Sample content: This article exists to exercise the publishing system. It is not presented as Patrick’s writing.
What to Measure Before Product-Market Fit
A sample founder’s guide to evidence that reduces product uncertainty before dashboards and growth metrics become persuasive distractions.
Before a product has earned repeat demand, polished dashboards can create confidence faster than understanding. Traffic rises, accounts accumulate, and feature requests arrive. None of those signals alone proves that the product solves a problem strongly enough to support a business.
Early measurement should make uncertainty smaller, not make the company look larger.
Prefer behavior that costs the user something
Copy link to section “Prefer behavior that costs the user something”Signals become more meaningful when they require effort, trust, money, reputation, or a change in habit. A visitor clicking a launch link is weak evidence. A user importing real data, inviting a colleague, returning without a reminder, or working around a missing feature to keep using the product is stronger.
This does not mean every early product needs payment immediately. It means distinguishing curiosity from commitment.
| Signal | What it may reveal |
|---|---|
| Completes setup with real data | The problem is worth initial effort |
| Returns to repeat the core workflow | Value persists beyond novelty |
| Invites someone needed for the job | The product fits a real process |
| Asks what happens if the product disappears | The workflow is becoming important |
| Pays or requests procurement steps | Value may support an economic exchange |
Each signal still needs context. A user may invite colleagues because setup is confusing. Measurement should lead to a conversation, not replace one.
Follow the core workflow, not every event
Copy link to section “Follow the core workflow, not every event”Instrumentation can expand faster than the product. Start with the smallest path that represents value: arrive, configure, complete the important job, and return when the need occurs again.
Look for where serious users stop. A low completion rate may indicate weak demand, but it may also reveal a broken step, a missing trust signal, or the wrong audience. Pair behavioral evidence with direct observation and interviews.
Avoid optimizing peripheral activity while the core remains uncertain. More notifications may improve return visits without making the product more valuable. A longer session may indicate engagement or confusion. Metrics need a model of the job the user is trying to do.
Measure learning velocity inside the team
Copy link to section “Measure learning velocity inside the team”The product is not the only system under test. Ask how quickly the team can turn an assumption into evidence and update its decisions.
Useful internal signals include:
- time from an important question to a real user observation;
- percentage of experiments with a decision scheduled in advance;
- recurring objections the team still cannot explain;
- features maintained despite no evidence of use or strategic value;
- customer segments that behave differently enough to require separate models.
These are not performance targets. They reveal whether the company is building a learning system or merely a shipping system.
Beware of averages too early
Copy link to section “Beware of averages too early”An early average can combine several unrelated realities. Ten enthusiastic technical users and ten confused occasional users do not form one representative customer. Segment by the problem, behavior, and context before drawing broad conclusions.
Outliers can be especially useful. The person who receives enormous value may reveal the narrow market worth serving first. The person who repeatedly fails may expose a mismatch the roadmap cannot fix.
Know what evidence would change the plan
Copy link to section “Know what evidence would change the plan”Write down the current product belief and the next evidence that would strengthen or weaken it. If no result can change the plan, the team is not measuring; it is collecting decoration.
Product-market fit is not a dashboard threshold that can be declared from one chart. Before it becomes obvious, founders can still behave rigorously: value costly user behavior, study the core workflow, preserve direct contact, and let evidence change the product.
