Discovery and user research

How does a research question differ from an interview question?

A research question states what the team needs to learn, such as why renewals fail. Interview questions elicit specific experiences and evidence; asking users the research question directly often invites guesses.

How should product assumptions be prioritized for investigation?

By consequence if false, uncertainty, and cost or time to learn. Test assumptions that could invalidate the investment before polishing details that matter only if the main proposition works.

What evidence supports a recurring user problem?

Concrete episodes, frequency, consequences, current workarounds, and observations across relevant users. Repeated requests for a feature are clues, but do not alone establish the underlying problem or its priority.

Why ask about the last time a user completed the task?

A recent concrete episode provides sequence, context, tools, and consequences that can be probed. General questions about what users 'usually' do encourage idealized summaries and omit exceptions.

Why is 'Would this easier workflow save you time?' a weak question?

It embeds the desired judgment and encourages agreement. Ask the participant to perform or describe the workflow, then examine actual effort, errors, and tradeoffs without supplying the answer.

What can a claim like 'I would definitely use this' establish?

Interest under the described scenario, not future adoption. Seek stronger evidence through realistic tasks, commitments, trials, or observed behavior while accounting for switching costs and missing product context.

How can an interviewer investigate a surprising answer without steering it?

Ask for an example, what happened next, or what led to that choice. Separate clarification from introducing an explanation the participant may simply agree with.

Why investigate spreadsheets, manual steps, and unofficial tools?

They reveal unmet needs, constraints, and the cost users already accept. A workaround may also be efficient or trusted, so compare the proposed product with its full value rather than assuming replacement is desirable.

How can a rare problem outrank a frequent annoyance?

It may cause much larger harm or block a critical task. Estimate affected users, frequency, consequence, and existing alternatives separately instead of ranking solely by mentions.

What makes a research recruitment plan representative of the decision?

Include the user types, contexts, and behaviors relevant to the target, including nonusers or failed users when needed. Recruiting only enthusiastic existing customers narrows the conclusions.

Why interview people who chose not to adopt?

They can reveal unmet prerequisites, switching costs, trust issues, or irrelevant value propositions absent from active-user interviews. Their reasons should be compared with observed market and product evidence, not assumed to represent everyone.

What can be learned from churned users that active users may not reveal?

The conditions under which the product stopped being useful, the replacement chosen, and failed recovery attempts. Account for recall and response bias, especially when only a small subset agrees to talk.

Why distinguish buyer, administrator, and end user in research?

They may have different goals, constraints, and authority. A workflow users like may fail procurement or administration, while a buyer's requested feature may add little value for daily users.

What should a participant screener avoid?

Questions that reveal the desired answers or select only people already favorable to the idea. Verify relevant behaviors and context without making eligibility depend on agreeing with the product hypothesis.

How can research incentives affect interpretation?

They support participation but can encourage misrepresentation or select people with particular availability. Use consistent incentives and screening, and distinguish compensated study participation from willingness to pay for the product.

What should participants understand before a research session?

Its purpose, what data or recordings are collected, how they will be used and accessed, and their ability to decline or stop. Use the organization's appropriate consent and retention process rather than treating participation as blanket permission.

How can a usability study avoid collecting unnecessary sensitive information?

Use realistic synthetic records or controlled accounts and ask participants not to reveal secrets. If real data is necessary, define access, handling, and retention in advance rather than relying on later redaction.

What does contextual observation add beyond an interview?

It exposes interruptions, handoffs, physical constraints, and unspoken steps in the actual environment. The observer can affect behavior, so record context and distinguish observed actions from inferred motives.

When is a diary study useful?

When relevant experiences occur over days or weeks and are difficult to observe in one session. Define lightweight prompts and timing, and account for missed entries and the burden of self-reporting.

Can ten qualitative interviews estimate what percentage of the market has a problem?

Not reliably from a small purposive sample. They can uncover mechanisms and patterns; prevalence estimates require an appropriate quantitative sampling and measurement design.

What does repeated discovery of the same themes suggest, and what does it not prove?

The current sample may be yielding fewer new themes for the scoped question. It does not prove coverage of all segments or edge cases; assess who is missing before stopping recruitment.

Why ask one concept per survey question?

A question such as 'Is the tool fast and easy?' cannot distinguish speed from usability. Use clear, bounded questions and response options that match the construct being measured.

Why specify a time window in a frequency question?

'How often do you export?' can mean different periods to different people. A defined window supports interpretation, though rare events and recall limitations may require logs or a diary instead.

Why can a survey sent to every customer still be biased?

Only some customers respond, and response propensity may relate to satisfaction or usage. Compare respondents with the eligible population and state coverage limits before presenting the result as universal.

How can survey order change answers?

Earlier questions can prime concepts, anchors, or dissatisfaction that influence later responses. Pilot the instrument, separate sensitive prompts appropriately, and randomize order where it helps without damaging comprehension.

How should raw research notes become findings?

Group evidence into patterns, retain counterexamples, and link each claim to source observations. Distinguish direct evidence, interpretation, and proposed action so a recommendation is not mistaken for something participants said.

Why should a memorable quote not determine priority by itself?

It may come from an atypical user or reflect an emotional episode rather than broad impact. Use the quote to illustrate a supported pattern and preserve its context, not as a substitute for prevalence or consequence.

Interviews favor a feature but usage is low. What should be investigated?

Instrumentation, discoverability, audience mismatch, workflow fit, and the difference between stated interest and actual effort. The disagreement is useful evidence to resolve, not a reason to discard either source automatically.

What makes a research segment useful?

A difference in needs, behavior, constraints, or decision context that changes product choices. Demographic labels alone may not explain the workflow, and overly narrow segments can overfit a small sample.

What should a journey map distinguish?

User goals, actions, channels, handoffs, friction, and supporting evidence across time. Mark unknown steps and differences between segments instead of presenting a polished invented journey as observed fact.

What does a service blueprint add to a user journey?

It connects visible user steps to backstage processes, systems, and owners. This helps locate failures caused by operations or dependencies that a screen redesign alone cannot fix.

How does an opportunity statement avoid prematurely specifying a solution?

It names a user need or obstacle and its context, such as difficulty reconciling disputed charges. 'Add an AI assistant' names an implementation before the need and success evidence are established.

How should prototype fidelity match the research question?

Use enough realism to test the uncertainty: rough flows for concept structure, realistic interaction for usability, and actual performance for latency-sensitive behavior. Visual polish cannot validate backend feasibility or long-term adoption.

What makes a usability-test task informative?

A realistic goal with necessary context, without telling the participant which controls to use. Observe completion, errors, hesitation, and recovery; asking someone to 'click Export' tests instruction following more than discoverability.

Why record every moderator intervention during a usability test?

Assistance changes what independent users could accomplish. Report aided and unaided completion separately and inspect where intervention was needed rather than counting all finished tasks as successful design.

What limitation comes with asking users to think aloud?

Speaking can alter pace and behavior, and people cannot always verbalize their reasoning accurately. Use it to learn interpretations while treating measured task time and stated motives cautiously.

How should a usability problem's severity be assessed?

Consider whether it blocks the goal, how often it occurs, recovery difficulty, and affected users. A visually obvious issue is not automatically more serious than an infrequent irreversible mistake.

Why include disabled users in relevant research rather than relying only on automated checks?

Real assistive-technology workflows expose interaction and comprehension barriers that scanners miss. Recruit for the service's users and contexts, and combine observed usability with standards-based accessibility testing.

What can a concept test establish before a working product exists?

Whether participants understand the proposition and see relevance or concerns under the described context. It does not establish actual performance, sustained use, or economic viability without further evidence.

What must be considered before a fake-door demand test?

User expectations, transparency, potential harm, and what a click actually measures. Provide an honest next step and avoid collecting commitments or payments under false pretenses; curiosity is weaker evidence than fulfilled demand.

What is a concierge test useful for?

Manually deliver the intended value to learn whether the outcome matters and what workflow is required. Track manual effort and exceptional handling so apparent user success is not mistaken for a scalable operating model.

What must be separated when interpreting a Wizard-of-Oz test?

The user's experience of a simulated capability and the feasibility of automating it. Human operators may hide latency, quality, or cost limits; disclose and manage the study appropriately and test feasibility independently.

What evidence is stronger than a compliment about a prototype?

A relevant action with real effort or tradeoff, such as completing onboarding, migrating sample data, or agreeing to a credible pilot. Even commitments need follow-through; they are not guaranteed recurring adoption.

When should engineering participate in discovery?

When assumptions involve data availability, latency, integration, reliability, or technical constraints. A small feasibility test can prevent research from validating an experience the system cannot economically deliver.

What does triangulating evidence mean in product research?

Compare observations from different methods or sources that have different weaknesses, such as logs, interviews, and support cases. Agreement raises confidence only if the sources are not all repeating the same biased sample.

How do you decide when a discovery effort has enough evidence?

Define the decision threshold and remaining material uncertainty, then assess whether another study could change the choice. Stop when additional learning has low decision value, not merely when a fixed number of interviews is reached.

What makes a research repository reusable rather than a pile of recordings?

Searchable findings linked to methods, dates, participant context, evidence, limitations, and decisions. Apply access and retention controls and mark stale findings when the product or audience changes.

How can a team actively look for evidence against its preferred idea?

Recruit relevant skeptics and failed users, test the hardest assumption, and define what result would stop or change the work. Record counterexamples rather than explaining each away after the fact.

What should a discovery readout lead with?

The decision supported, key evidence, confidence and limitations, and the next unresolved assumption. Include contrary evidence and distinguish observed needs from the team's proposed implementation.