Published numbers
The exit rate
Of everybody who deleted their Bearing account last month, how many said they were leaving because they met somebody. Published monthly, whether or not it flatters us.
How the number is made
Counted, not estimated
When somebody deletes their account they are asked, on the confirmation screen, why they are going. One of the answers is I met someone. Answering is optional and there is no retention offer anywhere on that screen — no discount, no “are you sure” on top of “are you sure”, and no thirty-day limbo. A published exit rate is only honest if leaving is as easy as arriving.
The deletion destroys everything we hold about that person. What it writes, in the same transaction, is a single anonymous row: the reason they picked, the metro, and how many days they were a member. No identifier of any kind. That row is the only thing that survives an erasure, so it must not be a way to reconstruct one, and this page is the reason it exists at all.
The figure above is a count of those rows for one completed calendar month, divided by all of them for that month. There is no modelling, no weighting and no adjustment. You are reading a fraction of two integers.
What it misses
The parts of this number that are not good
Publishing a metric and then not publishing its holes is the same trick as not publishing it.
- It only counts people who deleted. Somebody who pauses their account, or who cancels the subscription and leaves the profile sitting there, writes no row and is counted nowhere — not in the top of the fraction and not in the bottom. And pausing is the obvious thing to do when you have just met somebody, which means the ending this page exists to count is also the one most likely to go uncounted.
- It is self-reported and unverified. We take the answer at face value. We do not ask who, we do not ask for evidence, and we are not going to start.
- Answering is optional. Somebody who leaves without picking a reason lands in the denominator and not the numerator, which pushes the published figure down. We would rather understate this than force an answer out of somebody on their way out of the door.
- “Met somebody” does not mean “met them here”. Some of these people met somebody somewhere else while holding an account. We cannot separate the two and we are not going to pretend the wording does.
- Small numbers are noisy. Early on, one or two departures will swing the percentage wildly. The counts are printed next to the percentage for exactly that reason: read the integers, not the ratio, until the integers get large.
Why publish it at all
It is the only version of the claim that costs something
Every dating app says some version of “we want you to find somebody and leave”. It is free to say and nobody publishes whether it happens, because the number is a measure of churn and churn is the thing the business is built to avoid.
Here it is the other way round, and that is not virtue — it is the arrangement. The price is published with a ceiling, the feed is finite, and there is nobody on the cap table whose return depends on you still being here next year. A company that had promised its investors engagement growth could not print this page. We can, so we do, and you get to check the claim instead of taking it.
If this number is bad, it will be on this page being bad.