Okay, here’s the pitch: everybody is allowed to choose their top 5 or top 10 (whatever, just keep it smallish) episodes… and then you do this neato thing where you correlate the likes. You can, for any episode, show which other episodes people most often liked. Makes sense to me (because I did something like this once) but maybe an example is in order, using just numbers because I don’t want to write long episode titles…
Four people each indicate three favorites. Person One favorites episodes 101, 102, and 103. Person Two favorites episodes 101, 103, and 200. Person Three favorites episodes 101, 200, and 201. Person Four favorites episodes 101, 200, and 300.
Then you get a mapping of episode correlations, like:
- 101 / 102: 1
- 101 / 103: 2
- 101 / 200: 3
- 101 / 201: 1
- 101: / 300: 1
- 102 / 103: 1
- 103 / 200: 1
- 200 / 201: 1
- 200 / 300: 1
That’s the backend nonsense. What the end users get are two-way recommendations. When someone goes to episode 101 on the product page, it can say “people who liked this episode also liked episodes 200 and 103. Maybe you wanna buy that, HMMMMM?” And Person One, without going to a particular product page, might see something like “based on your favorites, you might also like episode 200.”