Hunch

A hunch is an intuition about what's true or what will work. It may have no data behind it at all, and that's fine: hunches are where many good ideas start.

For example, a designer on the rewards team says, "I think people ignore rewards because points feel abstract." Nobody has measured that yet. It's a hunch, and it's worth writing down because it suggests what to test next.

Why it matters

Every team runs on hunches, but most of them stay in people's heads. Some quietly turn into plans without ever being checked. Others get dismissed because the person who had them wasn't senior enough to be heard.

Writing hunches down makes them visible and fair. Anyone can raise one, from design, research, product or engineering. Each hunch has a perspective: an external hunch is about users and is read with UX metrics; an internal hunch is held by a business function and is read with that function's own KPI. Once it's on the record, the team can decide which hunches matter most, test them, and see clearly which ones held up and which didn't.

Capture the hunch in the person's own words, note what made them think it, and be clear about how much depends on it before anyone builds on top of it.

How we write one

Every hunch is written as one sentence that names the belief and the intuition behind it:

We think a belief because an intuition.

We think people ignore rewards because points feel abstract.

Writing it this way separates what someone believes from why they believe it, so the team can test the belief without having to defend the person. Every hunch is written with the questions that would prove it or kill it, so the test is ready when the hunch is.

How it differs from similar ideas

  • An assumption always cites data, such as low tab visits, and draws a conclusion from it. A hunch may rest on intuition alone.

  • A test is how a hunch gets checked. The hunch is the belief; the test is the plan for finding out whether it's right.

  • A signal is interpreted evidence. A hunch that holds up under testing is supported by signals, but the hunch itself is never evidence.

  • A concept is the proposal a hunch is written about. The concept comes first; the hunch is the belief about whether that concept will work.

A good hunch

  • States one belief in plain words

  • Gives the intuition behind it, even if it's only a feeling

  • Could be shown wrong by a test

  • Contains no solution

  • Links to the opportunity it tries to explain

  • Says who raised it

What happens to it

  1. Raised. Someone voices it, or Ray finds it in a meeting note, a chat thread or a retro.

  2. In test. A test is set up to check it, with an expected result.

  3. Supported. Test evidence matches the expected result, and signals link back to it.

  4. Contradicted. Test evidence goes against it. The hunch is kept, so the team remembers what didn't hold.

  5. Retired. It is no longer relevant, or a newer hunch has replaced it.


For Ray

This part tells Ray how to find a hunch in an input, such as a meeting transcript, a Slack thread, retro notes, a design critique or a kickoff doc, and how to store it.

How to find one

  1. Mark statements of belief. Read the input and mark every statement where someone suspects something is true or will work.

  2. Check for data. If the belief cites data, such as analytics or survey results, store it as an Assumption instead. If it doesn't, it's a hunch.

  3. Find the belief. Pull out the single thing being suspected. If several beliefs are bundled, split them.

  4. Find the intuition. Look for why the person thinks it: an experience, an analogy, a gut feeling. If none is given, leave it blank.

  5. Note who raised it. Record the person or role and their skill area.

  6. Judge the stakes. Check whether a concept, bet or decision already depends on it.

  7. Write the sentence. Use the format: We think [belief] because [intuition].

  8. Fill the record, score it, link it and store it using the rules below.

Cues to look for

  • Belief language: "I think", "my gut says", "I bet", "I suspect", "I have a feeling"

  • Explanations offered without numbers: "people probably...", "it's likely because..."

  • Analogies to other products: "it's like how airline miles..."

  • Side comments in critiques and retros that nobody followed up on

What doesn't count

<table xmlns="http://www.w3.org/1999/xhtml" style="min-width: 50px;"><colgroup><col style="min-width: 25px;"><col style="min-width: 25px;"></colgroup><tbody><tr><th colspan="1" rowspan="1"><p>If the statement...</p></th><th colspan="1" rowspan="1"><p>Store it as</p></th></tr><tr><td colspan="1" rowspan="1"><p>Draws a conclusion from data, such as "low tab visits mean..."</p></td><td colspan="1" rowspan="1"><p>Assumption</p></td></tr><tr><td colspan="1" rowspan="1"><p>Proposes something to build</p></td><td colspan="1" rowspan="1"><p>Concept</p></td></tr><tr><td colspan="1" rowspan="1"><p>Asks what the team should do</p></td><td colspan="1" rowspan="1"><p>Ask</p></td></tr><tr><td colspan="1" rowspan="1"><p>Reports what was seen in a test</p></td><td colspan="1" rowspan="1"><p>Observation</p></td></tr><tr><td colspan="1" rowspan="1"><p>Interprets test evidence</p></td><td colspan="1" rowspan="1"><p>Signal</p></td></tr></tbody></table>

The record

<table xmlns="http://www.w3.org/1999/xhtml" style="min-width: 75px;"><colgroup><col style="min-width: 25px;"><col style="min-width: 25px;"><col style="min-width: 25px;"></colgroup><tbody><tr><th colspan="1" rowspan="1"><p>Field</p></th><th colspan="1" rowspan="1"><p>What to fill in</p></th><th colspan="1" rowspan="1"><p>Values</p></th></tr><tr><td colspan="1" rowspan="1"><p>Statement</p></td><td colspan="1" rowspan="1"><p>The one-sentence hunch</p></td><td colspan="1" rowspan="1"><p>We think [belief] because [intuition].</p></td></tr><tr><td colspan="1" rowspan="1"><p>Belief</p></td><td colspan="1" rowspan="1"><p>What is suspected</p></td><td colspan="1" rowspan="1"><p>Short text</p></td></tr><tr><td colspan="1" rowspan="1"><p>Intuition</p></td><td colspan="1" rowspan="1"><p>Why the person thinks it</p></td><td colspan="1" rowspan="1"><p>Short text, or blank</p></td></tr><tr><td colspan="1" rowspan="1"><p>Opportunity</p></td><td colspan="1" rowspan="1"><p>The gap it tries to explain or close</p></td><td colspan="1" rowspan="1"><p>Link to Opportunity, or blank</p></td></tr><tr><td colspan="1" rowspan="1"><p>Raised by</p></td><td colspan="1" rowspan="1"><p>The person or role who voiced it</p></td><td colspan="1" rowspan="1"><p>Name or role</p></td></tr><tr><td colspan="1" rowspan="1"><p>Skill area</p></td><td colspan="1" rowspan="1"><p>Where it came from</p></td><td colspan="1" rowspan="1"><p>Skill area name</p></td></tr><tr><td colspan="1" rowspan="1"><p>Stakes</p></td><td colspan="1" rowspan="1"><p>How much depends on it</p></td><td colspan="1" rowspan="1"><p>High, medium, low</p></td></tr><tr><td colspan="1" rowspan="1"><p>Tests</p></td><td colspan="1" rowspan="1"><p>Tests that check it</p></td><td colspan="1" rowspan="1"><p>Links to Test</p></td></tr><tr><td colspan="1" rowspan="1"><p>Source</p></td><td colspan="1" rowspan="1"><p>The input and exact passage</p></td><td colspan="1" rowspan="1"><p>Link to Input</p></td></tr><tr><td colspan="1" rowspan="1"><p>Status</p></td><td colspan="1" rowspan="1"><p>Where it stands</p></td><td colspan="1" rowspan="1"><p>Raised, In test, Supported, Contradicted, Retired</p></td></tr><tr><td colspan="1" rowspan="1"><p>Confidence</p></td><td colspan="1" rowspan="1"><p>How clearly it was stated</p></td><td colspan="1" rowspan="1"><p>High, medium, low</p></td></tr></tbody></table>

New hunches start as Raised. A hunch moves to Supported or Contradicted only through a linked test and its signals, never by opinion.

Scoring confidence

Confidence measures how clearly the hunch was stated and attributed, not whether it's likely to be true.

<table xmlns="http://www.w3.org/1999/xhtml" style="min-width: 50px;"><colgroup><col style="min-width: 25px;"><col style="min-width: 25px;"></colgroup><tbody><tr><th colspan="1" rowspan="1"><p>Level</p></th><th colspan="1" rowspan="1"><p>When to use it</p></th></tr><tr><td colspan="1" rowspan="1"><p>High</p></td><td colspan="1" rowspan="1"><p>Stated directly by a named person, with a clear belief and the intuition behind it</p></td></tr><tr><td colspan="1" rowspan="1"><p>Medium</p></td><td colspan="1" rowspan="1"><p>Stated directly, but the intuition is missing or the belief is vague</p></td></tr><tr><td colspan="1" rowspan="1"><p>Low</p></td><td colspan="1" rowspan="1"><p>Inferred from an indirect remark, or the speaker can't be identified</p></td></tr></tbody></table>

If any part of the record can't be filled from the source, leave it blank and lower confidence by one level.

Judging the stakes

Stakes measure how much would break if the hunch turned out to be wrong.

<table xmlns="http://www.w3.org/1999/xhtml" style="min-width: 50px;"><colgroup><col style="min-width: 25px;"><col style="min-width: 25px;"></colgroup><tbody><tr><th colspan="1" rowspan="1"><p>Level</p></th><th colspan="1" rowspan="1"><p>When to use it</p></th></tr><tr><td colspan="1" rowspan="1"><p>High</p></td><td colspan="1" rowspan="1"><p>A concept, bet or decision already depends on it</p></td></tr><tr><td colspan="1" rowspan="1"><p>Medium</p></td><td colspan="1" rowspan="1"><p>It shapes an open test or an opportunity the team is actively exploring</p></td></tr><tr><td colspan="1" rowspan="1"><p>Low</p></td><td colspan="1" rowspan="1"><p>Nothing depends on it yet</p></td></tr></tbody></table>

A high-stakes hunch that hasn't been tested always goes to the review digest, whatever its confidence.

Linking

  • Check for duplicates first. If an existing hunch states the same belief, add the new source to it instead of creating another.

  • Link to the opportunity it feeds from and its source input.

  • Link to the tests that check it, and to any concept, bet or decision that depends on it.

  • Note conflicts. If two hunches predict opposite things, link them as conflicting so one test can check both.

  • Note promotions. If data later backs the belief, link the new Assumption to the hunch rather than rewriting it.

When to store and when to ask

  • Store automatically when confidence is high and stakes are low or medium.

  • Add to the review digest when confidence is medium or low, when stakes are high, or when it contradicts an existing hunch or open test.

When to flag

  • A concept, bet or decision depends on a hunch that has never been tested

  • A hunch has been In test for more than 30 days with no signals linked

  • A contradicted hunch is still linked to an active concept or bet

  • Two hunches conflict and no test checks either

  • A hunch is being treated as an assumption without any data behind it

Worked example

Input: a transcript from the weekly rewards design critique.

"Honestly, I don't think people ignore rewards because they can't find them. I think points just feel abstract. It's like airline miles: you have 40,000 of them and no idea what that means."

What Ray finds:

  • "Points just feel abstract" is the belief.

  • "It's like airline miles" is the intuition: an analogy, with no data.

  • "I don't think people ignore rewards because they can't find them" pushes against the existing Assumption that cardholders don't know rewards exist, so the two are linked as conflicting.

  • No concept depends on it yet, but it feeds the opportunity about seeing point value, so stakes are medium.

Stored record:

<table xmlns="http://www.w3.org/1999/xhtml" style="min-width: 50px;"><colgroup><col style="min-width: 25px;"><col style="min-width: 25px;"></colgroup><tbody><tr><th colspan="1" rowspan="1"><p>Field</p></th><th colspan="1" rowspan="1"><p>Value</p></th></tr><tr><td colspan="1" rowspan="1"><p>Statement</p></td><td colspan="1" rowspan="1"><p>We think people ignore rewards because points feel abstract.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Belief</p></td><td colspan="1" rowspan="1"><p>Points feel abstract to new cardholders</p></td></tr><tr><td colspan="1" rowspan="1"><p>Intuition</p></td><td colspan="1" rowspan="1"><p>Analogy to airline miles; no data cited</p></td></tr><tr><td colspan="1" rowspan="1"><p>Opportunity</p></td><td colspan="1" rowspan="1"><p>New cardholders need to see what their points are worth, but today the dashboard only shows a balance.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Raised by</p></td><td colspan="1" rowspan="1"><p>Senior product designer</p></td></tr><tr><td colspan="1" rowspan="1"><p>Skill area</p></td><td colspan="1" rowspan="1"><p>Design</p></td></tr><tr><td colspan="1" rowspan="1"><p>Stakes</p></td><td colspan="1" rowspan="1"><p>Medium: feeds an active opportunity</p></td></tr><tr><td colspan="1" rowspan="1"><p>Tests</p></td><td colspan="1" rowspan="1"><p>None yet</p></td></tr><tr><td colspan="1" rowspan="1"><p>Source</p></td><td colspan="1" rowspan="1"><p>Design critique transcript, October 2026, 14:20</p></td></tr><tr><td colspan="1" rowspan="1"><p>Status</p></td><td colspan="1" rowspan="1"><p>Raised</p></td></tr><tr><td colspan="1" rowspan="1"><p>Confidence</p></td><td colspan="1" rowspan="1"><p>High: stated directly by a named person, with the intuition given</p></td></tr></tbody></table>

Flag raised: added to the review digest because it conflicts with an existing assumption, with a prompt to set up one test that checks both.

Edge cases

  • Hunch with data. "I think it's abstract, and only 12% visit the tab" cites data. Store the conclusion from data as an Assumption and the remaining intuition as a Hunch, linked.

  • Stated as a solution. "I bet a progress bar would fix it" names a solution. Store the progress bar as a Concept and ask what belief it rests on.

  • A hunch from a senior voice. Seniority doesn't raise confidence or stakes. Score it like any other hunch.

  • Test results arrive. Don't edit the hunch. Link the signals, then move it to Supported or Contradicted.

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