The situation: the Search channel hits a ceiling
Amsterdam Data Academy trains career switchers and working professionals in data, AI and Python through online courses and bootcamps. The Google Ads account was performing well. The problem was the auction itself.
Data and AI training is a saturated paid search market. Generic terms like “AI agent course” pull huge search volume, but most of it is research traffic, not buying intent. CPCs on those terms sit around 20 euro. To take a meaningful share there you have to bid at the budget level of the large training providers and aggregators. That does not fit ADA, and it does not need to.
Search did what search does well: it caught prospects who were already actively looking for ADA or a specific bootcamp. Any growth on top of that had to come from outside the search auction.
Client Details
CLIENT NAME
Amsterdam Data Academy
INDUSTRY
Education
SERVICES
Google Ads · Tracking & Analytics · Reporting
TIMELINE
2024 – 2026
The approach: new signals, new funnel
Three changes that reinforce each other.
Sharpened the conversion signals. The Google Ads account already ran multiple conversion actions with values attached, so smart bidding had a workable picture of what mattered. Moving to Demand Gen, we sharpened it further through the tracking setup: conversions such as lead-magnet forms, Calendly bookings and contact forms received values calibrated to lead quality. That gave smart bidding an even clearer read on what points towards enrolment.
Built a lead-magnet portfolio. Most education advertisers treat their landing pages as direct response: ask for a call straight away. That works for in-market prospects, but it misses everyone earlier in the decision process. Together with ADA we built a portfolio of lead magnets, each tied to a specific course interest, such as an AI jobs report. In April 2026, lead-magnet forms made up 49% of all primary conversions, against 31% in Q4 2025.
Shifted budget from Search to Demand Gen. Search has a ceiling. Demand Gen captures attention before buying intent exists, with visual creatives on YouTube, Discover and Gmail that interrupt and inform. We scaled Demand Gen from just over 200 to 800 euro per month, while Search kept running to protect brand traffic and cover high-intent terms. The cost per lead dropped to 17 euro.
-76%
COST PER LEAD
3,7x
LEADS PER WEEK
8,1
ROAS
What each layer delivered
With the new structure in place, every part of the funnel had its own job.
Demand Gen became the main source of new leads. Visual creatives on YouTube, Discover and Gmail drew attention at the top of the funnel, from people who were not yet searching for “Python course”. Lead-magnet forms went from a handful a month to more than twenty in Q1 2026, with a CPL that stayed under 20 euro.
Search became the safety net at the bottom of the funnel. Branded campaigns caught people who already knew ADA, so they did not end up with a competitor. Bootcamp and Course campaigns picked up the searches from people ready to enrol. Not another euro wasted on expensive top-of-funnel terms where ADA was outbid anyway.
The lead-magnet portfolio was the conversion mechanism. Each piece of content was tied to a specific course interest, so leads came in already partly qualified. 14.7% of leads turned into a sales conversation, and 51% of those conversations closed as a deal.
The result: a steady lead stream that builds on itself
Lead volume nearly quadrupled. Cost per lead fell by three quarters. Across all of 2025 the account returned 8.1x ROAS, and roughly half of all sales conversations became an enrolment. Calendly bookings held steady, which confirms the new lead mix is the right one.
The next phase was clear: scale Demand Gen further, add audience-specific lead magnets, and feed deal data back into Google Ads so smart bidding optimises on actual enrolments, not just leads.
ADA started out paying top prices to bid in a search auction. After the shift, it had a lead stream that built on itself.