"proportional" covers 2x-10x. Again, the calculation is not public (or: I didn't find it), and I have no idea what the revenue from Google Shopping in the affe

"proportional" covers 2x-10x. Again, the calculation is not public (or: I didn't find it), and I have no idea what the revenue from Google Shopping in the affected countries in the given years were, Google Shopping is not AdSense or AdWords so it's not going to be huge. But the method is known: the starting point of the fine is sales (revenue, not profit), and is up to 30%. Since this went on for many years in multiple countries, we can imagine it will be on the higher side. This is also increased by the duration of the infringement, with a multiplier calculated on the basis of the days of participation in the infringement. If we expect Google Shopping to run with a 4% profit margin on sales, a fine >30% is well within 2x-10x of the benefits gained. But: your worry is "they get caught only X% of the time". This is _also_ considered in the fine guidelines, because the fine will be increased with each subsequent violation by the same entity. Every time you get caught, you will be punished worse for all subsequent violations, so each fine changes the loss/gains calculation for all your future anticompetitive behaviour.