Skip to main content
← User Interview
Spoke · Updated May 2026 · 8 min read

User interview best practices

12 rules distilled from hundreds of sessions. Recruitment, moderation, recording, synthesis, and the bias traps to avoid.

Why 12 rules

Most user-research mistakes aren't exotic — they're the same dozen mistakes everyone makes the first time. Recruit the wrong sample. Lead with future-tense questions. Fill silences. Skip the synthesis. Believe stated behavior over observed.

The 12 rules below cluster into four buckets — recruitment, moderation, recording, synthesis — plus a final rule on bias. Read them once, mark the three you currently violate, and fix those before the next round.

1

Recruitment

Screen aggressively, then trust the screen

A 5-question screener that auto-rejects people outside your ICP saves more time than any other moderation hack. Once someone passes the screen, treat them as the right participant — don't spend the first 10 minutes of the interview re-qualifying them.

2

Recruitment

Pay fairly and on time

$25–$50 for 30 min from consumers, $75–$150 for 60 min from professionals, $200+ for execs and specialists. Send the incentive within 24 hours of the call. Underpaying selects for participants who have nothing better to do — exactly the wrong sample.

3

Recruitment

Don't interview your friends

Friends, employees, and investors give you flattering signal. They want you to succeed, so they say what helps you feel good. The first 10 interviews of any research project should be with people who don't know you exist.

4

Moderation

Past tense, not future tense

"Tell me about the last time you tried to do that" beats "Would you use a feature that…" every time. Past behavior is data. Future intent is fiction. The Mom Test by Rob Fitzpatrick is the canonical reference here.

5

Moderation

Show, don't pitch

When showing a prototype, never describe what it does first. Hand it over and ask "what do you think this is?" Their interpretation is more valuable than your explanation. If they don't understand the prototype, that's the finding — not a problem to talk over.

6

Moderation

Wait four seconds after every answer

The most valuable sentence in the interview is usually the second one — the one they add when nobody fills the silence. Most interviewers panic at three seconds and ask the next question. Don't. Count to four. Watch what happens.

7

Moderation

Follow tangents for at least 60 seconds

When the participant goes off-topic, that's often where their actual concern is. Ride the tangent for at least a minute before redirecting. The tangent is honest; the prepared topic is dutiful.

8

Moderation

Distinguish observed from stated

When a participant says "I always check that screen first" but you watched them skip it three times in the screen-share, write down both. Stated behavior is what they think they do. Observed behavior is what they actually do. They're often different — and the gap is the finding.

9

Recording

Always record — with explicit consent

Recording lets you re-watch body language, pull verbatim quotes, and let teammates synthesize without sitting through every session. Disclose at minute 0, store securely, never share externally without re-consent. Tools: Zoom built-in, Otter, Fireflies, Grain.

10

Synthesis

Synthesize within 24 hours

Memory of context decays fast. Within 24 hours, write down the top 3 quotes, the surprising moments, and one hypothesis the interview confirmed or falsified. After 48 hours you'll remember the gist but not the texture — and texture is where insight lives.

11

Synthesis

Pull verbatim quotes, not paraphrases

Stakeholders trust direct quotes more than your summary. Time-stamp each quote so you can re-watch the moment for context. A research artifact with five verbatim quotes outperforms a 10-page report.

12

Bias

Watch for the four classic biases

Confirmation bias (hearing what you expected). Recency bias (last interview dominates synthesis). Selection bias (only enthusiastic people show up). Politeness bias (participants soften criticism). Naming the bias when you spot it cuts its power roughly in half.

Rehearse the 12 rules live

GoNoGo runs interview rehearsals against AI personas grounded in real conversations. You catch the leading questions, the rushed silences, and the politeness-bias signals you would have missed in a real session — without burning a participant slot.

Rehearse my interview →

Free · No credit card · up to 25 reports

Frequently asked questions

Which of these matters most if I can only do three?+
Past-tense framing (Rule 4), silence after answers (Rule 6), and same-day synthesis (Rule 11). Those three alone separate amateur interviews from professional ones. Everything else is incremental.
How do these change for B2B vs B2C interviews?+
Recruitment and incentives change (B2B prefers professional reciprocity over cash), but the moderation rules don't. The same human dynamics — politeness bias, recency bias, hypothetical-vs-actual gap — show up in both. The 12 rules apply to both audiences.
Can I skip recording if it makes participants uncomfortable?+
You can — but expect to lose ~40% of the value. Without recording, you're writing notes during the conversation, which means you're not fully listening. If a participant declines recording, take written notes and immediately after the call do a 10-minute brain-dump while it's fresh.

More on user interviews