


CarProf Autokeskus is a car dealership and service center in Estonia. The company sells cars for different budgets, provides car selection and inspection services, and operates its own repair shop. The business had been active for a long time, but advertising was not delivering predictable results.
Our goal was to rebuild the analytics and optimization system in Meta Ads from the ground up: set up proper Lead tracking, restore complete learning data for the algorithm, filter out bots, and move campaigns away from the “cheap leads at any cost” model toward real lead quality.
In practice, the budget was being spent, but the system could not learn from real outcomes.
We did not “tweak” the old campaigns. We rebuilt the system from scratch.
From that point on, every lead was actually being recorded.
One of the main issues was bot leads. We:
The goal was simple — filter out junk before it reached the sales manager.
We rebuilt targeting:
The focus shifted from “cheap” to “actually reachable by phone.”
We analyzed:
Only after that did we scale the combinations that were bringing in real, live customers.
The main emphasis was on trust and transparency. We were not just selling a car — we were selling peace of mind and safety.
The second focus was a wide selection for different budgets: first cars, family cars, and premium options.
The third direction was the in-house service center: diagnostics and repairs, so the customer is not left alone after the purchase.
We launched:
Video worked well for warming up the audience and building trust. Static creatives drove faster decisions and lead submissions.
In one of the more effective periods from the provided ad account data, we saw 436 website leads with a reach of 58,045 and a cost per lead of €1.11.
A full restart instead of patching the old setup.
Correct event configuration gave Meta real learning signals.
Rebuilding the forms and submission logic filtered bots out at the entry point.
Technical bot protection and strict audience filtering significantly reduced spam.
Optimizing for lead quality rather than lead volume made it possible to scale only live-performing combinations.

