These subscribers aren't real, and neither is the reason they left. Whoever built this file decided who cancelled by running a simple formula, then saved the answer as the churned column. So this dashboard doesn't show you why people leave Netflix. It shows you their formula.
You can see the formula in the numbers. Of the 1,636 people who watched 5 hours or more and logged in within the last 30 days, almost none left. Of the 873 who watched under 5 hours and stayed away more than 30 days, every single one left — no exceptions. Real customers never split that neatly.
Two consequences worth knowing. A prediction model built on this would be right about 99% of the time, which sounds excellent but only means it rediscovered the formula — it learned nothing about people. And age, gender, region, device and favourite genre make no difference whatsoever: every one of those groups leaves at roughly the same 50% rate, so any story you tell about them would be noise.
Useful as a puzzle to reverse-engineer. Not useful as a forecast.
Each square is a group of subscribers. The number is the share of them who left — green means almost nobody, red means almost everybody. Drag the dotted lines to move the cut-offs and watch the four boxes update.
Five yes/no checks, each worth some points. Add up the points and call anyone above the line a leaver. Change the weights and cut-off to see how close you get to the real answer.
The share of each group that left, longest bar first. Click any bar to narrow the whole page down to those people; click it again to undo.
The shape of the people who left, laid over the people who stayed. Where the red and green humps sit apart, that measure separates them.
Monthly fees attached to churned subscribers in the current view.
Sort any column. Search matches id, region, tier, device, genre and payment.