Measure: Testing

# Testing AI Skill Measure Area · Testing Block (formerly Questioning) · Decision Map · Oct 23, 2026 (v1.1) --- ## 1. What the Skill Does The Testing skill helps teams put a hunch against real evidence. It is the third block in the Measure area of Glare's Decision Map, between Hunches and Findings. Hunches set it up. Findings read what it returns. The skill still runs as glare-measure-questioning so existing installs keep working. This block used to be called Questioning. Writing questions now happens in Hunches; Testing picks how to check them. A test is any deliberate check of a hunch against evidence the team didn't write itself: a study, a standard instrument, an expert review, live behavior, or reading data the team already has. Every test has three parts. | Part | What it holds | Examples | |---|---|---| | Techniques | How you test | First click, task success, prototype test, expectation probe, data analysis, Wizard of Oz | | Stacks | What you ask, ready-made | Research Stacks (SUS, SEQ, UX-Lite, NPS) and Design Stacks (Website, Mobile App, Product, E-commerce, Marketing) | | Tools | Where you run it | Helio, Maze, Optimal Workshop, UserTesting, Lookback | Before a test runs, the skill rates how strong its evidence will be. | Rating | What the test captures | Examples | |---|---|---| | Weaker | Stated opinion or intent | Preference questions, likelihood to use, NPS | | Moderate | Behavior in a simulated setting | First click, task success on a prototype, tree testing | | Stronger | Behavior or commitment in the real world | Feature stub clicks, presales, Wizard of Oz | ### The Hunch Rule No hunch, no test. Open-ended learning before anyone has a hunch is Collecting, in Define. An interview with no hunch behind it is Collecting; an interview that probes a specific hunch is Testing. Running the same test across versions or segments is Comparing, in Focus. The skill also names a leading and a lagging metric. Behavioral and Attitudinal metrics lead: they can be read in a test before launch. Performance and Intelligence metrics lag: they are read in the live product after launch. Naming both up front makes sure the lagging check still happens once the leading test looks good. ## 2. Business Benefit Every hunch gets a test that fits it, and every result carries its source and strength. That keeps the team from treating a survey answer like proof of behavior. This helps teams: - pick the fastest test that can settle the hunch - say up front how much the result can be trusted - use standard instruments without breaking their scores - plan the after-launch check before the first test runs - set up a test that Focus can repeat later ## 3. Skill Output When used correctly, the skill produces a Test Plan. The example below shows how this works for a mobile banking dashboard. | Field | Example Output (Mobile Banking Dashboard) | |---|---| | Hunch | Moving balance above the promo cards will raise first-click success on balance for habitual users | | Technique | First-click test on the redesigned home screen | | Stack | SEQ (Single Ease Question), asked as written after the task | | Tool | Helio, 100 habitual mobile banking users | | Evidence Rating | Moderate: behavior in a simulated setting | | Leading Metric | Success (first click on balance), read in Helio before launch | | Lagging Metric | Sessions ending without an action, read in analytics 60 days after launch | | Failure Mode to Watch | Rewording SEQ to fit the brand voice. Changing a standard instrument breaks its score. | | Next Step Handoff | → glare-measure-findings to call the hunch when results come back | ## 4. Prompt Strategies The prompts below show different ways to use this skill. Each example uses a mobile banking dashboard update. ### Prompt 1: Match Entry: Pick the right test *"Our hunch: moving the balance above promo cards will raise first-click success for habitual mobile banking users. Using the glare-measure-questioning skill, pick a technique, a Stack if one fits, and a tool, rate the evidence, and name the leading and lagging metrics."* **Why this works:** It matches the test to the hunch and its metric, and says how strong the answer will be before any money is spent. **Best for:** - turning a ready hunch into a Test Plan - teams that default to the same test every time ### Prompt 2: Strength Entry: Upgrade weak evidence *"We were going to ask users which mobile banking home screen they prefer. Using glare-measure-questioning, tell us how strong that evidence is for our hunch about finding the balance, and suggest a test that captures behavior instead."* **Why this works:** Stated preference is weak evidence for a Use call. The skill moves the team toward a test of what people do. **Best for:** - plans built on surveys alone - large decisions resting on stated intent ### Prompt 3: Data Entry: Test with what you already have *"We have six months of mobile banking analytics and support tickets. Using glare-measure-questioning, show us how to read them against our hunch about the buried balance, what that test can and can't prove, and when we'd still need a study."* **Why this works:** Reading existing data against a specific hunch is a real test, and often the fastest one. **Best for:** - tight timelines - hunches about live products --- *Glare Framework · glare-measure-questioning (Testing) · Measure Area. Handoffs: glare-measure-hunches · glare-measure-findings · glare-define-collecting (no hunch yet) · glare-focus-comparing (repeat the test) · glare-ux-metrics*

⬇ Download SKILL.md