tl;dr There are four classes of GTM signal: product behavior, anonymous site intent, CRM lifecycle, and ecosystem exhaust. Only one of them is yours. Product signals are the highest-fidelity class, the only class nobody can rent out from under you, and the class almost nobody actually runs. This opens a four-part field guide. Part 1: learn to read the class you own, then run one Play on it.
A decent field guide does one job. It gets you from "I saw a bird" to "I saw a female goshawk, and here is what she is about to do." Identification, then behavior, then response.
Growth teams mostly do not have one. We have dashboards, which are closer to a museum: everything pinned, labeled, and dead.
Here is what took me too long to admit. I could name every intent vendor on the market and quote you their per-seat pricing, and I could not tell you the three strongest signals inside my own product. I was buying binoculars to watch other people's yards while the best specimens in the field sat on my own feeder.
So this is part one of a field guide. Four parts, one signal class each. We start with the class you own.
Every GTM signal I have ever worked with falls into one of four classes. Product behavior: what users do inside your product. Anonymous site intent: who is on your site and what they read. CRM lifecycle: what your own pipeline history says is about to happen. Ecosystem exhaust: hiring, funding, tech installs, community chatter, the stuff that happens off your surfaces.
Most taxonomies stop at fidelity: how strongly does this signal predict intent. That axis matters and it is the one everybody argues about. The axis everybody forgets is ownership. Who controls the instrument. Who can reprice it, degrade it, or sell it to your competitor on the same afternoon.
Plot the four classes on those two axes and the chart makes the argument for me. One class sits alone in the good corner.
THE LEGEND (interactive: the four signal classes plotted fidelity vs ownership; selecting a class opens its specimen card) Four classes, two axes. The vertical axis is fidelity (how strongly the class predicts intent), the horizontal axis is ownership (who controls the instrument). - Product behavior: highest fidelity, hands not cookies; owned outright; covered in Part 1 (this piece). - Anonymous site intent: medium fidelity, depends on match rates; rented from identity vendors; covered in Part 2. - CRM lifecycle: medium-high fidelity, your own history; yours on paper, enrichment rented; covered in Part 3. - Ecosystem exhaust: low-medium fidelity, everyone's weather; rented by definition; covered in Part 4. Positions are qualitative, not measured. The series map: each class gets its own part.A product signal is a user telling you what they want with their hands. Not their cookies. Their hands.
It is not one event. Single events are noise; a pageview is a pageview. A signal is a pattern with parts: a surface, a behavior, and a window. I write them the way Bryn logs them:
pricing → comparison → repeat (7d). Someone kept returning to your pricing page and your comparison page inside a week.
invite → stall → second-seat (48h). Someone sent a teammate invite, the teammate never activated, and the inviter came back looking within two days.
export → limit → retry (24h). Someone hit a plan limit, and instead of leaving, tried again the next day.
Read those back. Each one is a person mid-decision. The fidelity is not subtle. Nobody hits a plan limit twice by accident.
SPECIMEN CARD (static figure: one product signal rendered as a field-guide specimen card) Specimen card, Plate I, product signal. One specimen, fully identified. NAME, BY ITS PARTS: pricing then comparison then repeat (7d). HABITAT: your own product and site event stream. Owned ground. No identity vendor, no match rate, no per-seat rent. CALL: a buyer mid-decision. Kept returning to price and plan comparisons inside one week. High fidelity: hands, not cookies. RECOMMENDED RESPONSE (THE PLAY): if it fires on a fit account, account owner gets a note with the three pages read. Pre-approved once, run every time. The etching: a pattern, not an event. Three parts: surface, behavior, window (7 days). Example specimen (illustrative, not a benchmark). Pattern notation as Bryn logs it.If product signals are this good, why does almost nobody act on them? Because of where they live.
Product events land in analytics tools, and analytics tools are built to answer "what happened last quarter," not "what should happen in the next hour." They are retro surfaces. Brilliant for funnels, useless for the Tuesday morning when pricing → comparison → repeat (7d) fires on your best-fit account.
So the highest-fidelity class in the field gets aggregated into a chart, reviewed monthly, and never once triggers an action while the human behind it is still deciding. We wrote about the cost of that gap when we tried to buy our way out of it. The short version: we owned the best signals we had, and we were the last to act on them.
Every other class in the legend is rented, in whole or in part.
Anonymous site intent rides on identity vendors and their match rates. CRM lifecycle is yours on paper, but its enrichment layer usually is not. Ecosystem exhaust is a subscription by definition, and this week's news made the point for me: the biggest intent vendors are now piping their intelligence into any agent that can call an API. Rented instruments keep getting easier for everyone to rent, including the two competitors in your deal cycles. That is not a criticism of the vendors. It is just what happens to ingredients.
Chris made the adoption-gap version of this argument on Wednesday: the companies scaling agents are the ones who answered the ownership questions first.
WHAT RENTED MEANS, CLASS BY CLASS (interactive: per-class owned/rented toggle with a plain-sentence consequences readout) Each class starts where it actually sits. Toggle any class to read what changes. - Product behavior, as it sits: owned. What rented exposes you to: nothing. No repricing, no match-rate decay, no resale to competitors. - Anonymous site intent, as it sits: rented. What rented exposes you to: vendor repricing, match-rate decay, and the same data sold to your competitors. - CRM lifecycle, as it sits: partly rented. The records are yours; the enrichment layer can reprice or degrade. - Ecosystem exhaust, as it sits: rented. A subscription by definition, and increasingly callable from anyone's agent. Qualitative and illustrative, not a benchmark.Your product's signal exhaust is the one class that does not work this way. Nobody can reprice it. Nobody can degrade the match rate. Nobody can sell it to the company you lost a deal to last month. Everyone has signals. Only one class of them is yours alone.
Here is the gift, and it costs twenty minutes.
Name one product signal by its parts. Surface, behavior, window. Steal one of mine or build your own; the naming is the point, because a named pattern is checkable and a vibe is not.
NAME A SPECIMEN OF YOUR OWN (interactive: compose a product signal from surface, behavior, and window; the panel writes the pattern the way Bryn logs it plus a first-draft Play) Three worked examples: pricing then repeat (7d): account owner gets a note with the pages read. invite then stall (48h): inviter gets a one-line nudge with a working invite link. export then limit (24h): a plan-fit note goes to the account owner before the third attempt. Suggested Plays are illustrative, not a benchmark.Then write down the one action you would pre-approve if it fired. One sentence. "If pricing → comparison → repeat (7d) fires on an account over 50 seats, the account owner gets a note with the three pages they read." That pair, pattern plus pre-approved action, is a Play. You can run it manually this week. A human checking one pattern every morning is a perfectly good v0.
Bryn is what it looks like when that stops being manual. Bryn is not another dashboard to watch. It is the governed execution layer that runs Plays through your stack: it watches product, site, and CRM together, and when the pattern you named shows up, it runs the Play you approved and logs every step. The field guide is the same either way. Learn the specimen, decide the response, run it.
Next in this series: the class you rent. Anonymous site intent, identity vendors, match rates, and what renting well actually looks like.
Stop watching signals. Start running them.
Source: https://www.civic.com/field-notes/the-class-you-own