Fun with dating and network matching models
Recently, I got into a discussion via twitter with Ryan Hoover on the key metric for success for @TinderApp, a new entrant in the game of matching romance seekers via online social networks. What got me started was looking at the following equation as an attempt to create the “one metric that matters” (his post here is generally very insightful aside from my one bone to pick):
The number of people connected driven by:
(# of profiles viewed) x (# of profiles liked) x (% of mutual likes)
This didn’t seem quite right to me, so I naturally started to draw. Here is what I came up with:
What this means is that, following Acquisition and Activation (topics for another post), Retention is defined basically as whether you can get them to stay long enough to generate a match or reward.
The time it takes to find a match is directly proportional to:
- The score distribution of each population (e.g., what % of men are perceived as very attractive by the opposite sex [hint: it’s 40%*])
- The acceptability threshold (what is the typical level of attractiveness needed for someone of the opposite sex to rate a “match” [note it would be fascinating to see how this differs between TinderApp and Bang With Friends]), and…
- The average # of views you are able to get out of a user
Sadly, only the last item is controllable by the application, so if we were to choose, it is this statistic that would get the “ONE METRIC THAT MATTERS (TM)” badge. That and whether people are actually hooking up, I guess. But that’s determined outside of the app.
… for now…
*Source: Hot or Not. Seriously. Some guys from Dartmouth wrote a paper on the score distributions a while back which I found fascinating. Check it out here.