A real one, with the names taken out. Including the twenty-five companies we removed and the reason written against each.
This page is a real Kickoff target list with the identities removed. Every number below is taken from the live working file — 119 retained screened companies across 10 countries and 9 outreach languages, built for a Turkish manufacturer selling agricultural inputs into European markets.
Retained screened is a precise term here and worth stating before the numbers. It means a row survived the removal passes set out below and carries a source, a verification date and a confidence grade. It does not mean the company has been qualified as a buyer. A third of this list is graded low confidence, and that grade is published rather than smoothed.
It is here because the aggregate claims on the rest of this site — companies screened, buyers contacted, markets covered — prove activity and prove nothing about judgement. What follows is the judgement: the columns each row has to fill, the evidence standard behind them, the 25 companies that were removed and the reason recorded against each, and what actually happened when the outreach went out, including the parts that failed.
Thirty columns per company, in five groups. The source-and-evidence group is the one that gates entry: a row that cannot carry a source and a verification date does not go on the list. There is one stated exception on this list, and it is the reason the count below reads 118 of 119 rather than 119 of 119 — a company retained on a named contact and a verification date without a source URL, recorded as an exception rather than quietly filled in.
| Group | Fields |
|---|---|
| Identity | Internal ID · company · website · HQ country, region, province, city |
| Segmentation | Segment · primary product target · end-market sub-segment · buyer persona class · the stated reason for that class |
| Contact | Named contact · title · email · LinkedIn · phone |
| Source & evidence | Source URL · last-verified date · confidence · ICP fit score · tier |
| Execution | Outreach language · priority · send state · wave · date sent · reply state · send history · notes |
| Measure | Count | What it means |
|---|---|---|
| Retained screened rows | 119 | across 10 countries, 9 outreach languages |
| Rows carrying a source URL | 118 of 119 | the one exception is logged as an exception, not left blank |
| Rows carrying a last-verified date | 119 of 119 | nothing undated |
| Named decision-maker identified | 88 | not a role inbox — a person |
| Verified email | 84 | bounce-tested before sending |
| Tiering | T1 38 · T2 60 · T3 7 | priority, not alphabetical order |
| ICP fit score | 105 scored · range 50–100 · mean 78 | a number, applied consistently |
| Buyer persona class | 74 direct buyer · 31 hybrid | whether they buy, specify, or neither |
Confidence is recorded per row and it is not flattering: 50 medium, 34 low, 21 high. A third of the list is explicitly marked as weak evidence rather than quietly presented as solid. Several rows carry a written qualifier instead of a grade — “confirmed capability, crop mismatch”, “capability-check only”, “best crop fit, pelleting not yet confirmed”.
This is the part that is usually deleted before a client sees the file. It is kept, with a tag and a written reason against every row, because a list you cannot audit is a list you cannot trust.
| Removal reason | Rows |
|---|---|
| Off-filter — multinational, outside the buyer definition | 9 |
| Off-filter drops on later verification passes | 6 |
| Buyer-filter fail — discovery row, not a buyer under the filter | 3 |
| Off-filter — nursery, wrong end-market | 2 |
| Off-filter — trade association, not a buyer | 1 |
| Off-filter — local subsidiary of a company already excluded | 1 |
| Merged — found to be an organisational unit of another listed company | 1 |
| Sector adjudication drop | 1 |
| Fabricated-entity risk — multi-source convergence check failed | 1 |
| Total | 25 |
That last row is the one worth pausing on. One company on this list, carried into an early pass on the strength of AI-assisted discovery, failed a three-source convergence check and was removed before anyone was contacted. That is one instance and it is what we can evidence — we do not publish a fabrication rate, because we have not measured one. What the instance does establish is that the convergence check is a standing step rather than a lucky read, and it is the plainest argument for why a target list needs a source URL against every line.
And the largest block of removals on that table is a failure of ours, recorded anyway. Thirteen of the twenty-five were companies we had already emailed. Two April batches went out to multinational corporate inboxes, a trade association, a local subsidiary of a company already excluded, and two nurseries in the wrong end-market — all before the off-filter rule was written down. The rule exists because that happened, and the removals sheet is the only reason you can see it.
Outreach ran in waves, in the buyer's language — English 30, Turkish 26, Italian 13, Romanian 12, Hungarian 11, Greek 10, Czech 7, Polish 7, Slovak 2. Every send is logged with its date and its outcome, including the outcomes nobody publishes:
The file is versioned, with a changelog recording each supersession and what it changed — including two versions issued the same day when an external review pass overturned the first.
Removed: every company name, website, contact name, job title, email address and phone number, on both the live list and the removals sheet. Kept: the column architecture, the counts, the distributions, the removal tags and the reasoning behind them, and the send and reply outcomes.
The client is not named. Confidentiality is the default on everything client-related, and a reference for this engagement is available under NDA rather than published here. What is on this page is the method, which is ours to show.
One disclosure, because it changes how this page should be read. This working file comes from an operating mandate with a company where Kickoff's founder holds a disclosed directorship. It is published to show the work product, not as an independent client endorsement.
Thirty fields per company across identity, segmentation, contact, source-and-evidence, and execution. The load-bearing group is source and evidence: a source URL, a last-verified date, a confidence grade, an ICP fit score and a tier on every row. In the list shown here, 118 of 119 rows carry a source URL and all 119 carry a verification date.
Every row needs an independently opened source before it enters the list, and convergence across sources is checked before any contact is made. On this list that check removed one company for fabricated-entity risk before anyone was approached. That is one instance, which is what we can evidence; we do not publish a rate. It is also why every row carries a source URL rather than a note saying where it probably came from.
Yes, with the reason recorded against each. The list shown here retains 25 removed companies across 13 distinct removal tags, including thirteen recording our own process failure. A shortlist without its rejections cannot be audited, and an unauditable shortlist is just an assertion.
Two cold touches, then hold. On this list, eight companies sit at “no reply after two touches — HOLD” with the reason written into the row: a third cold approach is a throttle risk that damages the sender domain and the client's name. Stopping is a logged decision, not a lapse.
Aggregate numbers prove activity. The removals sheet is where you can see whether the judgement behind them is any good — which is why it is on this page and not in a folder.
Talk about the work >