3 ms·
It occurs to me that you could literally retrospectively measure the odds that someone's first experience with the site was due to a hypercontributing outlier m
by patio11 16y ago
It occurs to me that you could literally retrospectively measure the odds that someone's first experience with the site was due to a hypercontributing outlier member. My intuition is that a) they contribute a hugely disproportionate amount of the value on SO (pick any metric you want for this), b) they contribute a hugely disproportionate amount of value relevant to someone's first use experience, and c) if you segment users into "first use was influenced by a hypercontributor" and "first use was not influenced by a hypercontributor", there is a measurable difference in retention rates in favor of the first group.
Proving or disproving these assumptions would be pretty important for strategic direction at a QA site, incidentally, since the world if I'm right looks very different from the world if I'm wrong. Since StackOverflow's data set is public, you could do it as a weekend project. In general, though, this is the kind of stupidly powerful thing you can go with the data you already have in your startup.
- terrym 16y agoI think your intuition is correct - that user contributions generally follow a power law, so when you randomize 1-to-1 interaction among all members, the likelihood of a person in the long-tail getting feedback from another user in the long-tail is high. A good essay on this: http://www.shirky.com/writings/powerlaw_weblog.html http://www.shirky.com/writings/powerlaw_weblog.html