The Class You Filed

Part 3 of Signals: A Field Guide. Part 1: The Class You Own. Part 2: The Class You Rent.

Plate III. The class you filed.

Here is what took me too long to admit. Last year our CRM held a few hundred closed-lost accounts. I had never once asked which of them had come back to our site.

I knew the count. I could have given it to you in a pipeline review without looking, because the count was what the CRM was built to show me. I could not have told you which of those accounts had read our pricing page that month, because the record lived in one tool and the visit lived in another, and the question that joins them had not occurred to me as a question.

The CRM was the one class I owned outright, and the one I read least. I read the product data because it was interesting. I paid for the site data because it was new. The pipeline history I already had, and I treated it the way you treat a cabinet in a hallway. Things went in. The drawer stayed shut.

Plate III: the lifecycle signal

Part 1 was the class you own: what users do inside your product, with their hands. Part 2 was the class you rent: who is on your site, resolved by somebody else's graph. Plate III is what your own pipeline history says is about to happen.

Same three parts. Surface, behavior, window. The difference is that here the surface is a state. A stage, a status, a date on a contract. The behavior is what an account does next while it sits in that state. The window is how long the state has been true, or how long you have before it stops being.

closed-lost → pricing return (6mo+). The boomerang. An account that told you no more than six months ago is reading the price again. Something changed: the budget, the vendor they chose instead, the person who said no. You do not know which. You know it was not nothing.

eval → no activity (30d). The stall. An account sitting in an evaluation stage with thirty days of silence around it. Not a dead deal. A deal nobody is moving. The stage says evaluating. The calendar says nobody is.

renewal T-90 → seat drop (14d). The window. Ninety days out from a renewal, the account removes seats inside two weeks. The date was always on the calendar. The behavior is what makes it a specimen.

trial-expired → docs return (90d+). The second look. A trial that ended more than a quarter ago, and somebody from the account is back in your documentation. Nobody reads docs for fun.

And the negative specimen, because every plate needs one. closed-won → careers page. A customer reading your jobs page is a customer hiring, or a customer's employee leaving. It is not a buyer. It looks like intent on a dashboard because the account is a customer and the page resolved. Knowing the common ones is most of the skill.

The fidelity of this class is medium-high and specific about the wrong thing. It tells you what an account did with you. It does not tell you what the account wants now. That is the whole reason the class needs a join: a record from the cabinet plus a live signal from plate I or plate II. A filed record alone is a filing. A filed record that comes back to your pricing page is a specimen.

Why the cabinet stays shut

The CRM is a system of record. That is its job, and it does the job. It was built for forecasting and reporting, it is read in pipeline review, and it is optimized for what pipeline review wants, which is stage counts: how many in evaluation, how many closed this quarter, how many lost and to whom.

Stages are where deals go to be counted. Nothing in the cabinet is built to keep watching a deal after it has been counted.

Closed-lost is the clearest case. The word reads as an ending. The software treats it as a terminal state: the deal leaves the pipeline, the forecast drops it, the dashboard moves on. The account knows none of this. To the account it was a pause. Anweledig Labs picked somebody else, or nobody, and put you in a folder of their own. A year and a half later the somebody else is up for renewal and their folder opens. That moment lands on your pricing page, and the record that could have told you who they were is filed under a word that means over.

We made the point in the four handoffs that deals die in the gaps between stages. The cabinet is where the gaps are archived. Every closed-lost record is a handoff that did not happen, written down and never reread. And the year we tried to buy our way out of the signal gap, the CRM was the one tool in the stack we never thought to point the new eyes at. It was the reference. It was not the feed.

Yours on paper

Part 1 put the four classes on two axes, fidelity and ownership, and I said the CRM was yours on paper. Here is what the paper covers, layer by layer.

The records are yours. Stages, dates, loss reasons, contract values, the owner's notes. Nobody can reprice them or degrade them. Export everything tomorrow and it is all still true.

The enrichment is rented. Titles, headcount, funding, tech installs: the fields that make a record routable came from a vendor, landed in the record, and go stale on the vendor's schedule rather than yours. The title you are routing a renewal note to might be two jobs old.

The activity log is usually rented too, which is the layer I underestimated. The sequencer logs the send. The dialer logs the call. The CRM gets a summary if the integration is healthy and a gap if it is not. The record says contacted. What was said lives somewhere you pay for monthly.

The join is yours if you make it. A closed-lost record matched to a visit on your own site is a thing you own next to a thing you own. Nobody else has that pair.

What rented exposes here is specific. A renewal-window Play built on the enrichment layer alone keeps firing on stale titles, so the note goes to the champion who left in March. The same Play joined to renewal T-90 → seat drop (14d) fires on something the account is doing right now, whatever the title field says.

The math of the cabinet

Here is the arithmetic, and it is small numbers multiplied.

Take the closed-lost accounts from the last twelve months. Take the fraction that come back to a surface you can see inside the window. Take the fraction of those that still fit the ICP, because the ICP has moved since they were lost and some of them moved too. Then pick the window, because it is a choice and nobody else is going to make it.

Multiply and you get a small number. Smaller than the cabinet. Larger than zero. And every one of them is an account that already knows your name, already sat through the demo, already told you why not. The loss reason is in the record. The first line of the follow-up is already written.

Move the sliders and you will find what I found. The result is never big enough to build a quarter on. It is always big enough to be embarrassing that nobody was watching. And it goes stale the way every returned visit does: the account is on your pricing page for minutes, not weeks, which is the whole argument of the edge is the minute after.

The platforms are putting agents inside the CRM this week, and I am for it. Chris wrote on Tuesday about why the controls have to cross the tools once the pipeline runs across six of them. That is his argument and not this one. Mine is smaller. An agent that can read the cabinet is progress, because the cabinet was never the problem. Reading it was.

Reading well

Three rules, and they are the whole of part 3.

Treat closed-lost as a state, not an ending. A state can change. An ending cannot. The word your software chose does not decide which one it is. The account does. Build the watch list from the drawer you thought was closed.

Never read the cabinet alone. Join it to a live class before it fires anything. A record tells you what an account did with you. A site signal or a product signal tells you what it is doing now. Neither is a Play by itself. closed-lost is a filing. closed-lost → pricing return (6mo+) is a specimen. The arrow is the join.

Set the windows on purpose. T-90, thirty days, six months: none of those numbers is right, and any of them beats none. A window nobody chose is a window nobody owns, and an unowned window is how a renewal becomes a surprise. Follow-up has no owner is the ancestor of this rule. The thing with no owner does not get done, and a window is a thing.

The gift

Do this Monday. It costs half an hour.

Export closed-lost from the last twelve months. Sort by fit, take the top twenty, and put them on a watch list your identity or analytics tool can match against. That is the join, built by hand.

Write one specimen by its parts. Steal closed-lost → pricing return (6mo+) off the card. Pre-approve one action: "If it fires, the account owner gets a note with the original loss reason and the pages read, and nothing else happens until they answer."

Then find the renewal cohort due in Q1 and write down the T-90 date for each one. On a calendar, in a spreadsheet, on the wall. You are not building a Play yet. You are opening the cabinet.

How Bryn runs the class

Bryn is not another dashboard to watch. It is the governed execution layer that runs Plays through your stack. It watches the CRM alongside the site and the product, so a lifecycle state is one input to the score rather than a report to read later. The state feeds the timing axis and the Play's conditions: closed-lost more than six months, renewal inside ninety days, evaluation quiet for thirty. When a filed account comes back to a surface you can see, the Play you approved runs, and the record shows the state Bryn read, where it read it, and the live signal it was joined to. acct_2210 · closed-lost (8mo) → pricing return · ran reopen-note-v1 to owner. One line, replayable, with the loss reason attached.

If your own agents would rather ask than read, they can. Monday's piece, the agent your agents call, covers that. The answer comes back with the same parts.

The cabinet is not a bad place for a record to live. It is a bad place for a record to stay.

Next in this series: the class everyone rents. Ecosystem exhaust: hiring, funding, tech installs, everyone's weather, and what to do with a signal your competitor is reading at the same minute.

Part 3 of Signals: A Field Guide. Part 1: The Class You Own. Part 2: The Class You Rent. Part 4 next week.