User Research on a Startup Budget: What's Worth Doing

There's a misconception that user research requires a dedicated researcher, a usability lab, and a substantial budget. It doesn't. Most of the research methods that are actually useful at early stage are accessible to any founder with a calendar and a willingness to watch people struggle with their product.

The things you're trying to learn at early stage are also different from what you're trying to learn later. At early stage, you're validating whether the problem exists, whether your solution makes sense, and where your onboarding breaks. These don't require statistical significance — they require directional signal from a handful of honest observations.

What You're Actually Looking For at Early Stage

At the beginning, you're not optimizing. You're discovering. The questions are broad:

• Are users able to understand what the product does when they first encounter it?
• Can they complete the core task without getting stuck?
• Do they understand the value after using it?
• Would they use it again?

For these questions, five to eight users per segment is often enough to see the major patterns. The research method isn't about sample size — it's about rigor in the observation and honest interpretation of what you see.

At optimization stage (post-product-market-fit), the questions are more specific: should the button be labeled X or Y, which version of the onboarding flow has better activation, is this error message causing drop-off. These require quantitative methods — A/B tests, large enough samples for statistical significance. The expensive, rigorous version of research belongs here.

Moderated Usability Testing Without a Lab

A moderated usability session is you watching someone use your product while they narrate their thinking. That's it. No lab required.

Setup: Video call with screen sharing. Zoom, Google Meet, Teams — any tool works. Record with permission.

Recruitment: Email five to ten people who match your target user profile. Offer a $50-75 gift card for 45-60 minutes of their time. The incentive is necessary — it filters out people who are only participating because they want to be helpful to you personally.

The session structure:

1. Brief intro (5 min): Explain what you're doing (testing the product, not testing them), ask them to think aloud as they use it.

2. Context questions (10 min): How do they currently handle the problem your product solves? What tools do they use? How much time do they spend on it? This gives you baseline understanding before they touch the product.

3. Task scenarios (25-30 min): Give them 3-5 realistic tasks to complete with the product. "Imagine you want to do X — how would you approach that with this tool?" Then stop talking. Watch. Don't help. The frustration and confusion they express is your most valuable data.

4. Debrief questions (10-15 min): What was confusing? What would they need to see to use this regularly? What would they compare it to? How does it compare?

What you're looking for: Where do they click that surprises you? What do they try to do that the product doesn't support? What language do they use to describe the problem that differs from yours? Where do they express frustration?

Write down direct quotes and specific behaviors during the session, not interpretations. "She clicked on the logo three times expecting to go to the dashboard" is data. "She seemed confused" is interpretation.

Guerrilla Research

Guerrilla research means recruiting participants in the moment from contextually relevant places — coffee shops, coworking spaces, events — rather than scheduling them in advance.

The validity of guerrilla research is limited. Participants haven't been pre-qualified, they're in a context different from actual product use, and they're doing you a favor rather than participating in a structured study. This means you need to be cautious about how you interpret what you hear.

When guerrilla research is useful:

• Getting first impressions of a landing page or product homepage (5-second tests)
• Testing whether an explanation or value proposition is understandable
• Quick sanity checks on a design decision

When it isn't:

• When you need participants who actually have the problem you're solving
• When the task requires more than 10-15 minutes of engagement
• When you're making major product decisions

Think of guerrilla research as supplementary, not primary.

Remote Research Tools

Remote research is the norm now and the tooling is good:

Recruiting: User Interviews, Respondent.io, and Prolific are the main options for finding research participants quickly. You define the screening criteria, they find the people. Costs roughly $50-150 per participant depending on specificity.

Moderated sessions: Zoom or Lookback.io. Lookback has better session replay features. Zoom is free and everyone already has it.

Unmoderated testing: Maze, UserTesting, or Useberry. You set up a task flow; users complete it on their own and you get recordings plus behavioral data. Good for specific task completion questions at moderate scale.

Card sorting and tree testing: Optimal Workshop is the standard tool here, good for navigation and information architecture decisions.

Survey research: Typeform or Google Forms. The barrier is low; the quality of data is only as good as the questions, which is the hard part.

How Many Users You Actually Need

This question gets overthought. A practical guide:

| Research Type | Sample Size | Notes | |---|---|---| | Moderated usability testing | 5-8 per segment | For identifying major usability issues | | Early problem discovery interviews | 10-15 | Until you stop hearing new themes | | Unmoderated task testing | 20-30 | For directional signal on specific tasks | | Survey research | 50+ | For basic patterns; 200+ for segmentation | | A/B testing | Hundreds to thousands | Depends on effect size and baseline rate |

The "five users" rule from Jakob Nielsen applies specifically to finding usability issues in moderated sessions — it's not a general-purpose rule for all research types.

The honest answer: do more research than you're currently doing, because most early-stage teams do almost none. Five sessions where you genuinely watch someone struggle with your product will change how you think about it more than 500 survey responses will.

The Bias Problem

The biggest research risk for founders is confirmation bias — structuring research to validate what you already believe. It manifests in question design ("didn't you find the onboarding clear?"), session facilitation (jumping in when users struggle instead of watching), and interpretation (remembering the positive responses more vividly than the negative ones).

Some mitigation: have someone who isn't the founder conduct at least some of the sessions. The founder's emotional investment in the product is visible to participants and affects what they say. An outside researcher, even a contractor doing 10 sessions, often surfaces things a founder-led process misses.

Also: watch the recordings yourself afterward, alone, without someone next to you who is also invested in the outcome. The things you see when you're not narrating in real time are often the most important things. Founders who regularly share their research findings with outside advisors — whether through a peer network or a platform like Founderboard — get better at distinguishing genuine signal from the patterns their own bias is constructing.