Key takeaways
- Match method to the question. A survey cannot fix a navigation argument you never watched anyone use.
- Five good conversations beat a hundred survey answers nobody reads.
- Store evidence where the next designer can find it — date, method, clip, decision.
The roadmap argued for a feature nobody had asked for. The survey said people "somewhat agreed" with a statement nobody understood.
That is what guessing looks like when it wears a Google Form. User research for startups does not require a lab, a recruiter, or a research ops hire. It requires the right method for the question, a handful of honest sessions, and a place to store what you learned so the next argument starts with evidence — not vibes.
Pick the method after you write the question
Run the question-to-method selector before you recruit anyone.
| If you need to know… | Use | Skip |
|---|---|---|
| Why people abandon a step | Five moderated task sessions on the live route | A satisfaction survey |
| Which label people understand | Five-second test or unmoderated first-click | Longitudinal diary study |
| Whether the offer matches the buyer | Structured interviews with recent wins and losses | Analytics alone |
| If the page is findable | Tree test or open card sort (even on paper) | Heatmaps without tasks |
| What to fix first on a leaky site | Representative-route audit plus three sessions | A/B tests before basics work |
Write the question in plain language: We need to decide ___ by ___. If the method cannot answer that sentence, pick a different method.
How to brief a remote creative team is the handoff mirror: research output is input to a brief, not a slide deck that dies in Drive.
Five conversations beat a survey you will not read
Surveys scale opinions about statements you wrote. They rarely explain behavior on a confusing form.
For most small teams, five moderated sessions on the critical path answer more than a hundred checkbox rows:
1. Recruit people who match the buyer — not only friends who will praise you. 2. Give a task, not a tour: "Try to request a quote as you would on a Tuesday morning." 3. Think aloud. Silence is data; prompting is allowed when they stall. 4. Record with consent. Timestamp the moment they sigh, scroll back, or open the menu twice. 5. Debrief in thirty minutes: three findings, three fixes, one owner.
Stop at five unless new sessions repeat the same surprise. Saturation beats sample size theater.
The Nielsen Norman Group's guidance on why you only need to test with five users is still the honest math for qualitative discovery — not for precision metrics, but for finding the big leaks.
Recruitment constraints are part of the design
Small teams fail research by recruiting nobody like the buyer.
Practical sources:
- Recent customers — win or loss — with a short incentive
- Inbound leads who opted in — one question, not a fishing expedition
- Partner networks and industry forums — disclose who you are
- Walk-ins or booth conversations — capture notes the same day
Write a screener with three must-match traits and three disqualifiers. "Uses the product weekly" matters more than "likes technology."
Remote is fine. A phone call and a screen share beats waiting for a lab. Proof on a website without fake case studies applies here too: store real quotes with permission, not invented personas.
Build an evidence repository — not a pile of clips
Every session gets one row:
| Field | Example |
|---|---|
| Date | 2025-02-10 |
| Method | Moderated task test |
| Participants | 5 HVAC owners, 2–20 employees |
| Question | Can they find pricing and request a quote in under three minutes? |
| Finding | 4/5 opened menu twice; pricing label said "Solutions" |
| Decision | Rename nav; add pricing to home |
| Artifact link | Clip + transcript excerpt |
Store clips, photos of paper sorts, and screenshots in one folder with the same naming convention. Future you — and the next agency — should open row twelve and know what changed on the site and why.
Without the repository, research becomes folklore. Folklore loses to whoever speaks loudest in the next meeting.
Connect findings to buildable fixes
A finding without a fix is a note. Translate:
- Observation — what happened
- Issue — why it matters to revenue or trust
- Recommendation — specific copy, nav, or layout change
- Severity — blocks primary action vs polish
- Owner and date — who ships the fix and when you will re-test
Design thinking workshops help when the question is strategic — positioning, service mix, which offer to lead with. Task tests help when the question is mechanical — can they finish the form. Match the workshop to the decision.
Schedule a re-test after the fix ships — even three sessions. Research is a loop, not a report that ends in applause. If the same confusion reappears, the fix was cosmetic.
What you can run yourself — and when to hire the route
You can run five sessions, maintain the repository, and refuse roadmap items with no evidence row. You can rename nav labels without a rebrand.
What stays unfinished is the fix on the site: copy, structure, forms, and measurement — a conversion repair when the leak is proven, or a full website when the evidence says the whole path is wrong. Web design needs the repository in the brief so work does not restart from zero.
CLICK.BLUE has turned research into shipped pages since 2014 — more than fifty businesses, more than a hundred websites. If the next debate is about what users do — not what the room prefers — get started with the question, the method, and the last five findings in hand.
