iPrice: Engineering PPC Automation at Massive Scale

Great experience! Tomas has outstanding technical abilities, know-how, productivity and work ethics. He adopted immediately in completely different culture of Southeast Asia.

Matteo Sutto

CMO at iPrice

The “Impossible” Problem

iPrice had already built massive success through organic traffic alone.

But when it came to paid search, agencies delivered a blunt verdict:

To promote a business this large, you’ll need $100k–$200k just for tooling.

That budget wasn’t acceptable—especially just to test whether PPC could even work.

The real challenge wasn’t ads.

It was setup in low-cost mode and later attribution, and economics.

• 5+ billion products with a constantly changing portfolio
• 7 markets: Malaysia, Indonesia, Singapore, Thailand, Vietnam, Philippines, Hong Kong
• Affiliate-based revenue with 30+ day conversion lag
• Heavy ad-blocking across APAC → Google Analytics was unreliable

Most teams would stop here.

An Automated PPC Engine

Within 6 months, I had:

🎖 A fully automated campaign builder running across all 7 countries

🎖 Profit conversion goals maintained automatically, without enterprise tooling

🎖 Improved attribution and tracking leaks fixed across the entire business

🎖 SEO performance improved using SEM learnings

And all of it was built with an agile, test-first approach.

How I Did It

1. Building the Infrastructure (Without Fancy Tools)

Instead of buying expensive platforms, I built what was actually needed:

• Automated campaign and ad group generation
• Dynamic structure based on daily product changes
• Scalable across 7 countries
• Designed to optimize for revenue and revenue, not just clicks

The system was built to adapt, not over-engineer.

Since product feed was not available, I used the Kibana landing page exports and created my own “dirty feed”. Scraper enhanced it with extra details from the category pages (like number of products, cheapest item, biggest discount etc.)

The second input was ad builder from a template and keyword generator. The ad template builder was later used in SEO to test what variation drives higher CTR and later revenue per session.

All of the inputs were processed in Power Query where I imported the results directly to Google Ads.

2. Attribution Built for Reality (Not Theory)

Getting the right conversion inputs is as important as the right structure and setup. First, I started with importing the revenue data on session level via measurement protocol to Google Analytics. But I faced two issues:

First, hitting 10 million hits per property per month.

Second, heavy ad-blocking across APAC → Google Analytics was unreliable

What I implemented instead:

• ElasticSearch-based attribution from logs (more reliable than GA in ad-blocked regions)
• Exit-click tracking exported from Kibana
• Click IDs merged with:
  • Affiliate transaction files (HasOffers, WTF, CJ, etc.)
  • Stored traffic sources (GCLID)
• Revenue imported directly into Google Ads

This uncovered tracking issues across all channels, not just PPC.

Result: The business stopped losing money from misattributed traffic—company-wide.

• Extrapolated RPS per exit click, calculated using:
  • Historical subcategory RPS
  • Merchant-level CR
  • AOV × commission rate
• Data loaded into Power BI
• Subcategory-level optimization using:
  • tCPA (exit clicks generate volume easily)

This turned delayed, messy revenue into actionable bidding signals.

Results (6 Months)

• Automated PPC campaigns across 7 countries
• Scalable system handling 5+ billion products
• No enterprise-level tooling required
• Tracking fixed across all acquisition channels

Bonus: SEO uplift driven by SEM insights

The entire SEM evaluation framework was later presented at PPCEE (Central Europe’s largest PPC conference), showing how paid search insights can systematically improve SEO.