The Class Everyone Rents

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

Plate IV. The class everyone rents.

Here is what took me too long to admit. For about two years I treated a job posting as a buying signal. A company opened a RevOps role, the alert fired, and a note went out that week. It felt like fieldcraft. Then a prospect forwarded me the thread, the ordinary way, and three of our closest competitors had sent nearly the same note in the same week. Same posting. Same alert. Four notes in one inbox, all congratulating a stranger on a hire she had not made yet.

The posting was real. The team was growing. What I had misread was whose signal it was. It was not ours. It was the company's weather, published to anyone with a subscription, and all four of us had one.

Plate IV: the exhaust 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. Part 3 was the class you filed: what your own pipeline history says is about to happen. Plate IV is what the company did in public.

Same three parts. Surface, behavior, window. The surface is the world outside your stack: job boards, funding databases, DNS and script telemetry, the press. The behavior is what the company did where anyone could see it. The window is how long ago, and how long it stays true.

hiring → RevOps role (30d). The posting. A company opened a role you sell to inside the last month. Somebody has budget and a problem large enough to hire for.

funding → Series B (60d). The round. Money arrived in the last two months and a plan arrived with it. Everyone knows the same day, because it was announced so that they would.

tech → competitor install (90d). The install. A script showed up on their site that says they picked somebody, or are trying somebody. It tells you the category is live at that account. It does not tell you the deal is open.

leadership → new CRO (45d). The hire. A new revenue leader has forty-five days of wanting to change something. Also forty-five days of being sold to by everyone who noticed.

And the negative specimen, which matters more on this plate than on any other. headcount → decrease (90d). Layoffs. Closed offices. This is not a buying signal read backwards. It is a suppression signal: the account is telling you, in public, that this is not the quarter. PredictLeads' own category guide files decreases_headcount_by and closes_offices_in under cost cutting, apart from the expansion events, and says the two groups need handling differently. They do. One shortens your window. The other closes it.

Pick an exhaust signal. The panel names it by its parts, says who else can read it at the same minute, what it actually tells you, and the fidelity read.

Specimen, by its parts	Who else reads it at the same minute	What it actually tells you	Fidelity
hiring → RevOps role (30d)	You, every competitor with the same subscription, and any agent with the vendor's MCP tool.	A team is growing and has budget for a role you sell to. Not that anyone there is looking at you.	Low. Company weather.
funding → Series B (60d)	Everyone. The round was announced so that everyone would know.	Money arrived and a plan arrived with it. Not where the money goes.	Low. Company weather.
tech → competitor install (90d)	Anyone running script or DNS detection on the same domain, which is every vendor in the category.	The category is live at that account. Not that the deal is open.	Low to medium. Evidence of use, not of a decision.
leadership → new CRO (45d)	Everyone who follows the announcement, and every agent that reads it.	A new leader wants to change something. Also, is being sold to by everyone who noticed.	Low. A person changed, not a buyer.
headcount → decrease (90d)	Everyone. Layoffs are public by nature.	This is not the quarter. A suppression signal, not a buying signal read backwards.	High, for what it is: a reason to hold.

Qualitative. Nobody's counts are used or implied. Windows are choices; pick your own.

Figure 1. Five exhaust specimens, each read by its parts, each with the honest answer to who else can read it at the same minute.

What the weather tells you

Exhaust is about the company, not the buyer. That is the whole fidelity argument and it fits in a sentence: nobody on the team typed anything.

A product signal is a person doing something with their hands in your product. A site signal is a person on your pricing page. A filed record is a person who once told you no. Exhaust is a press release. A posting says a team is growing. It does not say anyone there is looking at you, or at your category, or at anything except the hire they need to make. A round says money arrived. It does not say where it is going.

So plate IV is the lowest-fidelity class in the legend, and it is that way by design, not by failure. It is population weather. Weather is useful. You dress for it. You do not propose to it.

Everyone's weather

Then there is the ownership axis, and here plate IV sits alone at the far end.

Part 2 said the class you rent is rented from one vendor at a time, on your own domain. Plate IV is rented by everyone at once. The same posting, the same round, the same install, sold to you and to everyone who sells against you, from the same handful of providers, refreshed on the same schedule. There is no version of the feed that is yours.

This summer made that literal. PredictLeads exposes its job openings, financing events, technology detections, and news events through one MCP server at mcp.predictleads.com. Coresignal shipped an MCP server with OAuth in August, so an agent connects to its company, employee, and job data in one click and no API key. Demandbase lists native MCP support so an assistant pulls its enrichment and intent data by prompt. Clay's MCP server puts its providers and your team's prebuilt workflows inside ChatGPT, Claude, and Codex. None of this is a complaint. These are good tools. We have bought more than one of them, and I would again. Monday's piece covered the general case: your vendors ship MCP servers now.

The specific case is the same-minute problem. When the signal is public and the tooling is universal, your competitor's agent reads the posting in the same minute yours does. Speed of read is zero edge. I argued in the edge is the minute after that the edge lives after the signal, in the doing. That still holds. Plate IV is where you find out it was never in the seeing.

Everyone's weather is nobody's edge.

What makes it yours

The join. It is the only thing on this plate that is yours, and it is enough.

Take the posting alone. hiring → RevOps role (30d). Everyone has it, everyone sends the note, and the note is about the hire because the hire is all anyone knows. Somebody who has never heard of you is being congratulated by four strangers.

Now take the posting and a return to your pricing page. hiring → RevOps role (30d) + pricing → repeat (7d). The second half is plate II, on your domain. Nobody else's agent can see that visit. The company is hiring for the role you sell to and somebody there is reading your price this week. That is not weather. That is a specimen, and it is yours because half of it happened on your ground.

Or the posting and a record from your cabinet. hiring → RevOps role (30d) + closed-lost (14mo). Plate III. An account that told you no last year is staffing the function you sell to. Every vendor knows the posting. Only you know the no, and why.

Part 3's rule was never read the cabinet alone. Generalize it: never read the weather alone. Exhaust is a modifier. It moves the timing axis (the posting makes this the right month) and it writes the suppression list (the layoff makes it the wrong one). It should almost never fire a Play by itself, because a Play fired on exhaust alone is a note everyone sent.

The exhaust signal is fixed. Choose what it is joined to. The panel says what fires, who else could have fired it, and the record line Bryn would write.

• Alone: hiring → RevOps role (30d). Fires a congratulations note about the hire. Every competitor with the feed could have fired it, and did. The same note your competitors sent.
• + plate II: hiring → RevOps role (30d) + pricing → repeat (7d). Fires a note about the role and the pages read. Only you could have fired it; the visit is on your domain.
• + plate III: hiring → RevOps role (30d) + closed-lost (14mo). Fires a reopen note that names the original loss reason. Only you could have fired it; the no is in your cabinet.
• + plate I: hiring → RevOps role (30d) + invite → second seat (48h). Fires a note to the owner about the trial account adding seats while hiring. Only you could have fired it; the seats are in your product.

Qualitative and illustrative, not a benchmark. Anweledig Labs is a fictional account.

Figure 2. One posting, three joins. What fires, who else could have fired it, and the line Bryn writes (illustrative, not a benchmark).

Reading well

Three rules for plate IV.

Exhaust modifies; it does not fire. Put it on the timing axis and in the conditions. A posting shortens the window on a Play you already run. A round raises the ceiling on one. Neither is the trigger. The trigger comes from a class you own or filed, where a person did something.

The inverse is the half everyone skips. Every team I know has a hiring alert. Almost none has a layoff hold. headcount → decrease (90d) should pause every outbound Play on the account, automatically, because the day it matters is the day nobody is checking. The suppression list is the most owned thing you can build on a rented feed. Nobody else's agent is deciding not to send.

Age the signal on purpose. A posting is a 30-day fact. A round is a 60-day fact. An install is a 90-day fact. None of those windows is right, and any of them beats a feed that treats a two-year-old detection like this morning's. Use the vendors' timestamps; every record carries first seen and last seen. The window is the one part of an exhaust specimen you get to write.

Chris wrote yesterday about planning 2027 around the hour with no owner. A suppression rule is the smallest version of that: a decision that fires when nobody is in the room. It belongs on Approve until you trust it, which Tuesday's piece, approve mode is not a speed bump, argued better than I will here.

The legend, complete

Four plates, two axes. Fidelity on one. Ownership on the other.

Plate I, product behavior: high fidelity, yours outright. Plate II, anonymous site intent: medium fidelity, rented from one vendor at a time, on your own domain. Plate III, CRM lifecycle: medium-high and specific about the wrong thing, yours on paper. Plate IV, ecosystem exhaust: low fidelity, rented by everyone.

Four plates on two axes. Pick a plate for its one-line read, its part in the guide, and its rule.

Plate	Fidelity	Ownership	Rule	Part
I. Product behavior	High	Yours outright	Own it. Name one specimen by its parts and pre-approve one action.	The Class You Own
II. Anonymous site intent	Medium	Rented from one vendor, on your domain	Watch it. The visit is yours; the name is rented. Match it to something you own.	The Class You Rent
III. CRM lifecycle	Medium-high, about the wrong thing	Yours on paper	Open it. Never read the cabinet alone.	The Class You Filed
IV. Ecosystem exhaust	Low	Rented by everyone	Let it set the weather. Never read the weather alone.	The Class Everyone Rents (this part)

Positions are a reading, not a measurement. Illustrative, not a benchmark.

Figure 4. The legend. Four plates on two axes, each with its rule and its part.

Own the class you own. Watch the class you rent. Open the class you filed. Let the class everyone rents set the weather. That is the whole guide, and I notice that most of it is instructions about which plate not to fire from.

The year we tried to buy our way out of the signal gap, most of what we bought was plate IV. More weather, better forecast, same four notes. Intelligence became an ingredient a while ago and the vendors have now finished the job by making it callable. That leaves the join, which nobody can sell you because it runs through your product, your site, and your cabinet.

The gift

Do this Monday. It is two rules on a feed you already pay for.

Pick one exhaust signal you are already buying: a posting in a role you sell to, a funding round, a tech install. Write it by its parts. Then write the join. hiring → RevOps role (30d) + pricing → repeat (7d) fires. hiring → RevOps role (30d) alone does not. Write both lines down so the second one is a rule and not a mood.

Then write one suppression rule from the inverse. headcount → decrease (90d) holds every outbound Play on that account until the window closes.

Two rules, both yours, on a feed everyone has.

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 product, the site, and the CRM, the three classes where somebody did something, and it takes exhaust as an enrichment input: a modifier on the timing axis and an entry on the suppression list, never a trigger by itself. A posting shortens the window on a Play you approved. A layoff holds it. Every run logs the join, so the record shows the weather Bryn read, where it read it, and the owned signal it was joined to. acct_4471 · hiring → RevOps role (12d) + pricing → repeat (7d) · ran intro-note-v2 to owner. One line, replayable, and the first half of it was public.

If your own agents would rather ask than read, they can. The agent your agents call covers that, and the answer comes back with the same parts, join included.

That is the field guide. Four plates, one legend.

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