01 · Start with a person
Define the person and problem before you define a product.
Make the first version of your idea specific enough to investigate: who might experience the problem, in what situation, and what task are they trying to complete? Keep this as a hypothesis. An idea brief, a market description, or a confident founder hunch is not evidence that the need exists.
Customer discovery is about learning how people handle the problem now, what alternatives they use, and whether they might pay for a better option. Ask open questions and listen before you explain or pitch your proposed solution.[1]
Ask about recent behavior, not a hypothetical future
- “Talk me through the last time you handled this.”
- “What part of your current workflow takes the most effort or causes friction?”
- “What happened the last time this became urgent?”
- “What words do you use when you describe this problem to someone else?”
Follow their story. Ask what they tried, what happened, and what they do when their first choice is unavailable. Avoid turning the conversation into a feature pitch or asking whether they “like” an idea they have not used.
Different counts answer different questions
- Before writing code, the report recommends 3 real conversations about the problem, not the solution. [4]
- It says 3–5 real conversations can help answer four early validation questions before committing to product development. [4]
- It describes five customer interviews as more revealing than any tool combination for context, causes, and nuance. This is the report’s comparison, not a guarantee that five interviews settle every question. [8]
- It recommends 20 conversations with target buyers before writing marketing copy. [3]
02 · Investigate current behavior and urgency
Learn how often the problem happens and what makes it matter now.
Ask what the person did most recently, what it cost them in time or effort, and what they did instead. Look for the event that changes the situation from inconvenient to urgent. A problem’s trigger may be a deadline, a failed handoff, an unexpected cost, or another event the customer describes; do not assume one before you listen. The discovery work is to identify that specific trigger and the customer’s own language for it. [3]
Use the four severity labels in order
- AnnoyanceA minor irritant; the person can continue without changing much.
- WorkaroundThey use a manual step or another tool to get around the problem.
- BlockageThe problem slows or prevents important work from moving forward.
- FailureThe work or intended outcome breaks down altogether.
For an existing product’s dependency check, ask: “What would break in your business if our product disappeared tomorrow?” [3] For a pre-product idea, ask about the customer’s current workflow and what happens when their present workaround fails; do not imply that a product already exists.
Keep the planning assumptions visible
The research snapshot assumes a founder has 10–20 hours per week available and uses 15–30 minutes as an assumed length for a “real conversation.” These are planning assumptions, not universal facts or requirements. Fit the work to your own schedule and the conversation. [3]
Reach people personally and be clear about the stage
The report recommends personally contacting early users within 24 hours during the first weeks instead of relying on automated welcome messages. It also recommends finding specific, trusted communities rather than broadcasting to broad generic groups. Those are recommendations in the report, not independently measured guarantees. [2]
Say what you are exploring and why, in direct language. Ask for feedback before launch or just after it; a thoughtful voice note can reveal more than a dashboard of uncontextualized signups. The report recommends listening before selling.[2]
“If you've ever struggled to find the right person to build with — would you try it and give me brutal feedback?”
A context-specific collaboration invitation in the research report
03 · Research the market
Look at the need, the alternatives, and who experiences it.
Keep desk research focused on four areas. Use it to shape better questions for people, not to replace conversations with them. [8]
- Pain-point discoveryFind how people describe the difficulty, how they work around it, and what makes it urgent. Search specific community discussions and compare what you read with interview accounts.
- Demand validationLook for signs that people are actively searching, switching, booking, pre-ordering, or spending to solve the problem. Treat each measure as one signal, with its limits.
- Competitor analysisReview competitor websites to understand their promises and offers. Read 1–3-star G2 and Capterra reviews for repeated unmet needs and missing workflows; those comments are clues, not representative market research.
- Customer segmentationSeparate people by their situation, urgency, existing workaround, or ability to reach them. A narrower segment makes it easier to learn whether the same problem repeats.
The report lists Google Trends, competitor websites, Reddit and other community forums, and free-tier AI research tools as low-cost starting points. It recommends spending close to zero on tools until you have paying customers. [7] [8]
Its separate budgeting guidance gives $0–$5,000 for one early-stage research project. If initial screening looks promising, it cites $5,000–$15,000 for a focused survey or $59 for an advanced AI report, while saying five-figure studies suit later, post-product-market-fit work. These are attributed ranges and examples from the report, not a recommended spend for every founder. [7]
The report also argues that a poorly scoped $20,000 survey may tell less than a $59 AI report using live competitive data. That is its comparison, not an independently verified result; scope and source quality still matter. [7]
04 · Run a small experiment
Test one important assumption with a small, observable action.
An MVP is the simplest version of a product that delivers its core value. It can help test an idea in the real world without building a large product first.[1] A landing page or email signup can measure click-through, conversion, or pre-orders, but those measures have different strengths and should not be treated as proof by themselves. [1]
Combine qualitative interviews with quantitative evidence. Signups alone are weak. Stronger signals include a meaningful reply, someone taking time for a 30-minute problem discussion, a referral, a booking, a pre-order, or willingness to pay. These signals show different kinds of effort; none alone proves that a sustainable business exists. [1] [4]
Before an MVP, decide what result would justify the next step. One example gate in the report is a small number of target users completing the core action and at least one person willing to pay the stated price. Write down the audience, action, price, time window, and threshold before you run the test so you can compare the result with the original decision. [4]
05 · Set the evidence gate
Decide what would change your mind before coding or spending more.
State the next decision in plain language: what will you build, buy, or stop doing if the evidence is positive, mixed, or negative? A gate should name the people you need to reach, the core action they need to complete, and what commitment would count as meaningful. The report’s pre-MVP example combines target users completing a core action with at least one person willing to pay the stated price.[4]
Ask whether there is a believable path to the first 10 customers. If you cannot name where those people might come from and how you will reach them, test that distribution assumption separately from the problem itself. [4]
Do not call an idea validated just because people signed up, an AI tool generated a brief, or one person praised the concept. A generated brief can suggest hypotheses and questions; the answers must come from customers, observable behavior, and experiments you actually ran.
After a product exists · Revalidate separately
Early exploration and product-market-fit checks answer different questions.
The following cadence is for current customers and an existing product. It is not a substitute for pre-product discovery or a target for an unlaunched idea.
Monthly alignment questions for current customers
- “What changed in how you do this work in the last 90 days?”
- “What problem feels more dangerous now than it did last quarter?”
- “What would break if our product disappeared tomorrow?”
The report recommends these alignment questions and says to avoid feature-focused prompts in the monthly conversation. [3]
Quarterly Dependency Audit
The report describes a quarterly audit of 5–10 active customers. It uses only the Dependency Test question: “What would break in your business if our product disappeared tomorrow?” Its target is at least 60% of those customers in Blockage or Failure. Keep that audit question and target distinct from monthly alignment prompts and early idea exploration; they are source-backed recommendations, not a universal law. [3]
Use tools as aids, not verdicts
Named tools can help organize work, but their reported claims need context.
ValidatorAI
The research report describes ValidatorAI as using behavioral data from 300,000 founders and says its method examines what moves founders from an idea to action before companies are built—framed as whether founders do what they say they will. In a separate statement, it describes real-time data from 293,460 founders’ ideas.[5] The snapshot does not explain how the two reported counts relate, so they are kept separate here. These are the report’s descriptions of a tool-provider method, not independently verified performance or proof that an individual idea is validated.
CoFounder.AI
The report presents CoFounder.AI’s working framework as Approve, Delegate, or Direct, with the AI proposing work for the founder to approve much of the time. It says the tool keeps a running context record covering stage, industry, goals, decisions, deliverables, and connected tools. The listed integrations are QuickBooks, Stripe, HubSpot, Slack, and Gmail. [6] These are attributed descriptions in the research snapshot, not an endorsement or independently checked integration test.
Prototype costs and capacity estimates
The report describes Figma’s free tier as suitable for MVP development. It estimates no-code subscriptions at £200–500 monthly versus £20,000–50,000 for custom-coded MVPs, and says most no-code platforms can handle 1,000–10,000 users before migration, when speed issues typically emerge. [9] These are cited estimates, not guarantees, current quotes, or a prescribed budget; the right choice depends on what you need to learn and what the tool can actually do.
Citation notes
Citation markers and source availability
This guide preserves the supplied markers [1]–[9]. The dispatch snapshot did not include the bibliography or source URLs, so those entries cannot be reproduced or linked without the full report. No missing titles, publishers, dates, or links have been inferred.
- [1] Citation marker present in the research snapshot; bibliographic entry and URL were not supplied.
- [2] Citation marker present in the research snapshot; bibliographic entry and URL were not supplied.
- [3] Citation marker present in the research snapshot; bibliographic entry and URL were not supplied.
- [4] Citation marker present in the research snapshot; bibliographic entry and URL were not supplied.
- [5] Citation marker present in the research snapshot; bibliographic entry and URL were not supplied.
- [6] Citation marker present in the research snapshot; bibliographic entry and URL were not supplied.
- [7] Citation marker present in the research snapshot; bibliographic entry and URL were not supplied.
- [8] Citation marker present in the research snapshot; bibliographic entry and URL were not supplied.
- [9] Citation marker present in the research snapshot; bibliographic entry and URL were not supplied.