Context unknown
What is happening around the behavior?
Possible sources: Interviews, observation, support conversations, qualitative research
Product Management Foundations · Learning chapter
Understanding Users and Markets
Learn what evidence a product decision actually needs.
Understand users, behavior, segments, alternatives and market context, then choose research methods based on the uncertainty you need to reduce—not because “user research” is on a checklist.
30 current learning unitsDecision-first evidence modelHypothetical worked exampleReal lesson preview
The core idea
Research is useful when it reduces uncertainty behind a real decision
Start with the decision and its unknowns. Then choose evidence that fits the question: direct user stories for context and motivation, behavioral data for observed patterns, surveys for carefully designed structured responses, and market or competitor evidence for category context. Every source has limits. The goal is not certainty; it is a better-supported next decision.
The chapter question
What do I need to understand about users and the market before I can make this product decision with less avoidable uncertainty?
In the previous chapter, you learned to ask what decision a PM activity is supposed to improve. Now ask the next question: what evidence would make that decision better?
10× learning asset
Evidence-to-Decision Research Map
Do not begin with a favorite method. Walk from the decision to the unknown, select evidence that can address it, make its limits visible, investigate contradictions, and state the decision implication explicitly.
- 1
Decision
What product choice is actually pending?
- 2
Unknown
What must become clearer before that choice improves?
- 3
Evidence source
Which source is capable of answering that unknown?
- 4
Can tell you
What claim can this evidence reasonably support?
- 5
Cannot tell you
What will still be uncertain after you collect it?
- 6
Contradiction
What evidence would challenge the current story?
- 7
Decision implication
What changes in the decision if the evidence holds?
- 8
Next uncertainty
What is now the most important unresolved question?
Start with: what decision is actually pending?
“Learn everything about our users” is not a decision. A stronger brief is: “Understand why new team admins stop before inviting collaborators, so we can decide whether the next work belongs in onboarding, product value, or acquisition.”
Name the uncertainty
Different unknowns need different evidence
Prevalence unknown
How common is the pattern?
Possible sources: Behavioral analytics, structured quantitative data, appropriately designed surveys
Segment unknown
Who experiences the problem differently?
Possible sources: User/account attributes, behavior grouping, interviews, context data
Alternative unknown
How is the job solved today?
Possible sources: Interviews, workflow observation, competitors, substitutes, market scans
Opportunity-size unknown
How much user or economic opportunity might exist?
Possible sources: Bottom-up models, current customer data, credible external sources, industry data
Causal unknown
Why did behavior change?
Possible sources: A design that can distinguish competing explanations; observation alone often cannot prove causality
Evidence fit, not hierarchy theater
Ask what each source can—and cannot—establish
There is no universal ladder where behavior data always beats interviews or quantitative evidence always beats qualitative evidence. Evidence quality depends on the question.
Interviews
Can help tell you
Reveal context, sequence, motivations, workarounds, language and decision process.
Cannot establish alone
Establish population prevalence, prove causal effects, or guarantee future behavior on their own.
Product analytics
Can help tell you
Show observed product behavior, frequency, funnels, cohorts and segment differences.
Cannot establish alone
Explain motivation, invisible off-product workarounds, or whether correlation is causal on its own.
Surveys
Can help tell you
Collect structured self-report, attitudes, experience signals and segmentation variables.
Cannot establish alone
Become truth because the respondent count is large. Sample, wording, response bias and design still matter.
Competitor and substitute research
Can help tell you
Show alternatives, category conventions, positioning, visible capabilities and market-structure signals.
Cannot establish alone
Prove your users want the same feature, that the competitor feature works, or that your users value its approach.
Market sizing
Can help tell you
Make opportunity assumptions explicit and compare rough orders of magnitude.
Cannot establish alone
Prove product-market fit, obtainable share or future adoption. A model is not a forecast disguised as certainty.
Interpretation guardrails
Context matters more than labels
Research becomes weaker when convenient labels replace the evidence underneath them. Use segments, personas, competitors, trends, and B2B/B2C context only when they explain a difference that can change the product decision.
Segment by meaningful difference
Useful segments explain differences in problem, behavior, context, constraint, desired outcome, or ability to adopt. Demographics or job titles are not automatically decision-relevant segments.
Treat personas as representations, not proof
A persona or archetype can summarize research-backed patterns. It is not evidence by itself, proof that a market exists, or a permanent description of every person in a segment.
Read competitors and trends as context
Direct competitors, indirect alternatives, manual workarounds, and non-consumption can all matter. A trend matters when it changes behavior, constraints, economics, capabilities, regulation, distribution, or competitive structure—not because it is fashionable.
Use B2B / B2C context without stereotypes
Inspect buyer versus user roles, procurement, multi-person decisions, account versus individual behavior, usage frequency, scale, privacy context, and available behavioral data. Do not assume B2B is rational or B2C is emotional.
Contradiction is a feature
When sources disagree, investigate the explanation
Interviews say a feature is critical, but usage is low.
Do not automatically choose the dashboard or the interviews. The disagreement creates competing hypotheses: the feature may be hard to discover; the interview sample may overrepresent a niche; instrumentation may be incomplete; the feature may matter rarely but critically; users may solve the job outside the product; or stated importance may be inflated.
Next move: identify which explanation would materially change the decision, then collect evidence that can distinguish those explanations.
Worked example · explicitly hypothetical
Should a small B2B project tool build a client portal?
Several agency customers request “a client portal.” Treat the request as evidence about an attempted solution, not as a requirement.
Pending decision
Should the team invest in a dedicated client-facing collaboration experience?
Unknown 1 — underlying job
Conversations with agencies that share updates externally could reveal different jobs: reduce status-update work, collect approvals, avoid full workspace access, or create a more professional presentation layer.
Unknown 2 — actual behavior
If instrumented, inspect shared links, exports, repeated copy/paste, comments and external invites. These are hypothetical evidence sources, not claimed current product data.
Unknown 3 — segment
Client-service agencies may differ from internal teams, and agencies with many concurrent clients may differ from freelancers. Do not collapse them into one persona by default.
Unknown 4 — alternatives
Email, PDFs, spreadsheets, messaging tools, status calls and other project tools can all compete for the same job. Competition is broader than direct products.
Unknown 5 — market relevance
Ask how many relevant teams may exist, whether the target segment pays for adjacent workflows, and which established constraints or alternatives matter. Do not invent a TAM number.
Contradiction to investigate
Suppose interviewees say a portal is important, but current shared-link behavior is rare. The current flow could be too hard, the request could come from a narrow group, “portal” could be shorthand for another need, or the job could happen outside the measurable product.
Illustrative implication: do not commit to a generic client portal yet. Narrow the opportunity to the agency/client-approval workflow and collect one discriminating evidence source before choosing solution scope.
Mini practice
What evidence would you seek next?
Make the call before opening the reasoning. The goal is to practice evidence fit, not memorize one “best” method.
Scenario A: Users drop during onboarding. What evidence would you seek next?
Inspect where the drop happens and talk to recent affected users about what they were trying to do. Behavior shows the pattern; conversations can reveal context. Neither source answers the whole question alone.
Scenario B: Five interviewees describe the same workflow problem. What remains unknown?
How broadly the pattern applies, which contexts or segments differ, and whether the problem is important enough to change the pending decision. The next move is not automatically ‘interview one more person.’
Scenario C: A competitor launches a feature your team does not have. What should you investigate?
Whether your users face the underlying job, how they solve it today, and whether the competitor changes expectations or alternatives. A competitor feature is market evidence, not a requirement.
Scenario D: A survey says 70% of respondents would use a feature. What should you inspect?
The sample, wording, selection bias, current behavior and workaround, and whether respondents showed any commitment beyond stated intent. Do not convert stated intent into a universal behavior forecast.
Just enough research
Research effort should match the decision
Consider the cost of being wrong, reversibility, current evidence, urgency, affected user scope and any material safety or regulatory risk. Do not turn those factors into a universal score.
Stop or pause when additional research is unlikely to change the current decision enough to justify its cost—or when the remaining uncertainty is better resolved by a real product test.
Small reusable scaffold
Write the research brief in six prompts
- Decision
- What choice is pending?
- Unknown
- What must we learn?
- Evidence
- What source could answer it?
- Limitation
- What will remain uncertain?
- Contradiction
- What would challenge the current story?
- Implication
- What changes if the evidence supports or weakens the hypothesis?
Research quality
Red flags that should slow down the story, not speed up the slide deck
- The research question has no pending decision.
- Participants do not match the behavior or context you care about.
- Leading or hypothetical questions dominate the evidence.
- Only supportive evidence is retained while contradictions are dismissed.
- Sample limitations are hidden.
- Market-size assumptions are presented as facts.
- Competitor features substitute for user evidence.
- AI output cannot be traced back to raw sources.
- The synthesis makes stronger causal claims than the evidence supports.
- The findings cannot explain what decision changes.
AI can accelerate processing; it does not remove verification
AI can help summarize transcripts, suggest codes or clusters, extract data and draft synthesis. Keep source traceability, review outliers, compare outputs with raw evidence, and follow informed participation where appropriate, confidentiality, sensitive-data minimization, company policy, and ethical handling of recordings or transcripts. Before uploading research material to an external AI tool, check what data is necessary and how the provider handles it.
Real CraftUp lesson
Why user & market understanding matters
Try the quiz questions taken directly from the published CraftUp lesson for this module.
Quick practiceThe guide continues below
Want another challenge? Practice another round.
Real CraftUp curriculum · no sign-up required
Full curriculum
30 lessons, grouped by the learning job they serve
The real lesson order stays intact. The phases below add orientation without pretending the first six lesson titles are a procedural playbook.
Phase 1 — Understand the user and market context
What context surrounds the product decision?
- 1Why user & market understanding matters
- 2The difference between user needs and wants
- 3How to calculate TAM, SAM, and SOM for market sizing
- 4Understanding competitor types
- 5How to run competitive analysis
- 6How to scan for market trends
Phase 2 — Model who, what job, and where friction occurs
Whose problem is this, in what context, and what are they trying to accomplish?
- 7Understanding user segmentation basics
- 8How to create practical user personas
- 9What is Jobs-to-Be-Done (JTBD)?
- 10How to create customer journey maps
- 11How to uncover user pain points
- 12Prioritizing customer problems
Phase 3 — Choose evidence methods
Which method fits the uncertainty I need to reduce?
- 13Qualitative research methods overview
- 14How to run good interviews
- 15Common interview mistakes
- 16How to design effective surveys
- 17How to extract insights from quantitative data sources
Phase 4 — Combine and interpret evidence
How do I build a clearer picture without pretending incomplete evidence is certainty?
- 18Triangulation of insights
- 19How to synthesize research into actionable insights
- 20What are fast, scrappy research methods?
- 21User research tools for solo product managers
- 22Using AI to process research
Phase 5 — Make research useful in real product work
How does research actually influence a product decision?
- 23The continuous discovery habit
- 24Customer empathy in practice
- 25User archetypes vs stereotypes
- 26When to do 'just enough' research
- 27Research in B2C vs B2B
- 28Red flags in research
- 29How to present research insights to stakeholders
- 30Why research gets ignored
Make the research change something
A synthesis is an interpretation, not objective truth
Keep the chain inspectable: source evidence, segment or context, limitation, contradiction, interpretation, and decision implication. A recurring theme can be plausible and useful without being proven causal.
Research has impact when the team can say what changes because of it: investigate further, deprioritize, reframe the problem, choose a segment, identify a risk, or design a product test. A polished deck is not the outcome.
Specialist handoffs
Go deeper only when the method becomes the main question
Behavior-first customer interviews
Go deeper on question construction, concrete stories and interview repairs when interviewing itself becomes the job.
JTBD interview depth
Use the specialist owner when you need a full jobs-to-be-done interview method rather than an introduction to evidence fit.
Research triangulation
Go deeper on combining evidence sources, disagreements and blind spots without declaring ‘qual + quant = truth.’
TAM / SAM / SOM
Work through market-sizing assumptions in the dedicated guide instead of turning this chapter into a sizing tutorial.
Need a discussion guide rather than method depth? Use the Interview Script Generator. Need survey-writing depth? Use the survey-design guardrails. Sample adequacy belongs in the interview sample-size and saturation guide.
Next learning step
Research informs strategy. It does not choose strategy for you.
Once you understand users, alternatives, constraints, segments, evidence quality and market context, the next job is deciding where the product should focus and which trade-offs follow from that evidence.
