Are lookalike audiences still worth using for a local service company?
Sometimes, but the constraint that binds locally is geography, not similarity. A lookalike is intersected with your service area, so a tight percentage in a small metro can collapse to an audience too small to deliver efficiently. Seed quality matters more than percentage: build the seed from completed high-value jobs, not from every lead. In many local accounts, broad targeting with strong creative performs at least as well.
Why lookalikes behave differently at local scale
A lookalike ranks the population of a country by similarity to your seed and gives you the top slice. Then your location setting cuts that slice down to your service area. For a national advertiser the slice is enormous. For a contractor covering a few dozen zip codes, a tight percentage can leave a pool small enough that frequency spikes within days and cost per result climbs for reasons that have nothing to do with quality.
This is the practical error behind most disappointing lookalike tests. People assume a tighter percentage means better people. Locally, a tighter percentage mostly means fewer people.
Seed on outcomes, not on submissions
The seed defines what similar means. If you seed on everyone who filled out a form, you are asking Meta to find more people who fill out forms. That is a real behavior and it is not the behavior you want.
Better seeds, roughly in order: customers with completed high-value jobs, customers on a maintenance plan, all paying customers, appointments set, qualified leads. Meta's documented minimum seed size is small, but performance near that floor is noise. If your list of completed high-value jobs is thin, use all paying customers instead of dropping to a worse definition of good.
Test it against the honest baseline
The comparison that settles this is not lookalike versus interest targeting. It is lookalike versus broad, in the same geography, with the same creative and the same optimization event.
Run both, give each enough time to leave the learning phase, and compare on your downstream funnel rather than on cost per lead. In accounts where the audience is already small, broad frequently wins because the delivery system has more room to find people and the creative is doing the qualifying anyway.
Where a local lookalike does earn its place
Two situations reliably justify one. First, when you are expanding into an adjacent market and have no history there: a lookalike seeded on your existing best customers gives delivery a starting hypothesis in a territory where you have none. Second, for a specific high-ticket service where your best customers share unusual characteristics that broad targeting dilutes.
In both cases, keep the seed refreshed on a schedule. A list uploaded once and never updated slowly stops representing your business. Automating that refresh from the CRM is a small job with a long payoff and it uses the same customer segmentation that drives customer intelligence and the audience sync side of Meta integration.
Topics: lookalike · audiences · targeting · first-party data · Meta
Have a version of this question about your own business?
The useful answer usually depends on which systems you run and how they're connected. That's a conversation, not a blog post.