Lookalikes work by copying a seed list. So we built the best seed we could think of: every person who actually closed on a house in Florida and Texas in the last year. Public county records, verified, $0.
Each row is one real person who bought a home, with the six columns Meta matches on: fn, ln, ct, st, zip, country. No emails or phones — we do not have them, and Meta does not need them.
These are not leads and not intent scores. Every row is a recorded, arm's-length purchase from a county appraisal roll.
Feed this to Meta and ask for more people like them. This is the whole point of the exercise.
Download seed-florida.csv 4.2 MBSame idea, Texas geo. Texas is a non-disclosure state so there is no sale price to band on — this is every residential deed transfer in Harris and Dallas counties.
Download seed-texas.csv 3.7 MBPeople who already bought. They are out of market for years. Owner-occupants only — absentee and investor buyers are left in on purpose, since they are the ones most likely to buy again.
Download exclude-recent-buyers.csv 14.8 MBA different behaviour to model. The lists above are people who bought once, recently. These are people who own three or more residential properties they do not live in — landlords and small operators who buy repeatedly, year after year.
Built the same way and at the same cost: Florida's statewide assessment roll for all 67 counties, plus the Harris and Dallas appraisal rolls. Owner-occupied homes are stripped out using the homestead exemption, so every row is a genuine investment property. Company-owned portfolios are resolved to a named human through the Florida corporate registry.
The one to start with. Big enough to be real investors, small enough that a rebate on every purchase actually moves them. Out-of-state owners excluded.
Download seed-investors-bestfit.csv 1.4 MBStatewide, all 67 counties. The broadest investor seed we can build from public record.
Download seed-investors-florida.csv 2.0 MBSmaller because Texas has no statewide roll and the Dallas file ships no owner city, so only Harris carries a full mailing address.
Download seed-investors-texas.csv 0.3 MBA tighter, higher-intent seed. Fewer rows, but every one is someone who buys as a business.
Download seed-investors-portfolio6plus.csv 0.5 MBAll investor rows in one file, if you would rather segment inside Meta.
Download seed-investors-all.csv 2.3 MBDo not apply the recent-buyer exclusion file to these. Investors buying again is the entire point — excluding recent purchasers would remove your best prospects.
Same six columns, same upload path. Expect a slightly lower match rate than the buyer seeds: about a third of these rows are company-owned portfolios resolved to an officer, and that person's mailing address is sometimes an office rather than a home.
Three separate audiences, one per file. Name them clearly.
The columns are already named the way Meta expects. Leave email and phone unmapped. When it asks where the data came from, choose partners, not "directly from customers" — that is the accurate answer and it matters.
Two lookalikes: one from the Florida seed, one from the Texas seed. Location United States, size 1%.
Do not build a lookalike from the exclusion file. That one is only ever used to exclude.
Florida lookalike into the Florida ad sets, Texas into the Texas ad sets. Then add the exclusion list to every prospecting ad set — including the ones with no lookalike in them.
The seed is a model input. Meta studies those people and finds others who resemble them. The seed members are never the target — they are the training example. Someone six months before they buy looks almost identical to someone who just bought, and that earlier person is exactly who we want.
The exclusion is a delivery filter. Those specific individuals do not see the ad, because they already did the transaction.
So the exclusion list is not there to protect the lookalike ad sets — Meta handles that. It is there for everything else in the account: DPA prospecting, metro carousels, broad and interest targeting. None of those have any idea who recently closed.
Honest read on size: 391,163 people against a FL + Houston + Dallas adult pool north of 30 million is a small slice. The reason we think it is worth more than the raw percentage suggests is that Meta optimizes toward people who engage with housing content, and someone who was house-hunting six months ago still looks exactly like that — so recent buyers are probably overrepresented in who we currently reach. That is a hypothesis. Once the list is uploaded we can measure the overlap and find out.
Meta blocks lookalike audiences and audience exclusions on anything filed under the Housing special ad category. If a campaign is flagged Housing, none of this is available inside it. Use these in the non-Housing campaigns.
These names exist to be hashed and matched inside Meta. That is the entire purpose. Not a call list, not an email list, not to be exported anywhere else.
All free public downloads. LLCs, trusts and corporate buyers are filtered out, so these are consumers. Both spouses are included wherever the record shows two owners.