Findings

Calling the hunch, then turning evidence into directionData is not a signal. A chart of drop-offs or a stack of survey scores only says what happened. Findings are where those results become evidence: first a verdict on the hunch that was tested, then a design signal that says what to do next.Without findings, teams drown in dashboards and debates. With them, every result ties back to the hunch it was meant to test and to the user need and business goal behind it.What Findings answersEach facet ends in a block that judges one quality, as people experience it. Findings closes Measure, and the quality it judges isusefulness: "this helps me do what I came to do."Question it answersWhat it checksWhat it hands upDid the design do what we intended for people?The hunch's prediction against the UX metric it namedA signal: useful or not, and for whomFindings is accountable forUsesignals: whether people can understand the design, find their way, and finish the task. A single test can also turn up Need or Prefer signals. Record them, and pass them to the block that judges them: Need signals go to Define: Situations, and a preference between options goes to Focus: Bets.Usefulness here is the quality the block judges. It is not the same thing as the Usefulness UX metric, which is one attitudinal score that can feed a Use signal.Findings in MeasureFindings are the output of Measure. Concepts name what is being designed. Hunches state the belief and its questions. Testing puts them against real evidence and records how. Findings read what came back.Findings receivesFromFindings producesThe hunch, its type, its perspective, and its metric (a UX metric, or a Product or Business KPI for an internal hunch)HunchesA verdict: Supported, Challenged, Refuted, or No callThe Test Plan and its results, with sources and evidence ratingTestingA Use signal: behavior, metric, context, and direction, with who it holds forThe user need and business goal behind the conceptConceptsA next move: one of the five decision typesWhat a Finding IsA finding is a tested result turned into direction. It opens with the verdict on the hunch, so everyone knows what the test settled, and then states the design signal.A strong finding:Names the hunch it answers, and calls itStates how strong the evidence is, from the Test PlanSays whether the call rests on the leading test, the lagging check, or bothSays who the result holds for, and where it splits by audienceLinks the result to a user need and a business goalEnds in a design signal and a next move the team can act onDesign SignalsA design signal forms when four parts come together: a user behavior is observed, a UX metric captures the reaction, the context explains why it happened, and a direction becomes clear. A metric alone shows what happened. Feedback alone describes what people said. The signal connects them.Glare has four signal types. Each bottom-row block judges one of them.Signal typeWhat it showsJudged inNeedWhether the concept connects to a real needDefine: SituationsUseWhether people can use it to get something doneMeasure: FindingsPreferWhether people choose it over the alternativesFocus: BetsAdoptWhether people keep using it over timeLead: ResultsTreat signal quality as a filter: a signal is usable or it is not. If it doesn't connect behavior to meaning, or doesn't point to a decision, it is still a metric.How to Turn a Test Into a SignalSTEP 1Call the HunchStart with the verdict. Read the results against the hunch and the metric it named, and weigh them by the evidence rating in the Test Plan.VerdictWhen to use itHunch statusSupportedThe metric moved the way the hunch predicted, on evidence strong enough for the decisionSupportedChallengedResults were mixed, or pointed somewhere elseChallengedRefutedThe results showed the hunch was wrongRefutedNo callThe test couldn't answer the hunch: the wrong technique, too small a sample, missing sourcesStays Ready to test; go back to TestingA Use call needs evidence of use: clicks, task success, errors, time, or a comprehension check. Stated preference alone is weak evidence for a Use call. It still counts as a Prefer signal, which Bets will need if the team compares options.A call made only on the leading test is provisional: once the change ships, read the lagging check against the same hunch and confirm or revise the call. An internal hunch is called the same way, against its Product or Business KPI and the function owner's numbers.STEP 2Translate Data Into a FindingRewrite the numbers as a plain statement of what happened and where the data came from.Data:48% drop-off at the payment step (checkout analytics).Finding:Analytics show nearly half of users abandon checkout when choosing a payment option.STEP 3Say Who It Holds ForSplit the result by the audiences in the test before you generalize it. A design can be useful for one group and not another, and the split often matters more than the average.For whom:Abandonment is highest for first-time buyers on phones (61%) and lowest for returning buyers with a saved card (22%).STEP 4Tie It to the User NeedName the user need the finding reveals or threatens, using the User Needs categories: Basics, Trust, Personal, Impact, or Feelings.User need:Basics. People need to pay without confusion or errors.STEP 5Tie It to Business ResultsShow how acting on the finding moves outcomes such as conversion, retention, efficiency, or revenue.Business result:Reducing abandonment raises checkout conversion and revenue.STEP 6Write the Design SignalPut the four parts together and name the signal type.PartExampleBehaviorFirst-time buyers hesitate and leave at the payment stepMetric48% drop-off; Error Rate rises on the card formContextThree payment options with similar labels appear at onceDirectionSimplify the payment options and label them by what people pay withSignal typeUseSignal:Three similar payment options stop first-time buyers at the payment step. Simplifying them should lower the error rate (UX metric) and raise checkout conversion (business KPI).STEP 7Choose the Next MoveA single test ends in one of five decision types. They are Findings' tactical next move, and the readout states which one the team chose.Decision typeUse it whenGoes toImplementSupported, on evidence strong enough for the size of the changeBuild, then Lead: Results checks the lagging measureRefine DesignThe direction is right, but the current version is not strong enough yetConcepts, then Testing againTest IterationThe evidence is useful but can't settle the callTestingRevisit LaterThe idea has potential, but the timing or context is wrongRecorded, with what would bring it backDo Not PursueRefuted, or not worth more investmentHunches, or Define if the need was wrongTwo cases go elsewhere:The team now wants to compare options, versions, segments, or time:hold the test fixed, name the variable, and go to Focus: Variables. When the comparison is read, the commitment is a bet, and a bet needs a Prefer signal for every option.Refuted, and the need itself looks wrong:go back to Define: User Needs or Situations.Why This Step MattersUX metrics tell you what happened.The verdict tells you what the test settled.The for-whom split tells you where it holds.Signals tell you what to do next.Without findings, teams stall in debates. With them, ideas earn their way forward.Links10 steps to turn user research insights into action, by Michele RonsenUX Research: How To Get From Data To Insights To Actions, by Ethan ParrySynthesizing UX Research: Making What's Mysterious Clear, by Nikki Anderson-StanierA Radical Proposal: Put UX Research In Charge, by Jared M. SpoolProduct Discovery Basics: Everything You Need to Know, by Teresa TorresThe 4 UX metrics that matter, by Nurkhon AkhmedovObed Rosassays,Its consistency in how my team delivers the outputs of their research and make non uxers understand it.AI PromptThis prompt helps you translate raw research data into a signal your team can act on.Start with a data set, test result, or research output you need to make sense of. It guides you to:Describe the behavior your data shows and name the sourceTie it to the specific user need it reveals or threatensConnect it to a business metric it affectsWrite a recommendation with both a user metric and a business metric attachedYou'll end with a completed signal that gives your team a clear direction and a recommendation ready to share.Use this after any test, research session, or analytics review when the team has data but not yet a decision.AI SkillsThe Findings skill file teaches your AI the full data-to-signal translation chain so it can help you close the loop between any data set and a clear, actionable recommendation.Load it when you need to go deeper on connecting findings to business results, cross-checking against the original hunch, or structuring findings so they travel beyond the design team. It gives your AI:The five-step translation chain from raw data through to a shareable signalThe user value mapping guide for tying findings to specific Honeycomb needsThe business results connection framework with a worked checkout exampleThe signal definition test for knowing when a finding becomes a signal versus noiseDownload the skill file below to use the full Findings framework with your AI assistant.

Related links

Vy Alechnavicius

Video tour of UX research methods beyond the standard interview and A/B test, including diary studies and participatory design. Useful when a team is over-reliant on the same two methods and missing other useful signals.

Purnimaa Arya

Walks through how data-informed design uses both quantitative and qualitative input to inform decisions and iterate fast. Useful when a designer is trying to balance numbers and user feedback in their daily process.

Jeff Sauro

Jeff Sauro examines five common NPS criticisms, including scale changes, predictive validity, and benchmark loss, with a balanced take on what to keep. Useful when teams debate dropping or changing NPS and you want a research-grounded view.

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