From 25 to 461 credit applications a month, at a third of the cost

The results

• 3.3xCost per lead went from $27 in January to $8.20 in August. July closed at $7.80, the best month.
• 14xMonthly leads went from 25 in the launch weeks to 461 in June, the peak, on a $5,700 budget.
• 2,000+Credit applications in eight months on $25,000 of ad spend. A blended cost per lead of $12, including the launch.

The context

Buy-here-pay-here dealers sell financing first and cars second. The customer has bad credit or none, the lead is a credit application, and Meta treats the ad as a credit ad: a special category that removes age, gender and narrow location targeting and limits what the algorithm can learn from. The audience is broad by rule, so the structure and the creative have to do the work.

The account had no lead program when I built it in January 2026. The dealer has two lots twenty miles apart with different inventory, and a customer base that is half English, half Spanish.

What I did

Stage 1: one campaign per lot, one ad set per language

January to March 2026

• One campaign per lot, CBO, each with a broad English and a broad Spanish ad set inside a 20-mile radius of the lot. Under the credit category there is nothing to narrow, so the split is by language and location, which is what the sales floor actually sees.
• Optimized on the pixel's lead event, lowest cost, no caps. Every creative runs in both markets, so the auction decides which lot's audience responds to it.
• Cost per lead went from $27 in the launch weeks to $17 in February and $11 in March as the ad sets left learning.

Stage 2: a video drop a month, mirrored to both lots

April to August 2026

• Ten videos a month from the content team, launched in both markets within two days of delivery. Each month gets its own ad set, cloned from the previous one with the targeting intact, so a batch is judged against last month's on its own cost per lead and the old ads stop competing for budget.
• Testimonial-style videos won: first person, the customer repeating the hook, the lot's address on screen. Cost per lead went from $17 in April to $12 in June with the budget at its highest.
• In July the client cut the budget in half. Leads dropped by a quarter, not a half: 461 in June on $5,700 became 367 in July on $2,900, at $7.80 a lead. The cut barely showed on the sales floor.

What did not work

• April. New budget and new creative in the same week, and cost per lead went from $11 to $17 while the new ad sets learned. Scaling and refreshing together costs a month.
• Meta caps an ad set at 50 ads, paused ones included. By May the launch ad sets were full of tests that could not be deleted because they had spend. A new ad set per month is now the rule, but one drop had to be rebuilt to learn it.
• The videos show one lot's address and run in both markets. It comes that way from the content team and it has not hurt cost per lead, but it is not how I would brief it.