Acquisition & Channel Monitor · Brief 02
A promotion bought volume the book could not keep
Comparison-site channel · August–September 2025 · lead surge → activation → day 90
The one line you configure
August was the channel's best month for volume. Ninety days on, is that cohort still here — and does the channel's cost per activation look different once it is scored on who stayed?
In a nutshell
It is not still here. Lead volume through the comparison-site channel ran 1.55× its baseline in August, 55% of it on paid search behind one promotion — and the cohort that activated out of it is churning at 7.5% inside ninety days against the book's own 1.8%. That is 4.2× the book, and because the cohort is roughly a third of the month's activations it drags the blended 90-day early-churn to 3.8%, which is exactly the shape the pack watches for: volume up and early-churn up together is low-quality acquisition, not growth. The subscribers who did stay bill at 0.60× the book's first-90-day revenue.
At a glance
What moved
The month it looked best, against how that month's cohort aged
The bars are what the channel reported at the time; the line is what those same cohorts did by day ninety. They peak one month apart, which is precisely why a volume dashboard cannot see this and a cohort read can.
How a third of the month moves the whole number
About a third of the month's activations at 7.5% and the rest at the book's 1.8% is what produces a 3.8% blended rate — and it is why the channel reads as twice the blended figure rather than four times it. The blended number is the one on the report, and it is the one that hides this.
The September cohort at day ninety, against the book
| Measure | This cohort | The book | Read |
|---|---|---|---|
| 90-day early-churn | 7.5% | 1.8% | 4.2× — one in thirteen gone by day 90 |
| First-90-day revenue per subscriber | 0.60× | 1.00× | The stayers bill less too |
| Activation volume, August | 1.55× | 1.00× | The reason the month was signed off |
| Sustained window, 15 Aug – 31 Oct | 4.5% | 1.8% | Not one bad month — a bad promotion |
Two independent quality measures moving together — they leave sooner and they bill less while they are here — is what separates a channel with a bad month from a channel buying the wrong subscriber.
Why the answer only exists ninety days later — and why it is still worth having
On the day, this channel was the month's best performer on every measure a volume dashboard carries. The quality signal cannot exist until the cohort is ninety days old. What autonomy buys is that the read happens on the day it becomes possible rather than at the next channel review — while the promotion is still live and the partner bounty for that cohort is still recoverable.
So what
The promotion is still live and the cohort behind it is still inside the window where the partner bounty can be disputed. Scoring channels on activations alone signs this off as the best month of the year; scoring them on activations that survive ninety days prices it correctly. Both readings exist in the same data — only one of them was being run.
The points that matter
Volume and early-churn moved together
The pack watches for exactly this pairing. A channel whose volume rises while the cohort it delivers churns faster is not growing the base — it is renting it.
1.55×volumeThe blended number is the one that hides it
At about a third of the month's activations, this cohort lifts the blended 90-day early-churn from 1.8% to 3.8%. The report shows 3.8% and nothing shows 7.5%.
3.8%blendedEven the keepers are worth less
First-90-day revenue per subscriber runs at 0.60× the book. The cohort is not just smaller by day ninety, the part of it that remains bills less.
0.60×revenue / subRecommendation
Retire the promotion on the comparison-site channel, and re-score the channel on activations that survive ninety days rather than on activations. On that measure the August surge is worth about a third of what it was credited with, and the bounty on the September cohort is worth disputing while it is still in window.
Questions it already answers
Was August actually a good month?
On volume, the best of the year — 1.55× baseline. On activations that survive ninety days, worth about a third of that, because the cohort churns at 4.2× the book and the survivors bill at 0.60× its first-90-day revenue.
Is this one bad month or the channel?
The promotion, not the channel. Early-churn runs 4.5% across the whole 15 August – 31 October window and 7.5% at its worst, against a 1.8% book — the shape follows the campaign, and 55% of the surge came in on one paid-search route.
Why did the monthly report not show it?
Because the report carries the blended rate. A cohort at 7.5% that is about a third of the month's activations produces a blended 3.8% — elevated, unexplained, and easy to read as noise until it is split by the channel that caused it.