If you like to follow Max and Caroline in 2 Broke Girls in their endless efforts to build a cupcake empire, you can't help but give it a try yourself. Unfortunately, the amount of available recipes is quite extensive (Chefkoch.de provides 385 recipes), so it would probably take at least one or two decision trees (cakes) to find the best recipe. If you have to choose between Rafaello cupcake, Black Forest cherry or even the chocolate muffin with marshmallow hat, you will definitely have serious first-world problems. 

That's why I'm setting smaller goals for myself first: I want to find a good basic recipe on which to base my own flavorful ones. So, in good statistician fashion, I proceed as follows: 

The best ten bids recipes from Chefkoch.de are quickly collected! Of course, these recipes are already a little cleaned up: special ingredients for the special taste (strawberry, Nutella, ...) have been removed. In the next step, the units (gr, pieces, tsp, pinches, ml) are still converted to a common unit gram. So for the next step you can access the following data.

Salt and baking soda seem to me to be a nice example of using the coefficient of variation as an alternative measure of dispersion, versus variance. But how do you come up with a recipe now? I simply take the mean percentage of each ingredient and normalize the result to 100%.