I like waffles and even though I haven't done any in-depth quantitative research I would like to claim that the majority of my fellow human beings likes them as well. But not all waffles are the same, Belgian ones for example are characterized by their caramel texture while the Swedish are crunchy and less sweet. I prefer the classic heart-shaped waffles, and wondered which was the best recipe for them. As usual, I've taken a data-driven and completely overengineered approach to solve this question.

tl;dr: take me to the recipe

1 - getting some data

For the perfect recipe, first of all I needed quite a lot of data. Starting with the search term "waffle", I employed a small scraper to save around 940 recipes from the Chefkoch.de platform, which is very popular in Germany. Again, I wanted to match the local taste, not the Swedish one…

After that, it got messy: Recipes are probably the most unstructured kind of data I've ever had to deal with. So I wrote a series of notebooks to standardize spellings of ingredients and quantities as much as possible. The key effort was to convert the most common ingredients to grams to allow for consistent analysis. I’ll push ne notebooks to Github once they are more or less readable.

Finally, the data was neatly parsed in a database and the actual quest for the perfect recipe began.

2 - the quest for the best recipe

The task was to gather a deeper understanding of the composition of waffles in order to be able to derive a perfect mixture of ingredients.

First, I asked myself how complex a good waffle recipe actually has to be. Of the roughly 800 recipes analyzed (I had to filter out quite a few non-sweet recipes), most are made up of six to nine different ingredients. My recipe should therefore be in this range of complexity.

So what ingredients make a perfect waffle? My recipe should not only please myself, but as many other people as possible - I don't like to eat waffles alone, so I occasionally pass them out in the neighborhood.

Parsing the ingredients I could answer this question quite easily by calculating the frequency of different ingredients related to all recipes. In the adjacent table Ilisted the most frequent ingredients.

I had so many questions: Like, wow important is vanilla sugar and does it help to add some cinnamon? I could have tried out all the variants in some sort of A/B test (with some negative side effects on my weight). Fortunately, Chefkoch.de also has a rating function and so I could rely on the wisdom of the waffle-crowd: I calculated the average rating of recipes for each of the top ingredients: Once for recipes with and once without said ingredient. The difference between the two ratings makes it easy to check whether, for instance, milk or water based waffles are rated better, or whether vanilla and cinnamon are more popular with the majority of people than waffles without.

The good news first: I'm not alone in my dislike of water-based waffles, and vanilla sugar should apparently be in every recipe. Cinnamon, interestingly, has no effect. I like it, so I kept it in my recipe.

Now that I had the ingredients of my recipe, it was a matter of determining their exact ratio to each other: I divided the recipes containing my desired ingredients into groups according to their Chefkoch.de ratings and calculated the weight-based ratios.

It turns out that really good recipes seem to contain more milk and a little less egg than moderately rated recipes, but at the same time also a little more sugar - people like it sweet…