After 17 years, we finally “cracked” a $100M churn problem at PayPal. Zero fancy tech. Just a spreadsheet, some simple SQL, and a physicist named Ben.

@benramsden studied physics @ Cambridge, and I seriously envy his ability to work through a thorny problem from first principles. So I gave it to Ben, and Ben gave it to his intern.

But first... Einstein said “If I had an hour to solve a problem, I'd spend 55 minutes thinking about the problem and 5 minutes on solutions.”
PayPal prob loses 1,000,000 merchants per year for various reasons, so step 1 is to narrow down the problem!
PayPal prob loses 1,000,000 merchants per year for various reasons, so step 1 is to narrow down the problem!
We narrowed it down by applying exclusions, crossing off things that look like churn, but aren’t. What can we rule out first? Account closures.
Almost nobody ever closes a PayPal account, because there’s no monthly fee. They just stop transacting. So we needed to look at accounts who “go dark.”
(By the time customers cancel it’s too late! They may have given up on your product months or years earlier - look for behavioral churn.) The next thing we excluded were “one-and-dones.”
Many businesses sign up for PayPal accounts, get paid a bit, and then “go dark.” We found that most of them are small / occasional users who don’t really need the product every day. That’s
1. Not much revenue and
2. Not addressable anyways.
Off the list.
1. Not much revenue and
2. Not addressable anyways.
Off the list.
What about new signups with legit businesses who have a bad first experience? That’s a serious problem! But it’s not churn. It’s an onboarding & activation problem. (That may sound like splitting hairs, but it’s an important distinction…)
(If you go on a first date with somebody, and they ghost you, that is not a “divorce” it’s an “activation” problem.)
We also had “false positives” who stop transacting for a while, and then predictably re-start. (e.g. halloween costume stores, monthly billers, Burning Man, etc.)
Also, a bit of churn is “non-regretted”: people we kicked off the platform for bad behavior. Rule them out.
Also, a bit of churn is “non-regretted”: people we kicked off the platform for bad behavior. Rule them out.
So now we’re down to tenured, well-behaved, non-seasonal merchants who “go dark.” Unfortunately, that’s still hundreds of thousands of accounts. We need to make another cut.
PayPal’s B2B revenue is quite concentrated: ~90% of revenue comes from ~10% of their merchants. So we focused on revenue churn rather than account churn. That narrowed the problem considerably! - maybe a couple hundred merchants per year.
So what’s left? Large merchants who have been with PayPal more than 3 months, and transact regularly. What’s happening with them? I wish it were that simple…
PayPal long ago fixed the major customer experience “killers.” Now it’s all “death by a thousand paper cuts” scenarios. How do you find a thousand paper cuts? You get a summer intern, that’s how!

Our diligent intern spent months reconstructing each history - logging into every system: Risk, compliance, customer service, etc. to see where we screwed up. (Getting access to systems was half the work!)
Next, he bucketed them into about 20 “killer” scenarios... and we had our “fancy predictive model” and ready to take action!
(By ‘fancy predictive model’ I mean Ben ran queries each week to flag merchants hitting any of 20 killer scenarios.) He'd send the results to customer service, who would call the customers and fix the problems.
The merchants we helped were delighted. (And the shareholders should be too!) So, if you have a churn problem, here’s how to focus your thinking:
1. Focus on behavioral churn, cancellation is too late
2. Separate activation from retention - use different fixes
3. Focus only on addressable & regretted
4. Account churn vs. revenue churn - is your revenue concentrated?
5. Then, dig into the details! (Intern optional).
2. Separate activation from retention - use different fixes
3. Focus only on addressable & regretted
4. Account churn vs. revenue churn - is your revenue concentrated?
5. Then, dig into the details! (Intern optional).
Since fixing at scale was too hard, we narrowed down the problem to solve at a sub-scale level.
Helpful? I send a 2-minute/week email to help founders learn enough about startup growth to be dangerous
. You can read them all & subscribe here. https://blog.startupcorestrengths.com/
Helpful? I send a 2-minute/week email to help founders learn enough about startup growth to be dangerous

Many thanks to @hnshah @davemcclure @dlevine815 @productjonas @IdaMannoh and @AnkitGordhandas for feedback on my draft!
And I finally managed to track down the intern! Michael Cooke @Michael22489026 Bet you didn't expect to be in a TweetStorm today.