Non-User Research: How to Interview the People Who Never Chose You

The bottom line: Every insight programme has the same structural flaw — it samples customers. The people who evaluated you and picked someone else, who never heard of the category, who solve the job with a spreadsheet, or who want the product but cannot get budget are invisible to in-app surveys, CRM lists, and support tickets. Non-user research fixes that, and it changes different decisions than customer research does: positioning, category education, pricing floors, and which objections your funnel has to defeat. Recruiting is the hard part; running the study is not, especially when an AI moderator can interview whoever shows up, whenever they show up.

---

The most expensive sampling error in product

Ask where your last ten insights came from. Almost certainly: in-app feedback, customer interviews, support tickets, NPS verbatims, win-loss on closed deals, churn surveys. Every one of those sources requires the respondent to have already chosen you.

The result is a well-known distortion — survivorship bias — with a specific product consequence. You optimise relentlessly for the preferences of people who were already willing to say yes, and you learn nothing about the far larger population who were not. Roadmaps built this way get better and better at serving a shrinking, self-similar audience, and the team is genuinely surprised when growth stalls while satisfaction scores rise.

Non-user research is the corrective. It is not a competitor teardown and not an analyst report; it is direct conversation with people outside your customer base about how they actually handle the job today.

---

The four types of non-user

They are not interchangeable. Each requires different recruiting, different questions, and informs a different decision.

1. Rejecters — evaluated you, chose otherwise

They know the category, they knew your name, and they said no. The most immediately actionable segment, because they can articulate a comparison. Beware the polite answer: "too expensive" is the socially easy reason and is frequently a proxy for "I did not believe it would work."

Informs: positioning, pricing, proof and trust gaps, competitive messaging.

2. The unaware — have the problem, never found you

They experience the job-to-be-done and have never encountered a solution like yours, or have never thought of the problem as solvable. The hardest group to interview well, because the moment you describe your product you have taught them the category and destroyed the very ignorance you were trying to measure.

Informs: channel strategy, category education, the language people use before they know your vocabulary.

3. The DIY crowd — solving it with something else entirely

Spreadsheets, email threads, a Notion page, an intern, or nothing at all. Almost always your biggest real competitor, and almost never the one your battlecards name. What matters is not that their workaround is worse, but what makes it good enough — usually zero procurement, full control, and no new tool to learn.

Informs: the switching-cost story, onboarding, integration priorities, and whether the category is actually a category.

4. The constrained — want it, cannot have it

Budget frozen, procurement blocked, security review failed, incumbent contract running for two more years, no internal champion. These people fail your qualification criteria but their reasons are systematically diagnosable.

Informs: enterprise readiness, security and compliance investment, packaging, timing of outreach.

---

How this differs from research you may already run

The distinction matters because win-loss and churn research both start from a record in your CRM. Non-user research starts from a market definition, which is why recruiting — not moderation — is the whole difficulty.

---

Recruiting non-users

You cannot email a list you do not have. Five approaches, roughly in order of quality:

1. Research panels with a screener. Fastest path to the unaware and DIY groups. Write the screener around the job, never the product: "how do you currently keep track of X?" rather than "do you use a tool like Y?" The second version recruits people who will tell you what they think you want to hear.

2. Closed-lost and no-decision records. Your best source of rejecters, and the least used. Reach out three to six months after the decision, when the outcome of choosing something else is known and defensiveness has faded.

3. Communities where the job is discussed. Subreddits, Slack groups, professional forums, LinkedIn. Recruit around the problem, offer a genuine incentive, and be explicit that you are not selling.

4. Paid social with a job-framed ad. Expensive per participant but the only reliable route to people who have never heard of you. Point it at the workaround, not the product.

5. Your customers' colleagues. The people in the same company who did not adopt after a rollout. Adjacent to feature adoption research and unusually easy to reach.

Two recruiting realities to plan for. First, incentives must be higher. Non-users have no relationship with you and no interest in your roadmap; expect to pay noticeably more than for a customer interview, and check the tax and threshold rules before you start issuing payments. Second, screeners leak. People will claim familiarity with a category to qualify for an incentive. Use a knowledge question with a plausible fake option among the real ones, and drop anyone who claims to use it.

---

Interviewing without contaminating the answer

The cardinal rule: learn how they work before you say what you do. Once you describe the product, every subsequent answer is a reaction to your framing rather than a description of their world.

A workable structure:

1. Current-state reconstruction. "Walk me through the last time you had to do X. What did you use? Where did it get annoying?" Concrete episodes only, never habits in general.
2. The workaround's virtues. "What do you like about doing it that way?" Skipping this is the most common mistake in the method. The strengths of the status quo are your actual competitive brief.
3. Cost of the current approach. What breaks, how often, who complains, what it costs in time or money. If nothing breaks, you have learned something important: this person is not a prospect and no messaging will make them one.
4. Trigger conditions. "What would have to happen for you to go looking for something new?" This produces the buying-trigger list your marketing team keeps guessing at.
5. Awareness probe, still without the pitch. "Have you ever looked for a tool for this? What did you find? What made you stop?" For rejecters, this is where the real objection surfaces.
6. Reveal, last. Only now describe the product, and treat everything after as reaction data — useful, weaker evidence.

For rejecters specifically, ask the counterfactual directly: "what would have had to be true for you to have chosen differently?" It reliably outperforms "why didn't you buy?", which invites the rehearsed answer.

---

Running it at scale

Non-user studies fail for a boring reason: they are slow, so they get cut. Twelve strangers × scheduling × an hour of moderation × transcription is a month of work for a study nobody has budgeted, and it dies in favour of another round of customer interviews that are easier to arrange.

An asynchronous AI-moderated study changes that arithmetic:

• No scheduling. Participants recruited from a panel, an ad, or a community open a link and start immediately, in their own time zone, at 11pm if that is when they are free. Recruiting non-users is hard enough without adding a calendar negotiation.
• Consistent probing across a heterogeneous sample. Non-user samples are messier than customer samples, and the temptation to guide the conversation toward your product is strongest exactly here. Koji's AI moderator asks the same questions with the same neutrality to participant 3 and participant 60 — no pitch creep, no interviewer bias drift late in the day.
• Structure for the comparable parts. Use structured questions for what must be counted — single_choice for the current workaround, multiple_choice for tools evaluated, ranking for what drove the decision, scale for satisfaction with the status quo — and open_ended with probing for everything else. One conversation yields both the distribution and the reasoning.
• Voice or text. Rejecters and busy operators often prefer talking; a five-minute voice conversation while walking gets you a completion where a form gets you an abandon.
• Cost that survives a budget review. A text conversation costs one credit and a voice conversation three, so a 60-person non-user study is a line item rather than a project. The economics are what make this research repeatable instead of a one-off.

Sample size: aim higher than for customer work. Non-user populations are more heterogeneous, so 25 to 40 completed interviews across your target types is a reasonable floor, weighted toward whichever type maps to your current strategic question.

---

Analysing it: compare, do not just summarise

The output of non-user research is only meaningful next to your user data. Three comparisons carry the value:

• Vocabulary. How non-users name the problem versus how you name it. Mismatches here are the cheapest fix in the whole funnel — usually a headline rewrite.
• Job decomposition. Which parts of the job you serve well versus which parts the DIY workaround handles that you ignore entirely. This is where roadmap gaps hide.
• Objection frequency. Rank the reasons non-users give, then check which ones your funnel currently addresses. The most common finding is a top-three objection nobody on the team had heard.

Then be disciplined about what the study cannot do. A non-user sample is self-selected volunteers responding to an incentive; it is directional evidence about a market, not a market-share estimate. Report it as themes with counts and sample sizes attached, and resist the pull to turn "9 of 30 mentioned procurement" into a percentage of the addressable market.

---

Related Resources

• Structured Questions in AI Interviews — the six question types used in this study design
• Survivorship Bias in Customer Research — the bias non-user research exists to correct
• Switch Interviews: The JTBD Method — the adjacent method for people who did change solutions
• Competitive Intelligence Interviews — what your own customers know about rivals
• How to Research Hard-to-Reach Audiences — recruiting tactics for difficult segments
• Churned Customer Interviews — the other half of the people who are not using you
• Feature Adoption Research — non-adopters inside your existing customer base