About Taste Twins

Why we built this

We kept getting book recommendations from algorithms that only knew our genre, not our actual taste - and kept getting recommendations from friends whose taste didn't line up with ours at all. Taste Twins started as a way to answer a simple question:

Out of everyone reading right now, whose taste lines up with mine?

Once you know that, their favorite unread-by-you book is a much better bet than anything a genre tag can offer.

How correlation actually works

We look at the books you and another reader have both rated, and measure how closely your ratings correlate using a real statistical method (Pearson correlation), the same one used in collaborative-filtering recommendation systems generally. Two readers who tend to rate the same books similarly - both loving the ones they loved, both lukewarm on the ones they were lukewarm on - get a high correlation. Two readers whose ratings run opposite get a low or negative one.

That number alone isn't the whole story, though: a correlation based on 5 shared books is much less certain than one based on 50, even if the raw number looks the same. So every correlation also carries a confidence level based on how many books you actually have in common - more shared books means we trust the number more. A high correlation from only a handful of shared books shows up as real, but lower-confidence, rather than being treated the same as a well-established one.

Ratings are one signal. If you and another reader both have a taste profile, built from likes and dislikes in your reviews, that overlap is blended in too.

See it for yourself

You can try this with your own reading history before creating an account. Upload your Goodreads or StoryGraph library and see a real correlation, no signup required.