Blinkist is a Berlin-based subscription app that condenses nonfiction books and podcasts into 15-minute reads and listens. It acquires new subscribers web-to-app on Meta, through a trial path and a direct-purchase path, and asked Churney to beat the custom conversion event its own team had built.
Blinkist sells annual subscriptions through a trial path and a direct-purchase path. Meta optimises inside a seven-day window. The trial converts on day eight.
Every subscription business carries some version of this gap, and pays for it whether or not it does anything about it. Users who convert after the window never feed back into bidding, so the algorithm never learns what they look like. A trial start is a proxy, not a customer. Budget flows daily towards whoever converts cheaply and early, and the users who fund the business are systematically underbid. None of it shows up in the dashboard, because the dashboard reports what the platform was told to chase.
Most case studies in this category beat an install event or a trial event. This one did not.
Blinkist's MarTech team had already built a predicted-value signal of its own. In practice the account ran best on conversion optimisation towards a custom event of Blinkist's own design, firing for trials and direct purchases with early cancellations removed. That is further than most subscription teams get, and that event was the control. The brief to Churney was to beat it.
"I had heard positive things from two other founders working with Churney, so I wanted to see whether they could outperform our internal signal. It seemed like a straightforward bet: if they beat our signal, we both win. If they don't, we carry on with our own, confident that we are already in a good place."
Holger Seim — Co-Founder and CEO, Blinkist
"We don't want you to mimic what we are doing, because we want you to do it better."
Holger Seim — briefing the test
Churney predicted each new user's long-term value within hours of their first session and delivered it to Meta as the value signal for bidding. The platform then had something it had never had on this account: a per-user estimate of what a trial or a purchase would actually be worth, at the moment the bid was placed.
Blinkist's data team connected its data warehouse through a secure, read-only share. No SDK, no changes to the app, no PII leaving the building. Blinkist's UA team kept full control of campaigns, budgets and creative throughout.
The US test ran Blinkist's own event in one cell and Churney's pLTV signal in the other, everything else identical. Within two weeks the pLTV cell was producing three times the direct purchases of the control.
Blinkist then asked for a second test in new markets, run hands-off from launch, to confirm the result travelled. It did.
"We had made changes to both variants during the first test, so we could not be certain whether the result came from the better event or from our own optimisation along the way. We wanted to be sure. That is why we ran a second test."
Holger Seim — Co-Founder and CEO, Blinkist
Both data teams reconciled the results at user level before anything was called.
US: D8 ROAS 64% higher on the pLTV cell. Twice the share of signups became paying subscribers, the direct-purchase rate ran at two to three times business as usual, and each subscription cost 20% less to acquire.
UK, Canada and Australia: D8 ROAS 31% higher on matured cohorts, hands-off from launch, with a higher share of direct payers.
The same spend now returns 64% more day-8 revenue in the US. The account can bank that as efficiency, or spend it as headroom: scale Meta at the return it has today, or hold today's scale at a return the channel could not previously reach. Blinkist is moving its business-as-usual Meta campaigns onto the Churney signal region by region, retiring its own event as it goes, and extending the signal to new markets.
"The most meaningful day-to-day change so far is that I'm checking the split a bit more closely than usual in order to learn as much as possible about the signal, for example when it leaves the learning period, when it will show the first good ROAS signals."
Heiko Mandler — Performance Marketing Manager, Blinkist
pLTV pays where value varies widely between users and early monetisation is sparse. It pays less where early transaction volume is dense enough for the platform to find high-value users on its own, and it optimises traffic quality, not product conversion, so pricing and paywall come first. Holger's own version is sharper.
"I would tell another founder that Churney is a no-brainer to test if two conditions are met. First, Meta spend is in the mid six figures a month, or on a trajectory to get there. Below that, the fixed cost of running it eats the efficiency the event gains back. Second, trials or monthly subscriptions are a significant share of transactions. That is where predicting which trialists convert and which subscribers renew generates real value. If a company sells mostly annual plans on direct purchase, I see less value conceptually. But I don't know what I don't know."
Holger Seim — Co-Founder and CEO, Blinkist
Your data warehouse has incredible value. Our causal AI helps unlock it.
