Meta lookalikes on a small customer list: 1% or go wider?

Say you've got around 800 past customers, mostly from the last two years, and you want Meta to find more people like them. The default advice is to start with a 1% lookalike because it's the closest match. But with a seed that small, a 1% audience in a mid-sized country can end up pretty narrow, so Meta keeps spending on the same few thousand people while frequency climbs.

I'd rather run a 3% to 5% lookalike next to broad prospecting and let the creative do the filtering, because seed quality matters more than the percentage. Honestly, I'm not sure the tighter match is worth the frequency penalty for most accounts. Has anyone seen a tight lookalike beat a wider one on cost per purchase with a seed that small?

Mira's right that with 800 seeds the seed is doing all the work, so check the match rate on the customer list before you pick a percentage. Meta only matches a share of uploaded emails and phones to real accounts, and that share can come in well below what you'd expect. If it's low, a 3% lookalike just widens a weak signal.

Also, a wide lookalike running next to broad prospecting will overlap heavily, so compare the two on cost per purchase over the same dates, not on reach.

Sasha, the match rate is the first thing to check. Also, I overstated my earlier point. 1% of a mid-sized country is usually hundreds of thousands of people, so frequency climbs because the budget is small, not because the audience is tiny.

Mira, a small budget spread over hundreds of thousands of people should give you low frequency, not high. Frequency climbs when weekly spend is big compared to how many people the ad can actually reach, so the number that decides this is weekly spend divided by the audience size Meta shows for the ad set.

That's why the 1% versus 3% to 5% choice comes down to spend. If you're putting in enough money that a 1% audience gets hit four or five times a week, the wider lookalike is probably the better call, because it spreads the same budget over more people. If frequency sits under two or three, the tighter seed is usually worth testing, since seed quality is what you're paying for. The seed still matters more than the percentage, but spend against audience size tells you whether the percentage is even the bottleneck.