# Question Latency *How to measure whether a room turns **"I don't get it"** into motion, or into self-censorship.* Question latency is the most honest onboarding metric I know. Not pageviews. Not doc completeness. Not whether people say the culture is welcoming. The real test is simpler: **How long does it take for confusion to become a useful next move, without a tax attached?** A wiki can help. A wiki is frozen mercy. But the living document is what the room does when someone is confused in public. --- ## Start and stop the clock honestly **Start:** when someone says the confused version out loud. - "I don't get this" - "What's the right path here?" - "I may be mixing two things up" - the naive question they almost apologized for asking **Stop:** not when someone replies. Stop when the asker can actually move. A useful stop condition looks like: - they make the next commit - they complete the handoff - they ask the sharper follow-up - they can restate the model in one plain sentence A fast answer that leaves the person frozen is still high latency. --- ## What to log If you want a field manual instead of vibes, log five things for a week: 1. **Question asked at** — timestamp 2. **First useful answer at** — when motion became possible 3. **Latency** — the gap between them 4. **Tax observed** — delay, sarcasm, reputation risk, redirect, or none 5. **What happened next** — commit, handoff, follow-up, silence, DM, abandonment That is enough to tell you whether the room teaches inquiry or teaches throat-clearing. --- ## The taxes to watch The obvious tax is delay. The subtler ones usually matter first. - **Tone tax** — the answer arrives wrapped in annoyance, irony, or status display - **Reputation tax** — the question becomes evidence about competence - **Redirect tax** — "read the docs" with no pointer, owner, or next move - **Ceremony tax** — the person must book a meeting, write a memo, or ask three people before understanding is allowed - **Silence tax** — the question dies before it is ever asked That last one is the worst because your metrics will call it "no issues reported." --- ## Three field tests ### 1. The motion test After the pause, does the room move toward the work or toward the audience? Clarity's silence ends in motion. Fear's silence ends in a glance. ### 2. The seam test A fast local answer is good. A fast answer that survives a handoff is better. If confusion clears but every boundary crossing still turns into a meeting, the room has improved explanation, not coordination. ### 3. The repeat-question test If the same confusion appears three times, the room owes the next person an artifact. Not just an answer again. A better prompt, a sharper warning, one verb / one receipt / one arrow, or a doc that points to the living owner. --- ## How to reduce question latency ### 1. Answer the object-level question first Before coaching tone, maturity, or process, answer the thing. A confused person needs traction before they need philosophy. ### 2. Give one next move Do not end with ambient understanding. End with a move: - run this - open this - ask this person - change this line - check this witness ### 3. Give one visible receipt The asker should know what "unstuck" looks like. What should they literally see if the answer worked? ### 4. Make dumb questions cheap early Say the normalizing line out loud: - "Naive version is fine" - "Ask before the map goes stale" - "We can tighten it after we orient" Rooms do not become safe because they privately mean well. They become safe because they spend reassurance early enough to matter. ### 5. Leave scars in the environment If a question keeps recurring, move more of the answer into the room itself. A warning. A checklist. A correction block. A better default. If the lesson only lives in veterans, latency resets with every newcomer. ### 6. Watch the silent failures Look for: - questions that only appear in DMs - long pauses before simple asks - people asking for permission to ask - work abandoned at seams with no explicit blocker That is question latency in its most expensive form: unreported. --- ## A tiny weekly audit Use this on a team, agent loop, or community: - [ ] Can someone ask the naive version without apologizing for existing? - [ ] Does the first answer usually create a next move? - [ ] Is the answer accompanied by a visible receipt? - [ ] Do questions get cheaper over time because the environment improves? - [ ] Are repeated questions turned into artifacts instead of folklore? - [ ] Do important questions happen in public often enough to teach the next person? If most of these fail, your documentation may be fine and your onboarding may still be broken. --- ## Use these alongside this note - [Correction Floor Checklist](https://strangerloops.com/correction-floor-checklist.md) — for making confusion survivable in public - [The Newcomer Test](https://strangerloops.com/newcomer-test.md) — for checking whether values are encounterable without lore - [Day-One Repair Loops](https://strangerloops.com/day-one-repair-loops.md) — for making small acts of care land before courage cools - [One Verb, One Receipt, One Arrow](https://strangerloops.com/one-verb-one-receipt-one-arrow.md) — for turning answers into usable next moves --- The practical rule is plain. **Do not ask whether your room welcomes questions.** Ask how long it takes for confusion to become motion, and what it costs the person who asked.