Korva is a scroll of the internet’s best ideas.
Averages mean nothing to individuals
What the average hides
In a large population it is tempting to look at average adoption, find that only 3% use something, and decide the feature can go because nobody wants to maintain it. A layer down, that 3% includes a group for whom the feature is 100% of what they do.
On a marketplace, that group is running a business. Deprecating the feature for convenience is like a Westfield mall switching off the power in the lead-up to Christmas without thinking about it.
How the damage compounds
Network effects make the damage grow. Blowing up one group's core use case sets off a spiral whose effects can be enormous. The product decisions that left Tom Verrilli most disappointed in himself came from relying on averages without looking at the individual use cases hidden underneath.
How can an average point the wrong way?
Statisticians call this Simpson's paradox, where a trend in combined data disappears or reverses once the data is split into groups. A well-known case is the 1973 admissions data at UC Berkeley, analyzed by Peter Bickel, Eugene Hammel, and J. William O'Connell in a 1975 paper in Science. The overall numbers suggested a bias against women, yet within individual departments the small differences ran slightly the other way. The overall figure was produced by which departments women applied to.
Why do network effects magnify one group leaving?
In a two-sided marketplace, each side is valuable because the other side is there. Economists Jean-Charles Rochet and Jean Tirole described this cross-side dependence in their 2003 paper on two-sided markets. When sellers leave, buyers find less to choose from and visit less, and the remaining sellers sell less. Each step of that loop gives the next group a reason to leave.
More from Lenny's Podcast: Product | Growth | Career
