A signal is evidence that has been read and given a meaning, so it can move a decision forward. It says what the evidence shows and whether it supports or contradicts what the team was testing.
For example, SecondBank's rewards test found that only 42% of new cardholders could say what their points were worth. The signal is that point value is unclear, which supports the team's hunch that points feel abstract.
Why it matters
Raw evidence piles up fast: notes, numbers, quotes. Without someone saying what it means, leaders either drown in detail or skip it and go with instinct. The same data can also be read two ways, and nobody notices until the decision is already made.
Signals turn evidence into something a leader can act on. Each one states its meaning plainly, points back to the observations and metrics behind it, and says which test it helps or hurts. That makes it easy to see how well a decision is supported, and by how much.
Write each signal as a clear reading of the evidence, judge how strong it is, and send it to the stage that needs it next.
How we write one
Every signal is written as one sentence that names the evidence, what it means and which test it affects:
The evidence shows an interpretation; supports or contradicts a test.
Low task success on the value question shows point value is unclear; supports the abstract-points test.
Writing it this way keeps the evidence and the meaning side by side, so anyone can check whether the reading is fair, and the last clause shows which way it pushes.
How it differs from similar ideas
-
An observation reports what was seen or heard, with no meaning added. A signal interprets one or more observations or metrics.
-
A UX metric is a measured value. A signal says what that value means for a test.
-
An assumption is a conclusion taken as true before checking. When a test checks an assumption and it holds up, it becomes a signal.
-
A hunch is a belief that may not rest on data. A signal always rests on evidence from testing or measurement.
-
A recommendation suggests what to do. A signal says what the evidence means; the recommendation acts on it.
A good signal
-
Cites specific observations or metrics
-
States one clear interpretation
-
Says whether it supports or contradicts a named test
-
Names whose evidence it is
-
Is judged for strength, not just stated
-
Goes to the stage that needs it next
What happens to it
-
Raised. Ray or the team reads observations and metrics from a round of testing, or from a bet's results, and states what they mean.
-
Judged. Its strength is scored from the sample, the benchmark and the audience mix.
-
Sent on. Signals from Findings go to Initiatives in Focus; signals from Bets go up to Business Goals in Lead.
-
Acted on. An initiative, bet or decision cites it as evidence.
-
Strengthened or weakened. New evidence adds to it or contradicts it, and its strength is rescored.
For Ray
This part tells Ray how to find signals in an input, such as a test readout, a findings deck, an experiment report, a research synthesis or a retro note, and how to store them.
How to find one
-
Mark statements of meaning. Read the input and mark every place where someone says what the evidence shows, means or suggests.
-
Find the evidence. Identify the observations and UX metrics behind each statement, and link to them. If none can be found, hold it as a Hunch instead.
-
Find the interpretation. State plainly what the evidence means, in one clause.
-
Find the test. Identify the test, hunch or assumption it bears on, and whether it supports or contradicts it.
-
Judge the strength. Score it using the strength scale below, based on sample, benchmark and audience mix.
-
Find where it goes. Mark whether it came from Findings, headed to an Initiative, or from a Bet, headed to a Business Goal.
-
Write the sentence. Use the format: [Evidence] shows [interpretation]; [supports or contradicts] [test].
-
Fill the record, score it, link it and store it using the rules below.
Cues to look for
-
Interpretive language: "this shows", "suggests", "tells us", "confirms", "points to"
-
A number or pattern followed by a meaning: "only 42% could... so value is unclear"
-
Statements that a test passed, failed, held up or was disproved
-
Synthesis headings in a findings deck: "key insights", "what we learned"
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>Reports what happened with no meaning attached</p></td><td colspan="1" rowspan="1"><p>Observation</p></td></tr><tr><td colspan="1" rowspan="1"><p>Gives a measured value with no reading of it</p></td><td colspan="1" rowspan="1"><p>UX Metric</p></td></tr><tr><td colspan="1" rowspan="1"><p>Interprets data that hasn't been tested yet</p></td><td colspan="1" rowspan="1"><p>Assumption</p></td></tr><tr><td colspan="1" rowspan="1"><p>States a belief with no evidence behind it</p></td><td colspan="1" rowspan="1"><p>Hunch</p></td></tr><tr><td colspan="1" rowspan="1"><p>Says what the team should do next</p></td><td colspan="1" rowspan="1"><p>Recommendation, linked to the 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 signal</p></td><td colspan="1" rowspan="1"><p>[Evidence] shows [interpretation]; [supports or contradicts] [test].</p></td></tr><tr><td colspan="1" rowspan="1"><p>Evidence</p></td><td colspan="1" rowspan="1"><p>The observations and metrics it rests on</p></td><td colspan="1" rowspan="1"><p>Links to Observation and UX Metric</p></td></tr><tr><td colspan="1" rowspan="1"><p>Interpretation</p></td><td colspan="1" rowspan="1"><p>What the evidence means</p></td><td colspan="1" rowspan="1"><p>Short text</p></td></tr><tr><td colspan="1" rowspan="1"><p>Direction</p></td><td colspan="1" rowspan="1"><p>How it bears on the test</p></td><td colspan="1" rowspan="1"><p>Supports, Contradicts, Mixed</p></td></tr><tr><td colspan="1" rowspan="1"><p>Test</p></td><td colspan="1" rowspan="1"><p>The test, hunch or assumption it affects</p></td><td colspan="1" rowspan="1"><p>Link to Test, Hunch or Assumption</p></td></tr><tr><td colspan="1" rowspan="1"><p>Audience</p></td><td colspan="1" rowspan="1"><p>Whose evidence it is</p></td><td colspan="1" rowspan="1"><p>Link to Audience; one or more of the four audiences</p></td></tr><tr><td colspan="1" rowspan="1"><p>Strength</p></td><td colspan="1" rowspan="1"><p>How strong the evidence is</p></td><td colspan="1" rowspan="1"><p>Strong, moderate, weak</p></td></tr><tr><td colspan="1" rowspan="1"><p>Origin and destination</p></td><td colspan="1" rowspan="1"><p>Where it came from and where it goes</p></td><td colspan="1" rowspan="1"><p>Findings to Initiative, or Bet to Business Goal</p></td></tr><tr><td colspan="1" rowspan="1"><p>Source</p></td><td colspan="1" rowspan="1"><p>The input and exact passage or slide</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, Judged, Sent, Acted on, Weakened</p></td></tr><tr><td colspan="1" rowspan="1"><p>Confidence</p></td><td colspan="1" rowspan="1"><p>How clearly the signal was stated and traced</p></td><td colspan="1" rowspan="1"><p>High, medium, low</p></td></tr></tbody></table>
New signals start as Raised. A signal can't be marked Sent until its strength has been judged.
Scoring confidence
Confidence measures how clearly the signal was stated and how well it traces to its evidence. Strength, scored separately below, measures how convincing that evidence is.
<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>The interpretation is stated, linked to specific observations or metrics, and tied to a named test</p></td></tr><tr><td colspan="1" rowspan="1"><p>Medium</p></td><td colspan="1" rowspan="1"><p>The interpretation is clear, but the evidence or the test it bears on has to be inferred</p></td></tr><tr><td colspan="1" rowspan="1"><p>Low</p></td><td colspan="1" rowspan="1"><p>The meaning is implied, or the evidence behind it can't be found in the source</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 strength
Strength measures how convincing the evidence is: how many people it came from, whether it beats a benchmark, and whether different audiences agree.
<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>Strong</p></td><td colspan="1" rowspan="1"><p>An adequate sample, a clear move against a benchmark or expected result, and agreement across more than one audience or method</p></td></tr><tr><td colspan="1" rowspan="1"><p>Moderate</p></td><td colspan="1" rowspan="1"><p>A reasonable sample or a clear benchmark comparison, but evidence from a single audience or method</p></td></tr><tr><td colspan="1" rowspan="1"><p>Weak</p></td><td colspan="1" rowspan="1"><p>A small sample, no benchmark, or evidence from internal voices only, such as the project team or stakeholders</p></td></tr></tbody></table>
Weight Customers and Participants above the Project Team and Stakeholders when judging strength. Weak signals are still stored, but can't carry a bet on their own.
Linking
-
Check for duplicates first. If an existing signal states the same interpretation for the same test, add the new evidence and source to it and rescore its strength.
-
Link to its evidence, the observations and metrics it rests on, and to its source input.
-
Link to the test it supports or contradicts. When a tested assumption holds up, link the assumption to the new signal and mark the assumption as checked.
-
Link to its destination: the initiative it was sent to from Findings, or the business goal it reports to from a bet.
-
Note conflicts. If two signals read the same evidence differently, or point in opposite directions, link them as conflicting.
When to store and when to ask
-
Store automatically when confidence is high and strength is moderate or strong.
-
Add to the review digest when confidence is medium or low, when strength is weak, when it contradicts an open test or an existing signal, or when it's the main evidence for a bet.
When to flag
-
A signal has no observation or metric behind it
-
An initiative or bet rests only on weak signals
-
Two signals read the same evidence in opposite ways
-
A signal contradicts a test that is still treated as passing
-
A signal was raised but never sent to an initiative or business goal
Worked example
Input: a findings deck from the SecondBank rewards value test, slide 6, "What we learned".
"Only 42% of participants could say what 5,000 points buys (expected 70%), and 7 of 10 scrolled past the balance. Point value just isn't landing. This backs up the abstract-points hunch."
What Ray finds:
-
"42% task success" and "7 of 10 scrolled past" are the evidence, already stored as a UX metric and an observation.
-
"Point value just isn't landing" is the interpretation: point value is unclear.
-
"Backs up the abstract-points hunch" means it supports the abstract-points test.
-
Ten participants against a 70% expected result is a clear miss, but from one audience only, so strength is moderate.
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>Low task success on the value question shows point value is unclear; supports the abstract-points test.</p></td></tr><tr><td colspan="1" rowspan="1"><p>Evidence</p></td><td colspan="1" rowspan="1"><p>Task success 42% (n=10); 7 of 10 scrolled past the balance</p></td></tr><tr><td colspan="1" rowspan="1"><p>Interpretation</p></td><td colspan="1" rowspan="1"><p>New cardholders can't tell what their points are worth</p></td></tr><tr><td colspan="1" rowspan="1"><p>Direction</p></td><td colspan="1" rowspan="1"><p>Supports</p></td></tr><tr><td colspan="1" rowspan="1"><p>Test</p></td><td colspan="1" rowspan="1"><p>Test the abstract-points hunch with 10 new cardholders; expect 70% task success</p></td></tr><tr><td colspan="1" rowspan="1"><p>Audience</p></td><td colspan="1" rowspan="1"><p>Participants: new cardholders, first 90 days</p></td></tr><tr><td colspan="1" rowspan="1"><p>Strength</p></td><td colspan="1" rowspan="1"><p>Moderate: clear miss against expected result, single audience</p></td></tr><tr><td colspan="1" rowspan="1"><p>Origin and destination</p></td><td colspan="1" rowspan="1"><p>Rewards value test Findings to the rewards visibility Initiative</p></td></tr><tr><td colspan="1" rowspan="1"><p>Source</p></td><td colspan="1" rowspan="1"><p>Findings deck, slide 6, October 2026</p></td></tr><tr><td colspan="1" rowspan="1"><p>Status</p></td><td colspan="1" rowspan="1"><p>Judged</p></td></tr><tr><td colspan="1" rowspan="1"><p>Confidence</p></td><td colspan="1" rowspan="1"><p>High: interpretation stated, evidence linked, test named</p></td></tr></tbody></table>
Next step: stored automatically and sent with the Findings to the Initiative to improve rewards visibility for new cardholders by end of Q1.
Edge cases
-
Mixed evidence. If the metrics support a test but the quotes contradict it, store one signal with direction Mixed, or two signals linked as conflicting. Don't average them away.
-
Signal without evidence. "Users clearly hate points" with nothing behind it is a Hunch. Store it that way until evidence is linked.
-
Assumption confirmed. When a test checks an assumption and it holds, create a signal from the result and link it to the assumption. Don't rewrite the assumption itself.
-
Signals from bets. After a bet ships, a signal such as "redemptions rose with tab visits" goes up to the business goal it serves, not back to an initiative.