Ask a revenue team what a good-fit company looks like and you will get a fluent answer in under a minute. Ask what a bad-fit company looks like and you will get a pause, then something about budget. The asymmetry is the whole problem: the first list is written down and the second lives in the heads of whoever has been there longest.
Anti-signals are that second list, made explicit and made observable. They are not a mood or a gut feeling. They are facts you can find from outside, before anyone picks up a phone, that predict a bad outcome.
Why they do more work than the positive filter
A firmographic boundary — size, sector, region, stage — usually leaves thousands of eligible companies. It orders almost nothing. Anti-signals cut, and they cut precisely where the expensive mistakes are, because they are derived from deals that actually went wrong rather than from a description of the deals you would like.
The positive profile tells you who could buy. The anti-signals tell you who will waste your quarter.
Where they come from
Not from a workshop. From your own closed-lost and churned records, read one at a time with a single question in mind: was there anything visible from outside, before first contact, that predicted this?
Most answers will fail that test, and they should. “The champion left” is real but was not visible in advance. “They had rebuilt the same thing internally eighteen months earlier” often was — in a conference talk, an engineering blog post, a job title that should not exist at a company that buys this. That one becomes an anti-signal. The first does not.
The ones that show up repeatedly
These are patterns worth checking against your own history rather than rules to adopt. Some will be wrong for your market — the exercise is to find out which.
| Anti-signal | Where it is visible | What it predicts |
|---|---|---|
| Built the same capability in-house | Engineering blog, conference talks, a job title that implies it | A long evaluation ending in 'we already have this'. |
| Hiring freeze or a visible layoff round | Postings withdrawn, public announcements, headcount pages | Budget exists on paper and cannot be spent this half. |
| Signed a competitor within the last year | Case studies, integration pages, public logos | You are being used to price a renewal. |
| No owner for the function you sell into | Team pages, absence of the role in any posting | No internal sponsor, so nothing progresses past interest. |
| Buys exclusively through a procurement framework you are not on | Public tender records, supplier pages | A twelve-month cycle you cannot shorten and probably cannot win. |
| Deep dependency on a stack you cannot integrate with | Job postings, status pages, public architecture posts | A technical blocker discovered in month two. |
Hard and soft
Not every anti-signal deserves to remove a company entirely. Two behaviours are worth separating.
Hard anti-signals remove the company from consideration. They are stable facts: a direct competitor, an excluded jurisdiction, an existing legal dispute. Nothing about them will change next quarter, and surfacing the company anyway just re-runs the same decision.
Soft anti-signals cap the tier. The company can still appear — useful when the list is genuinely thin — but never at the top, and always with the reason attached so the rep knows what they are walking into. A hiring freeze belongs here. So does a recent competitor signing, because renewals do come round.
Making them survive contact with the list
An anti-signal that lives in a document nobody opens is worth nothing. Three things make them stick.
- Each one is written as an observable fact with a place you would find it, not as an adjective.
- Each one carries the loss it came from, so nobody quietly deletes it in a tidy-up.
- Hard and soft are labelled separately, and soft ones carry an expiry.
- They are applied before the list is scored, not as a manual review afterwards.
- A rejected company keeps its rejection reason, so the decision can be argued with rather than repeated.
That last point matters more than it sounds. A filter that silently drops companies is impossible to debug, and the first time someone senior finds a good account missing, the whole list loses credibility. Keeping the rejections visible is what makes the filter arguable — and a filter nobody can argue with is a filter nobody trusts.
This is the half of the profile most teams skip, which is why it is the fourth part of the ICP structure rather than an afterthought. In Leads, anti-signals run before scoring and the rejected companies come back with their reasons attached — the list you did not get is as reviewable as the one you did.