You wanted more proof? Here’s all of it.

Veteran-owned window and door replacement, working with us since 2023. In 2026 so far, $86,000 of Meta spend has produced $1.06M in booked jobs — a 12.2x return, 68 new customers at $1,270 each, average job value $15,544.

The comparison is the interesting part. They spent roughly the same on Thumbtack that year. Thumbtack returned about 5x; Meta returned 12.2x, and the Meta customers spent close to twice as much per job. Rented leads compete on price. Owned ones do not.

An independent salon in Cherry Hill. In June we set up booking tracking and pointed the Meta campaign at completed online bookings, not clicks.

From June to October 19, 2026 it produced 343 online bookings at $7.14 each.

Three weeks, September 28 to October 19, 2026. $1,679 in ad spend produced $15,944 in tracked sales — a 9.5x return on 92 orders.

The athlete, Jax Forrest, came through our athlete NIL partner, Axe Sports Group. His ad did the selling.

Meta-attributed tracked sales from the presale campaign, not profit.

Before us, PlayerVault got 45 cents back per dollar; since May the account runs at 2.31x overall and 3.1x across August and September.

The method is unglamorous: every new drop gets its own launch on a small test budget, the ones that work get scaled, and the ones that do not get cut before they burn money.

Meta-attributed return multiples.

$31,984 in ad spend since January 2025 has produced $148,720 in signed work, a 4.65x return — with a further $260,000 sitting in open quotes. Quotes, to be clear, are not revenue; they are work that has been priced and not yet won or lost.

Meta is now their single biggest source of new customers at 43% of their customer base — ahead of referrals, ahead of Google, ahead of Thumbtack.

Earlier work, outside our current focus. Bonchon Philadelphia has taken 198 online orders straight from the ads, at $31.66 in ad spend per order.

The consistency matters more than the total. On the same $1,500 a month, they land 40 to 52 orders every month. Predictable is the point.

No return figure is quoted here on purpose: the ordering pixel does not pass an order value, so cost per order is the only number we can stand behind.