How do you read advertising data correctly? In a team that had worked with advertising data for many years, that was the question we chased most. When we could not find a tool that gave us that reading, and the meaningful insights that follow from it, we built our own system around our needs. We knew what we needed, or at least we thought we did.
Along the way, one question stayed with me: how could I go beyond what we already had? I believed that if I could reach more data, I would get more meaningful answers.
I started researching that question. The research first became blog posts, then kept growing. I never thought it would turn into a book. It moved forward through long rounds of questions, answers and debate with an AI assistant.
The book orders measurement methods on a ladder by the question each one can answer. When channels reinforce each other, their separately measured effects add up to more than 100%, so there is no single natural split; a sale can be divided among channels only under stated assumptions. The book treats a platform measuring the results of the ads it sells as a separate kind of error. It shows how attribution, experiments and models work together, and finally tests whether AI can make these distinctions. The source of every number in the book was opened and read; the one figure taken from a secondary source says so where it appears.
I set out looking for more data; the work carried me from the quantity of data to evidence and cause. The numbers we thought we understood were not saying what we thought.
Edip Ali Avcıbaşı, November 2026