Testing

Put Hunches to the TestTesting is where a hunch meets proof. Hunches state what the team believes about a concept and write the questions that would prove it. Testing puts those questions in front of real evidence: people using the design, a standard instrument, or data the team already has. Findings read what comes back.A good test shows not just what people say, but what they do. Measure runs one test to get an answer. Focus repeats that same test while one thing changes. Testing is the one test, set up well enough that Focus can repeat it.What Good Testing DoesSurfaces hidden user challenges.Avoids bias so feedback reflects reality, not team assumptions.Guides the design process with targeted insights.Translates raw curiosity into measurable data.Here’s how to start testing:Getting StartedBegin TestingMake It PracticalSee It in ActionApply and EvolveGETTING STARTEDWhy Testing MattersA test is any deliberate check of a hunch against evidence the team did not write itself. That covers more than usability sessions:Studies:first click, task success, card sorting, preference tests, expectation probesStandard instruments:SUS, SEQ, UX-Lite, NPS and the other Research Stacks, asked as writtenExpert review:PURE or a heuristic review, when there is no time for participantsLive behavior:a landing page, a feature stub, a presale, or a Wizard of Oz service that shows what people do, not what they sayData analysis:reading analytics, support tickets, or evidence from the Glare MCP against one specific hunchInternal hunches are tested too. Their questions go to the team, and their evidence is usually the business’s own numbers, such as past launch costs, support volumes, or sales data, read against the hunch and its Product or Business KPI.Every test has three parts. Questions the team writes for a hunch stay with the hunch. Stacks are picked off the shelf and run as written, because changing their wording breaks the score. A test can mix both: three custom questions plus a SEQ.PartWhat it holdsExamplesTechniquesHow you testFirst click, task success, prototype test, expectation probe, data analysis, Wizard of OzStacksWhat you ask, ready-madeResearch Stacks (SUS, SEQ, UX-Lite, NPS) and Design Stacks (Website, Mobile App, Product, E-commerce, Marketing)ToolsWhere you run itHelio, Maze, Optimal Workshop, UserTesting, LookbackThe line between Testing and Collecting is the hunch. Collecting, in Define, is open-ended learning gathered before anyone has a hunch. Testing starts because the team has one. An interview with no hunch behind it is Collecting; an interview that probes a specific hunch is Testing. No hunch, go to Collecting. Hunch, go to Testing.Findings then guide decisions and shape the next round of work.Evidence RatingTests don't all produce the same strength of evidence. What people say they would do is weaker than what they do, and what they do with a prototype is weaker than what they commit to with money or time. Rate each test before it runs, so Findings can say how sure it is.RatingWhat the test capturesExamplesWeakerStated opinion or intentPreference questions, likelihood to use, NPS, discovery surveysModerateBehavior in a simulated settingFirst click, task success on a prototype, card sorting, tree testingStrongerBehavior or commitment in the real worldFeature stub clicks, presales, mock sales, Wizard of Oz, conciergeMost Helio tests sit between weaker and moderate: fast, cheap, and early, which is often the right trade. Name it, and pair stated intent with a behavioral measure when the decision is large.Leading and Lagging MetricsA hunch names the UX metric it expects to move, and the type of that metric decides when a test can read it. Behavioral and Attitudinal metrics are leading: they show what people do and feel with the design before it ships. Performance and Intelligence metrics are lagging: they show what the live experience did at scale, after launch.TimingMetric typesWhere it is readExamplesLeadingBehavioral, AttitudinalHelio studies, prototype and concept tests, Research Stacks, before launchSuccess, Comprehension, Effort, Intent, Sentiment, DesirabilityLaggingPerformance, IntelligenceAnalytics, product data, and AI logs, after launchCompletion Rate, Drop-off Rate, Error Rate, Time on Task, Recommendation QualityExample:for the hunch “removing two steps from signup for first-time users will increase completion,” the leading test reads Success and Effort in a Helio prototype test before the change ships. The lagging read is Completion Rate and Drop-off Rate at step 3 in analytics after it ships.When the decision is large, test on both. The leading test runs first, because it is fast and finds problems while they are cheap to fix. Name both metrics in the Test Plan before the first test runs, so the lagging check still happens once the leading test looks good. Timing and evidence rating answer different questions: a feature stub is a leading test with stronger evidence.What Testing Is NotTesting has clear edges. It sits between Hunches and Findings and does not take over their work.Collecting:open-ended learning before a hunch lives in Define.Comparing:running the same test across versions, segments, or time is Focus. An A/B test is a comparison, and experiment is the word Focus uses for it.Findings:Testing records what came back and how it was captured. What it means is Findings' work.Testing keeps Measure honest. Every hunch gets a test that fits it, and every result carries its source.BEGIN TESTINGStart From a Hunch and Its QuestionsQuestions for a hunch are now written inHunches: question types, modes, the bias check, and the tie to a UX metric. Testing starts once a hunch has its questions. The steps below are kept as a worked example of that question writing. Step 8 is where Testing takes over.Each step follows the rhythm of Measure:Concepts→Hunches→Testing→FindingsSTEP 1Clarify the conceptStrong questions begin with clarity. Use the hunches and assumptions defined in your earlier activities to shape what you need to prove or disprove.At this stage, your concept is the focus, the product, flow, or message you want to understand. Your hunches are the beliefs that guide it. Your goal is to make those beliefs visible so you can test them.Ask yourself:What assumptions are we making about how users will act or respond?What part of this concept do we most need to understand?What outcomes are we hoping to confirm or challenge?What would success look like if our hunch is correct?Clarity comes from defining what you believe now, so you can see what needs to be questioned next.STEP 2Identify knowledge gapsNow turn those hunches upside down. This step surfaces what youdon’tknow, the spaces where understanding is weak or uncertain.Ask:What assumptions haven’t been tested?Where do we lack proof or confidence?What do we think we know but can’t yet show with data?What would surprise us if it turned out to be false?Identifying gaps creates focus. You move from broad curiosity to specific uncertainty, and that uncertainty becomes the seed of a testable question.STEP 3Define learning goalsLearning goals translate uncertainty into direction. They describewhat kind of learningis needed: exploratory, evaluative, or comparative.ModeWhen to UseLearning GoalExampleExploratoryEarly stages, when little is known.Discover unmet needs or behaviors.“Understand how users choose between two plan types.”EvaluativeMid-stage, when testing an existing concept.Measure clarity, usability, or comprehension.“See if users can complete checkout without confusion.”ComparativeLater stages, when choosing between options.Decide which version performs better or builds more trust.“Compare two headlines to see which drives more sign-ups.”Learning goals define intent. They help you decide what kind of question to ask and what kind of signal to look for.STEP 4Write open-ended research questionsOpen-ended questions explore user perception and experience without assumption. They should help you understand behavior, language, and decision-making.Examples:What do users expect when they first see this page?How do they describe the purpose of this feature in their own words?What steps do they take when they get stuck?Avoid confirming what you already think. Instead, use questions that let users reveal how they actually understand the concept.Avoiding Bias and Leading QuestionsBias often sneaks in when teams write questions that reflect their own expectations. It leads users toward “right” answers and creates false confidence. Good questions stay neutral. It focuses on learning, not validation.Bias Check:Is this question leading users to say “yes”?Does it echo our internal assumptions?Is the language emotionally charged or too technical for real users?Neutral questions surface what people actually experience. They turn curiosity into honest feedback.STEP 5Translate into testable questionsOnce your research questions are clear, turn them into measurable prompts. A testable question should show observable behavior or perception.Examples:Can users find the pricing information within five seconds?Can users complete checkout without help?Do users understand what this feature does after reading the headline?Testable questions point directly to data. They make it possible to gather real evidence, not opinions.STEP 6Align questions with UX metricsEvery testable question should connect to a UX metric. This is how a hunch's questions link to measurement and signals.UX Metric TypeWhat It RevealsExample QuestionExample MetricBehavioralWhat users do.Where do users click first?First Click SuccessAttitudinalWhat users think or feel.How satisfied are users after completing the task?Satisfaction or Sentiment ScorePerformanceHow efficiently users succeed.How long does it take to complete checkout?Time on TaskIntelligenceHow well AI features help people.Do the AI suggestions help users finish the task?Recommendation QualityA good alignment makes every answer measurable and every result comparable across projects.STEP 7Prioritize and RefineMost teams have more questions than time, people, or budget to answer them. Prioritization ensures your effort goes where it matters most.Consider these filters:FactorWhat It RevealsHow to Use ItGap SizeThe degree of uncertainty or risk in your current understanding.Focus on areas with the least clarity or highest unknowns.ImpactThe potential value of answering the question.Prioritize questions tied to decisions that affect users or outcomes directly.FeasibilityThe time, cost, and resources available.Choose what can realistically be tested in the current cycle.ConfidenceThe team’s belief in current assumptions.Address low-confidence areas first to strengthen overall direction.Questions that sit at the intersection of large gap + high impact + feasible effort should rise to the top. Refine them for clarity. Remove duplicates or overlap. Keep only those that lead directly to action.STEP 8: TESTING TAKES OVERMatch the Hunch to a TechniquePick the technique that gives the clearest evidence for the hunch, add a Stack if one fits, and choose the tool to run it in. Note the evidence rating as you choose.TechniqueWhat it showsEvidence ratingExample questionFirst click, tree testWhere people go firstModerate“Where do users click first on this page?”Task success on a prototypeWhether people can finishModerate“Can users complete checkout without help?”Research Stack (SUS, SEQ, NPS)How people rate the experience, asked as writtenWeaker“How easy was this task?” (SEQ)Feature stub, presale, Wizard of OzWhat people commit to in the real worldStronger“Do people click to start the trial?”Data analysisWhat the live product did, read against the hunchLagging read“Did step 3 drop-off fall after launch?”Open-ended learning with no hunch behind it belongs inDefine: Collecting.MAKE IT PRACTICALPlan Your TestUse this Test Plan to set up your next test. Fill it in before the test runs, so Findings can read the results against it.PartWhat to recordExampleHunch and questionsThe hunch, its type, and the questions written in HunchesRemoving two signup steps will increase completion for first-time usersTechniqueHow you will test itTask success on a prototypeStackAny ready-made instrument, run as writtenSEQ after the taskToolWhere it runsHelioEvidence ratingWeaker, Moderate, or StrongerModerateLeading metricWhat the test reads before launchSuccess, EffortLagging metricWhat to check after launchCompletion Rate, Drop-off Rate at step 3SourcesWho or what the evidence comes from100 first-time users on mobileWhen complete, the Test Plan travels with the results to Findings.Quick ChecklistBefore you run your next test, confirm the essentials:The test answers one hunch, and its questions are written.We picked the technique that fits the hunch.Any Stack is run as written.The evidence rating is named.The leading and lagging metrics are named in the Test Plan.Every result will carry its source.If we are comparing versions, segments, or time, we have moved to Focus.Strong testing is less about volume and more about fit. If each test is tied to a hunch, a technique, a rating, and a metric, you are ready to run it.How to Integrate Testing Into WorkflowTesting becomes powerful when it is built into your daily and weekly work. Instead of a one-time activity, it becomes a habit that shapes how your team checks every hunch and measures every outcome.Each round of testing strengthens clarity. You learn which tests settle a hunch, which ones stall, and which lead directly to measurable signals. Over time, this practice builds discipline: you stop guessing and start learning with purpose.Turn Testing Into a Repeatable RoutineWeekly Discovery ReviewBegin each week by identifying which assumptions or user behaviors need clarity.Frame these into questions you can test or observe within the week.Sprint PlanningPair every major design or product task with at least one measurable question. Make it visible on the board,  “What do we need to learn before we move forward?”Post-Test ReflectionEnd each sprint or test cycle by reviewing which questions produced usable signals. Note which ones were too broad or biased, and refine them for the next round.Share the Questions, Not Just the AnswersWhen findings are shared, include the questions that led to them. This helps others understand your intent and replicate the method.Build a Question LibraryOver time, collect well-framed, tested questions in a shared workspace. Reuse them as templates for future projects to keep consistency in your measurement approach.Testing drives measurable improvement across teams. When every hunch gets a test that fits it, design moves faster, research becomes actionable, and results connect directly to user needs.BenefitWhat It DeliversFocusTeams align around what to learn next instead of debating opinions.SpeedClear questions shorten the time between testing and decisions.ConfidenceEach question backed by metrics creates evidence leaders can trust.ScalabilityReusing question frameworks builds consistency across projects.Testing becomes the foundation for every signal you collect. When your process is disciplined, curiosity becomes proof, and design gains the credibility it deserves.SEE IT IN ACTIONHow one test turned confusion into clarityAPPLY AND EVOLVEGetting It to StickWhen testing becomes a regular practice, design moves from assumption to evidence. Teams stop debating what to do and start proving what works. Each test creates sharper hunches, clearer findings, and stronger signals that guide design.The goal is not just to run more tests; it’s to build a culture where every decision starts with a hunch and ends with measurable proof.Business ImpactConsistent testing creates measurable advantages across design and product teams.BenefitWhat It CreatesExample ImpactFaster FocusClear learning goals eliminate wasted research.Teams cut discovery cycles by identifying what to test within hours instead of weeks.Better DecisionsQuestions tied to metrics create credible findings.A single comprehension test clarifies which layout works best, saving development time.Cross-Team AlignmentA shared way of testing connect research, design, and product decisions.Product, design, and marketing use the same questions to test message clarity.Proven ValueMeasurable questions link user insight to business results.Improved comprehension leads to higher conversion and reduced churn.Testing helps leaders see evidence of progress and gives contributors confidence that their work drives impact.ReflectionBefore moving to Findings, take a moment to reflect on what you learned through testing.Ask yourself and your team:Did each test answer one hunch?Did we pick the technique that fits the hunch, and name its evidence rating?Did we name both the leading and the lagging metric?How did the results change what we decided to do next?What will we test differently next time?Reflection turns testing into an evolving skill. Each project becomes easier to frame, test, and measure with confidence.Next StepsTesting sits in theMeasurefacet of Glare, between Hunches and Findings. From here, Findings calls the hunch and turns the evidence into a design signal.If You Are HereMove ToWhat You GainConcepts → HunchesState the belief and write the questions that would prove it.A hunch ready to test.Hunches → TestingPut the questions against real evidence.Results with sources and an evidence rating.Testing → FindingsCall the hunch: Supported, Challenged, Refuted, or No call.A Use signal and a next move.Findings → Next movePick one of five decision types: Implement, Refine Design, Test Iteration, Revisit Later, or Do Not Pursue. To compare options, go to Focus: Variables.A clear step the team can act on.Each test you run moves you closer to confidence. When teams test with precision, they see patterns faster, share evidence more clearly, and build proof that lasts beyond a single project.Closing ThoughtGood tests are the foundation of design measurement. They connect curiosity to clarity, intent to evidence, and intuition to influence. When your team tests well, design stops being opinion and becomes proof that earns trust.ResourcesUser Research Questions | UX Research Field Guide, by User Interviewshttps://www.userinterviews.com/ux-research-field-guide-chapter/user-research-questionsThe Best User Research Questions (+ How to Ask Them), by Mazehttps://maze.co/blog/research-questions/A Step-by-Step Guide to Writing Better Research Questions, by dscouthttps://dscout.com/people-nerds/guide-better-questionsHow to ask good questions and level-up your user research skills, by Envoy Designhttps://medium.com/envoy-design/how-to-ask-good-questions-and-level-up-your-user-research-skills-19fd28b275a4Crafting good UX research questions for qualitative research, by PlaybookUXhttps://www.playbookux.com/crafting-good-ux-research-questions-for-qualitative-research/The NCredible Framework — UX research questions prioritised, by UX Insighthttps://uxinsight.org/a-practical-tool-for-prioritising-research-questions-and-ensuring-that-they-dont-all-end-up-in-a-usability-test/How to ask the right UX questions and avoid the wrong ones (Reddit)https://www.reddit.com/r/userexperience/comments/oorfjq/how_to_ask_the_right_ux_questions_and_avoid_the/How to write effective UX research questions (With Examples), by Hotjarhttps://www.hotjar.com/ux-research/questions/AI PromptThis prompt helps you turn a hunch into a Test Plan that will produce usable signals.Start with a hunch, design decision, or area of uncertainty your team needs to investigate. It guides you to:Pick a technique that fits the hunchAdd a Research or Design Stack if one fits, run as writtenChoose the tool and rate the evidence: Weaker, Moderate, or StrongerName the leading and the lagging metricYou'll end with a Test Plan ready to run, with its evidence rating and the metrics Findings will read.Use this before any test to make sure it can actually settle the hunch.AI SkillsThe Testing skill (it still installs as glare-measure-questioning) teaches your AI to put a hunch against real evidence: picking the technique, Stack, and tool, rating the evidence, and naming leading and lagging metrics.Load it when you need to match a hunch to a technique, upgrade weak evidence, or plan the after-launch check. It gives your AI:The technique catalog, with an evidence rating for eachThe Research and Design Stacks, and when to use themLeading and lagging metrics by UX metric typeThe line between Collecting, Testing, and ComparingDownload the skill file below to use the Testing framework with your AI assistant.

Related links

Jonathan Richardson

Jonathan Richardson on testing content as part of user research. Useful when teams overlook content during research.

Jack O'Donoghue

Jack O'Donoghue shows how to identify, apply, and validate mental models in UX work. Useful when a research team has a mental model in mind but no way to test it.

Sophia V Prater

Shows how two questions ("What are the objects?" and "How do they relate?") get stakeholders to spot their own assumptions and want user research. Useful when stakeholders resist research and you need a way to pull them into the problem.

Identify where decision quality breaks down

The Glare Design Assessment helps teams spot weak validation, stakeholder friction, alignment gaps, and assumptions that scale without measurable learning—so you have a clearer starting point for improvement.

About 5 minutes · Team-based · Diagnostic snapshot you can act on

Take the Design Assessment