but this fine is exactly based on making _this_ bad behaviour unprofitable, the amount is proportional to the estimate of the advantage that google got from its base behaviour.
That is for this instance where they were caught and it took a huge amount of resources, willpower, and luck for that to be followed through on to get to the fine.
Basically in this case it was breakeven minus lawyer fees if they get caught which only happens X% of the time.
That is why there are usually punitive premiums so that the business calculus never makes sense as the cost of doing it x risk of being caught is always a negative compared to potential profit.
TLDR: Fine should be 2-10x times benefit gained or higher to prevent even thinking about trying to get away with it.
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.
