview, a sales conversation and an internal summary is not three separate customer cases. Traceability in the evidence register helps reveal such connections.
Keep the brief narrower than a market claim
A reasonable conclusion might be that the offer suits people needing to perform a particular task who already possess the necessary basics. You can describe that concretely without claiming that “all teams lack this skill” or that most of the market wants the training.
If the insight informs context hints, preserve it as a factual use-case description. OpenAI’s targeting documentation describes hints as additional context about the offering. They are not exact-match keywords and do not guarantee an ad impression in a particular conversation. More interviews do not change that limitation.
Record the limitation the next writer actually needs to know. Perhaps every interviewee bought individually, or nobody had tried learning the task in a large group. This is more useful than the generic instruction “more research needed”.
Decide what is enough for the next decision
A bounded writing draft may be reasonable when the task is concrete, the offer actually supports it and uncertainty does not have to become a claim. Hold the conclusion when two plausible interpretations would require different offers or prerequisites. A focused follow-up is then needed before the writer chooses between them. When several uncertainties remain, the assumption-risk workshop helps choose which one could change the brief enough to investigate first.
Repeated answers may clarify the particular task described. They do not certify that the sample covers every relevant situation. Record which decision the material was sufficient for and what new information would cause the team to reconsider it. The targeting and relevance guides cover other parts of campaign planning.
Sources and scope
Draw useful but limited conclusions for an advertising brief when customer research contains only a few interviews.
Working methods and examples are editorial suggestions. Check current platform requirements and available features before implementation.