A worked example · 1 September 2026
33 companies matched. I killed 24 of them.
I have no clients yet, so I can't show you a client's pipeline. What I can show you is mine. This is one real batch from the targeting layer I'd build for you, including every company it threw away and why.
The kill rate is the product. Anyone can hand you 33 companies. The work is knowing which 24 would have wasted your week, and being able to say so in writing.
- Matched
- 33
- Killed
- 24
- Emailed
- 2
01 What I was looking for
The signal, written down before the search.
A signal is an observable fact that means a company needs you this week. It has to be checkable by someone else, or it's just a hunch with a spreadsheet around it. Here is the one I ran, exactly as it was defined before a single row was pulled.
- B2B SaaS — any vertical, Europe or the US
- 20 to 150 people — big enough to buy, small enough to decide
- Seed to Series B — funded, still founder-led
- A live sales-role job ad — the intent to sell, made public and dated
- And no sales leadership already in place — or a sales org under 90 days old
02 The kill log
24 kills, each with a reason I can defend.
This is the artefact most vendors never show you, because for most vendors it doesn't exist. A list arrives with the survivors and no record of the judgement. Every row below was removed on purpose.
| Killed because | What it means |
|---|---|
| 11 The job ad was too old to still be true | Roles posted three to five months earlier. A role posted in March is probably filled, which makes "I saw you're hiring" false — and checkable by the person reading it. The worst class of error in a cold email is the one the recipient can disprove in one click. |
| 8 They already had sales leadership | A VP Sales, Head of Sales, CRO or CSO already in seat, in one case three of them. These companies are not going from founder-led sales to a first motion. They already went. The premise of the email would have been wrong. |
| 4 My own check had missed the sellers | These four passed the first leadership check and died on the second. The first query searched only leadership titles — VP Sales, Head of Sales, CRO. Re-run against every sales-carrying title, it found existing teams: one had a five-person commercial org, another four account executives. The company was never the problem. My query was. |
| 1 Too mature to be founder-led | Founded 2011. Fifteen-year-old companies do not run founder-led sales, whatever the headcount band says. A second company, founded 2009, also failed this test — it is counted once, in the stale-ad row above. A company gets the reason it died of first, not every reason it could have died of. |
Every one of the 33, and what happened to it.
Twenty-four killed and two emailed leaves seven companies, and a page arguing that the discarded rows are the product has no business hiding them. So here is the whole batch, including the rows that make me look worse.
| What happened to it | Companies |
|---|---|
| Killed — the job ad was too old to still be true | 11 |
| Killed — they already had sales leadership | 8 |
| Killed — my own check had missed the sellers | 4 |
| Killed — too mature to be founder-led | 1 |
| No reachable contact — no address for any founder | 1 |
| Never resolved — parked as borderline, never revisited | 3 |
| Never resolved — fell out of the workflow unchecked | 2 |
| Qualified, then held — I could not confirm the country | 1 |
| Qualified and emailed | 2 |
| Total | 33 |
03 Four traps a bought list walks straight into
The failures are silent. That's what makes them expensive.
None of these show up as an error. Every one of them produces a full-looking column and a confident, wrong email. All four came out of this single batch.
-
Trap 01 · Job title inflation
"Founding" in a job title is marketing, not an org chart.
Two companies in this batch advertised a Founding SDR and a Founding Account Executive. Both already had a Head of Sales in place.
The word describes how the role is being sold to candidates, not the state of the company. Any rule that treats "Founding AE" as a first-sales-hire signal will generate false positives at scale, and each one is an email whose opening line is provably wrong. Only the separate leadership check catches it.
-
Trap 02 · Entity resolution
An acquired company resolves to its parent's job board.
Look up a company's open roles through its applicant tracking system and you may be handed the roles of whoever bought them. The data looks perfect. The company appears to be hiring eleven salespeople.
The tell is the volume: a 40-person company does not have ten open sales roles. When the count looks too good, the row is usually pointing at a parent company, and the email you're about to send describes someone else's business.
-
Trap 03 · The name trap
Two founders, similar names, different companies.
An enrichment step returned a founder name for one company that actually belongs to the CEO of a different one. Nothing errored. The column was populated, the name was real, and the person was real — just not there.
An email opening with the wrong person's name at the right company is not a typo. It tells the reader precisely how the list was built, and it is unrecoverable.
-
Trap 04 · Verification before send
The obvious address hard-bounced in 20 seconds.
The pattern-guessed address at the obvious domain came back
550 NoSuchUseralmost immediately. The same person at the company's actual sending domain was valid. One character of difference, caught before the send rather than after.Bounces are not just lost emails. They are the input to your sender reputation, which is the thing that decides whether the next hundred land in an inbox at all. This is why the guarantee on bounce rate is the part I'll stand behind without conditions.
04 What this is and isn't
What this page does not prove.
It proves the targeting layer works and that the judgement is written down. It does not prove revenue, because there isn't any yet — three emails went out of this batch and I am not going to dress up a reply rate from a sample that small.
Company names, people and domains are deliberately absent. These are real businesses that did not ask to appear in my marketing, and a service that sells research rigour does not leak its research. If you want to walk through the unredacted tables live, ask and I'll open them on a call.
05 Your turn
Want this run on your ICP?
Tell me what you sell and who buys it. Within two working days you get back the three signals I'd target first and why, in writing. No call, no obligation.