2,000+ Leads at ~$8: How Proper Optimization Delivered Scale for an English School

Key challenges before launch

• irrelevant optimization event
• unstable lead quality
• mixed learning directions
• insufficient data quality

How we solved it

• moved to trial lessons as the main goal
• implemented Conversion API
• segmented campaigns by direction
• created separate landing pages for each direction

Results

We generated 2,031 leads, including 1,761 website leads.

Additionally, we recorded 1,711 conversations and 1,417 applications.

The total budget was $42,414, and the average lead cost was ~$8.22.

Even at a scale of 2,000+ leads, the CPL remained stable.

Users interacted through several channels: the website, messages, and application flows.

• The correct optimization goal gave the algorithm a clear signal.
• Removing the secondary event reduced noise in campaign learning.
• Conversion API improved data quality and stability.

We changed the optimization goal

Instead of calls, we moved to optimizing for trial-lesson registrations. A trial-lesson sign-up is a high-intent action, so the algorithm began finding more relevant users.

Conversion API

• improved data transfer
• increased campaign stability
• accelerated algorithm learning

This made it possible to send more data back and make optimization more precise.

Campaign structure

• online
• offline — children
• offline — adults

This prevented mixed audiences and improved conversion rates.

Landing pages

We created separate landing pages for each direction. That improved conversion rates because each audience saw a more relevant offer.

Why it worked

A trial lesson is the closest action to an actual purchase.

Optimizing for that event gave the algorithm a clear signal and allowed us to scale without losing quality.