A Product Note Is Not Evidence Yet

A product team can collect hundreds of notes and still lose the reason a decision changed.

The interview quote sits in one document. The usability observation sits on a board. The roadmap records the conclusion. Weeks later, someone asks why the team changed direction, and the answer depends on whoever remembers the meeting.

The problem is not a shortage of notes. It is a broken chain between what happened, what the team thought it meant, and which belief changed because of it.

My thesis is testable: a traceable record that separates observation from interpretation and links both to an affected assumption should produce more explainable product decisions and fewer evidence-free reversals than a flat repository of notes. Measure source-linked decision changes, correction rate, time to explain a change, and reversals made without new evidence. If a simpler note works as well, use the simpler note.

Observation and interpretation are different product objects

The distinction sounds basic until a research session starts.

"The participant paused for eleven seconds after the fee appeared and returned to the cart" is an observation. "The participant thought the fee was unfair" is an interpretation unless they said so. The first can be checked against a recording. The second may be useful, but it carries the team's reasoning.

The GOV.UK guidance for research notes tells observers to record what they see or hear rather than personal interpretations. Its analysis guidance then treats observations, findings, and actions as separate stages.

That separation is not administrative purity. It creates a place for disagreement. A designer can accept the observation and challenge the interpretation. A researcher can add another session. A PM can change the affected assumption without editing the source.

When a tool stores all three as one note, the conclusion inherits the authority of the observation.

An insight needs somewhere to land

Research synthesis often ends with a polished finding. The next weak point is the handoff from finding to product judgment.

Consider the observation about the fee. The team might interpret it as weak price transparency. Which belief does that affect?

Perhaps the team assumed customers understand the difference between delivery tiers before reaching the summary. Perhaps it assumed the fee amount, not its timing, is the main source of abandonment. Those assumptions lead to different next tests and different product changes.

The GOV.UK alpha guidance recommends doing the minimum needed to test the riskiest assumptions. That becomes difficult when evidence lives in one place and assumptions live in another, or only in someone's head.

An insight becomes operational when a team can point to the belief it changes.

The Signal Trace

I use a five-part artifact to preserve the path from a research moment to a product decision:

The final choice matters. Product evidence is not only positive or negative. A signal can complicate an assumption by exposing a segment, context, dependency, or contradiction. "Complicates" protects the team from forcing ambiguous evidence into a score.

The artifact should stay lightweight. It is not a demand that every support ticket become a research report. A signal earns this treatment when it could change scope, priority, a success measure, or the next question the team asks.

I built Living Fieldbook to test the interaction

I wanted to know whether the Signal Trace could feel like a practice worth returning to instead of another form to complete.

I built Living Fieldbook as a mobile-first prototype. A short mission asks the user to capture a source and observation, write an interpretation, name the affected assumption, and choose how the signal connects. The completed trace moves into a fieldbook. A living habitat changes as evidence accumulates.

The habitat is a metaphor, not a score. It does not reward certainty or punish contradictory evidence. Its job is to show that product understanding grows through connected signals, including the signals that make a belief less comfortable.

I tested the complete interaction at 390 by 844 and 320 by 568. The audit changed the dropdown, disclosure layout, glass depth, bottom navigation, scrolling, and content density. The current prototype passes typecheck, lint, and a production build. It remains intentionally narrow. There is no account system, shared repository, or production research store yet.

That boundary matters because the purpose of this version is to test the reasoning loop, not to pretend the wider evidence-management problem is solved.

The workflow creates a privacy obligation

Research notes can contain personal and confidential information. A production fieldbook would need informed consent, access controls, retention rules, redaction, deletion, and a clear statement of purpose.

The same GOV.UK note-taking guidance requires informed consent and says notes should only be used for the agreed purpose. That constraint changes the product. Search, sharing, AI synthesis, and long-term memory cannot be added as harmless conveniences. Each changes who can access evidence and how far it travels from the original research context.

The product should also keep participant language distinct from generated or team-authored interpretation. An AI assistant might help cluster signals or surface related assumptions. It should never rewrite an inference as an observation.

Evidence should change a decision

The GOV.UK agile governance principles say decisions should be evidence-based and focused on user needs. A repository does not create that outcome by existing.

I would evaluate the Signal Trace against ordinary notes:

• Can another teammate identify the source of a decision change?
• Can they distinguish participant behavior from team interpretation?
• Can they explain which assumption moved and why?
• Do contradictory signals produce a clearer next question?
• How often does a roadmap reversal happen without new evidence?
• How much extra capture time does the structure require?

The last question keeps the proposal honest. Traceability has a cost. If the artifact is so heavy that people stop capturing evidence, it has failed.

The product principle is smaller than a research repository and more demanding than a notebook: preserve the chain.

A product note is not evidence merely because it was written down. It becomes decision evidence when the team can see what happened, what they think it means, which belief it affects, and what they should examine next.