Signals

What is a signal

A signal is any raw observation that might be relevant to a product decision. The key word is raw — don't filter at capture time. If it caught your attention, it belongs in the system. You'll decide what it means later, when you have enough signals to see a pattern.

A signal can be any of these:

• A verbatim quote from a user interview ("I always forget where I left off")
• A data anomaly from your analytics ("DAU dropped 18% the week after the last release")
• A teammate's observation during a demo ("everyone asked about exporting")
• A support ticket pattern ("five tickets this week about the same edge case")
• A competitor move you noticed in the wild
• A metric you're tracking that just crossed a threshold worth investigating

Signals are deliberately unstructured. They don't have to be actionable yet. That's what hypotheses are for. The goal at capture time is fidelity — get the observation in as close to its original form as possible.

“A signal isn't a decision and it isn't a hypothesis. It's evidence. Collect enough of it and the hypotheses write themselves.”

Capturing signals

There are three ways to get signals into Findry:

1. Quick note — press N anywhere in the app or click + Signal in the sidebar. Fill in the content, source, and tags. The whole flow takes under 30 seconds. This is the mode to use mid-interview, mid-meeting, or whenever something surfaces and you need to catch it before it slips.
2. Document paste — in the signal creation wizard, paste a full transcript, interview note, or research document. Use the built-in quote extraction to highlight and pull individual quotes as separate signals — each one becomes a child signal linked to the parent document. Useful after a batch of user research when you want to process everything at once.
3. From a test conclusion — when you close a test, Findry optionally generates signals from the conclusions you've written. These arrive pre-linked to the hypothesis the test was validating, which means the evidence chain is maintained automatically.

Signal metadata

Each signal has four fields. Only content is required; the rest are worth filling in when you have a moment:

• Content — the raw text. Can be a quote, a note, a number with context. Keep it close to the original phrasing. Don't summarize at this stage.
• Source — where it came from. Free text — "User interview, May 2, Ana K." or "PostHog: weekly_active_users" or "Slack thread #product-feedback". The source tells you how much weight to give the signal later.
• Tags — optional. Use these to group signals by theme, sprint, or source type. Tags are workspace-scoped and autocompleted from your existing set. Good tags make the filtering experience much more useful once your signal library grows.
• Confidence — how certain you are about this signal. Low / Medium / High. Defaults to Medium. A verbatim quote from a paying customer is High. A hunch from a Slack thread is Low. Confidence flows into the evidence strength calculation on any hypothesis you link this signal to.

Linking signals to hypotheses

Signals become evidence when linked to a hypothesis. A signal on its own is just an observation — the link is what makes it part of the reasoning chain behind a decision.

Open any signal and click Link to hypothesis — it searches your hypothesis list by title and belief statement. You can also pull signals in while creating a hypothesis by searching your signal library from the hypothesis form directly.

The link is bidirectional: the signal shows which hypotheses it supports; the hypothesis shows its full evidence base. Evidence strength updates automatically as you add or remove links — you don't need to do anything extra.

Searching and filtering

The Signals list supports full-text search across content and source fields, plus tag filters, source filters, and confidence filters. These compose — you can combine them to get very precise slices of your evidence library.

A few filters worth knowing about:

• Use the tag filter to surface all signals from a specific research round — tag everything from a sprint's user interviews with the same label, then filter by it when you're ready to write hypotheses.
• Use confidence + tag together to find "all High-confidence signals tagged onboarding" before writing a hypothesis about onboarding friction.
• Filter by unlinked to see your backlog of raw evidence that hasn't been connected to a belief yet. This is your most valuable queue — it's where hypotheses are waiting to be written.