Company matching is the least glamorous part of account-based advertising and the one most likely to be quietly destroying a programme. It is also invisible: nothing in day-to-day campaign reporting distinguishes "this account is not engaging" from "this account is not being targeted at all".
Why match rates decay
A list uploaded once and never revisited degrades for entirely mundane reasons.
- Companies get acquired. The acquired entity's page is merged or archived, and everything you were targeting through it stops resolving.
- Companies rename or rebrand. Common in the two years after a funding round, and a rename usually comes with a new domain.
- Your list was built from trading names. Sales calls it "Ravensbourne". The registered entity is "Ravensbourne Logistics Holdings Ltd" and the page is under a third variant.
- Subsidiaries and parents get confused. You target the parent; the buying committee sits in a subsidiary with its own page, or vice versa.
- Regional entities are separate pages. The DACH arm of a US company is often a distinct entity, and targeting the parent will not reach the people who would actually buy from you.
None of these produce an error. They produce a slightly smaller audience, which produces slightly cheaper delivery to the accounts that did match, which produces a report that looks fine.
What a bad match report looks like
Rough guidance for a list of mid-market and enterprise companies:
| Match rate | Interpretation |
|---|---|
| 90%+ | Healthy. Usually means the list was built from domains. |
| 75–90% | Normal for a name-based list. The gap is worth working through. |
| 60–75% | Something systematic is wrong — often a region or a segment failing as a group. |
| Below 60% | Do not launch. You are about to spend a quarter reaching half your strategy. |
Small companies and certain regions match worse for legitimate reasons, so read the rate by segment rather than in aggregate. A 78% overall rate that is 94% in North America and 51% in DACH is not a 78% problem; it is a DACH problem, and averaging it away is how it survives.
Fixing the four common failures
1. Use domains, always
Domain matching outperforms name matching by a wide margin. If your CRM holds website URLs, export those. Strip protocols and www, keep the registrable domain, and drop anything that resolves to a generic email provider — a row whose "domain" is a free webmail host is a data-entry error, not a company.
2. Decide the parent/subsidiary rule once, and write it down
There is no universally right answer. If you sell centrally, target the parent and roll subsidiary engagement up to it. If you sell into regional entities, target them separately and report them separately. What breaks programmes is doing both inconsistently, so that an account appears twice, gets double the delivery, and shows up in reporting as two half-covered accounts.
3. Add known aliases
Keep an alias column. When a match fails, the fix is usually one of: the registered name, the pre-rebrand name, or the parent. Recording the alias means the next upload does not fail the same way.
4. Re-upload on a schedule, not on a crisis
Monthly is enough for most programmes. Quarterly is acceptable for very stable lists. "When something looks wrong" is not a schedule, because the failure mode is specifically that nothing looks wrong.
A monthly routine
Twenty minutes, once a month:
- Export the current match report. Record the rate overall and by region and tier.
- Compare against last month. Any drop over about three points is a signal, not noise.
- Work the unmatched list. Most rows resolve with a registered name or a domain, and the fixes accumulate — the list gets more robust each month rather than decaying.
- Check for accounts that have gone to zero delivery while the campaign is still running. That pattern almost always means an acquisition or a page merge.
- Re-upload and note the new baseline.
It is dull. It is also the highest-return twenty minutes in the entire programme, because everything downstream — coverage, scoring, attribution — is computed against a list that has to be right first.