User Research for Startups: How to Talk to Customers Before You Build the Wrong Thing

Most startups skip customer research because they think they do not have time.

Most startups that fail did so because they built something people did not want.

These two facts are related.

The startups that move fastest are not the ones who skip research — they are the ones who have figured out how to run it quickly. Twenty customer conversations before you build a feature can save six months of engineering time. A 15-minute AI interview with five churned users can explain why your retention curve has a cliff at day 14. An hour of structured discovery conversations can tell you whether a market exists before you write a single line of code.

This guide is about how to do research at startup speed: fast, focused, and without a dedicated research team.

Why Research Feels Hard at Startups (And Why It Does Not Have to Be)

"We do not have a researcher." Most research that matters to early-stage startups does not require a trained researcher. It requires good questions, genuine curiosity, and the discipline to listen more than you talk. The methodology frameworks built into Koji handle the research best practices — you bring the curiosity.

"Our users are hard to reach." Every startup has some path to people experiencing their target problem — Reddit communities, LinkedIn, email lists, beta user groups. The bottleneck is usually scheduling 30-minute calls with 20 different people across different time zones, not finding them. Async AI interviews solve the scheduling problem entirely.

"Surveys are faster." Surveys are faster to deploy, not faster to produce insights from. A 20-question survey with 100 responses gives you a spreadsheet. A 15-minute AI interview with 20 participants gives you the stories, reasoning, and specific context behind the numbers — the things that tell you what to build.

"We will do research when we are bigger." The highest-value moment for research is before you have built anything, or right after you have shipped something. Waiting until you are bigger means you have already paid the cost of getting it wrong. The best time to talk to customers is always now.

With AI-moderated research platforms like Koji, the "we do not have time" objection largely disappears. Setting up a study takes 45 minutes. Collecting 20 responses takes 24–48 hours. Reading the automated report takes 30–60 minutes. Total investment: 3–4 hours for insights that would otherwise require 10–20 hours of calendar blocking for live interviews.

The Right Kind of Research for Each Startup Stage

Customer research is not one thing — it changes as your startup evolves.

Pre-Idea: Problem Discovery

If you are still figuring out what to build, the goal is to understand a specific population's pains, behaviors, and workarounds — without pitching anything.

The best framework here is the Mom Test approach, named after the book by Rob Fitzpatrick. The core principle: ask about their life, not your idea. "Tell me about the last time you had to handle this manually" reveals more than "Would you use a product that automated this?" — because people tell you the truth about their past, but lie (inadvertently) about their hypothetical future behavior.

In Koji, set your methodology to Customer Discovery with Mom Test principles. Structure your key questions around:

• "Walk me through how you currently handle this area"
• "What is the most frustrating part of that process?"
• "What have you already tried to solve it?"
• "What would it cost you — in time or money — if this stayed exactly as painful as it is now?"

You do not need 100 interviews at this stage. Six to eight conversations with the right people will tell you whether there is a real, painful, common problem worth solving. Platforms like Koji make it practical to run these eight conversations in 48 hours rather than two weeks.

Pre-Launch: Customer Discovery and Validation

You have an idea or an early prototype. The goal now is to validate that the right people have this problem, that they would prioritize solving it, and that your solution is directionally correct.

Jobs to Be Done (JTBD) is an ideal framework here. It focuses on understanding the progress participants are trying to make — the "job" they would hire a product to do. JTBD interviews reveal:

• What triggered them to start looking for a solution
• What alternatives they considered (your real competition is often not other startups — it is spreadsheets, workarounds, or doing nothing)
• What anxieties almost stopped them from switching
• What they had to give up to make a change

In Koji, the Jobs to Be Done methodology framework loads these principles into your AI interviewer automatically. Add structured questions to capture:

• Scale ratings (1–10) for how painful the current approach is
• Yes/no checks ("Before you started looking for a solution, were you managing this in a spreadsheet?")
• Open-ended questions for discovery moments ("Take me back to when you first realized you needed to do something about this")

For a complete guide to all six Koji question types — including how to combine scale, yes/no, and open-ended questions for maximum research value — see the Structured Questions Guide.

Post-Launch: Onboarding and Activation Research

You have users. Some of them are activating and becoming engaged; others are signing up and disappearing. The research question is: what is different about the ones who succeed?

Interview both groups — but separately. Do not pool insights from activated users and churned users in the same report, because the patterns you are looking for are the differences between them.

For activation research, focus on:

• The first impression and setup experience
• Where confusion or friction appeared
• What the activation moment was — and whether they actually had one

For early churn, focus on:

• What they hoped the product would do
• What happened when reality did not match that expectation
• What they did instead after leaving

In Koji, you can segment these studies by importing your activated and non-activated cohorts from your CRM or a CSV export. Personalized interview links can pre-fill context about each user's specific behavior to make the interview feel relevant rather than generic.

Growth Stage: Feature Validation and Power User Research

At this stage you have more data — and more assumptions. The research job shifts from "do we have product-market fit?" to "how do we deepen it?"

Power user interviews reveal what your best customers are actually using the product for — which is often different from what you intended. These conversations surface use cases, integrations, and workflows you did not design for, and reveal the features power users wish existed.

Feature validation research asks: before we build this, do users understand what it would do? Would they actually use it? Have they tried to do this another way already?

Mix question types for feature research:

• Open-ended: "Walk me through how you currently handle this"
• Scale: "On a scale of 1–10, how painful is that process?"
• Yes/no: "Have you ever tried to do this within our product?"
• Multiple choice: "If we built this feature, which of these would you most likely use it for?"
• Open-ended: "What would make this a must-have vs. a nice-to-have for you?"

How to Run 20 Interviews in a Week Without a Research Team

Here is a concrete timeline for a startup running customer research with Koji:

Day 1 (45–60 min): Set up your Koji study. Define your research question, complete the brief, and generate your interview link. The brief covers: problem context, target participant, methodology, and key questions.

Day 1–2 (30 min): Distribute the interview link. Email your user list, post in your beta community, DM participants directly, or add an in-app prompt for users who trigger a specific event.

Days 2–5: Koji's AI conducts interviews as participants click the link — day or night, at their own pace. You receive a notification as each response comes in and can see the transcript immediately.

Day 5–6 (60 min): Generate your research report in Koji. Read the synthesized themes, review the quantitative charts from your structured questions, scan representative quotes from open-ended responses.

Day 6 (30 min): Share the report with your co-founders, designer, or first engineer. Make a decision.

Total founder time: approximately 3 hours for 20 interviews worth of insight. That is less than a single afternoon, and it is the kind of research that prevents six-month mistakes.

Koji's Credit Model: Research at Startup Budgets

New Koji accounts receive 10 free starter credits on signup — enough to run several text interviews and experience the full product before committing to a plan.

After your free credits:

• Insights plan: €29/month for 29 credits — suitable for 15–29 text interviews per month
• Interviews plan: €79/month for 79 credits with unlimited studies and free report refreshes — suitable for startups running continuous research across multiple studies
• Enterprise: Custom pricing for teams running research at scale

For most early-stage startups, the Insights plan covers 2–3 focused studies per month. That is more research than most seed-stage companies currently run — and the kind of regular customer contact that compounds into a genuine understanding advantage over teams that only talk to customers quarterly.

Common Startup Research Mistakes to Avoid

Asking hypothetical questions. "Would you use a feature that did X?" People are poor predictors of their own future behavior. Ask about what they have actually done. "Have you ever tried to do X?" is a stronger question than "Would you pay for something that does X?"

Pitching instead of listening. The purpose of customer interviews is to understand the customer's world, not to get them excited about your solution. If you find yourself talking for more than 30% of the interview, you are pitching, not researching. Koji's AI-moderated format enforces this discipline by design.

Only talking to enthusiasts. Your most enthusiastic users are not your typical users. Talk to people who tried your product and stopped. Talk to people who fit your ICP but are still doing the thing manually. Talk to people who looked at your product and decided not to sign up. These are often your most valuable conversations.

Treating research as a one-time event. The startups with the deepest customer understanding treat research as a rhythm, not a project. Weekly or biweekly conversations — even just 5–8 per sprint cycle — create a compound intelligence advantage that is very difficult for competitors to replicate.

Frequently Asked Questions

Q: How do I find participants if I do not have many users yet? Before you have customers, find problem-space participants in relevant communities: subreddits, LinkedIn groups, Slack communities, and industry forums. Reach out directly with a clear explanation of who you are looking for and why. Offer a small incentive (€20–30 gift card) for 15 minutes of their time. Koji's async format makes recruitment easier — participants do not need to schedule a call, just click a link when they have time.

Q: How many interviews is enough to validate an idea? Five to eight interviews with the right participants typically surface the most important themes. You will notice you are hearing the same things repeatedly — that is the signal you have reached saturation on the key patterns. For higher-stakes decisions (pricing strategy, pivots, major feature bets), aim for 15–20.

Q: Should founders be doing research or should we hire a researcher? At the early stage, founders should do research — or at minimum participate in it. Customer empathy is a competitive advantage, and it cannot be fully delegated. AI-moderated research makes this practical: you are not spending 20 hours conducting live interviews, you are spending 1 hour setting up a study and 1 hour reading the AI-generated report.

Q: How do we prevent research from slowing us down? Scoped research is fast research. A study with a clear, narrow question ("Why do users churn at the end of the free trial?") can be set up in 45 minutes, collect responses overnight, and generate a report the next morning. That is not a slowdown — it is a shortcut that prevents you from building the wrong thing at full speed.

Q: Can Koji handle interviews in multiple languages? Yes. Koji supports interviews in multiple languages. If you are building for a non-English-speaking market, you can configure your study for the appropriate language and your participants will be interviewed in their native language — with analysis and reports generated in English.

Q: What is the difference between Koji and running a survey tool like Typeform? Surveys give participants a fixed set of questions with no follow-up. Koji's AI asks your key questions and then probes each participant's specific answers — following the reasoning, asking for examples, and surfacing the context that explains the patterns. The difference in insight quality is significant, especially for discovery and problem validation research.

Related Resources

• Structured Questions Guide: How to Use All 6 Koji Question Types
• The Mom Test at Scale: AI Interviews for Customer Discovery
• Jobs to Be Done Framework: The Complete Guide
• Lean User Research: How to Run Meaningful Research with No Time or Budget
• How Many User Interviews Do You Need?
• Customer Discovery Interviews at Scale