Grow the customer
Grow the customer
Your next growth is sitting in your customer list.
Winning a new customer keeps getting more expensive. The ones you already have are sitting in your own order history, and most brands barely look. We build the models that spot a customer drifting away before they're gone, prove the lift against a control group we leave alone, and hand the whole thing over running in your stack.
- How it runs today
- Hand-built segments, churn spotted after the fact
- What breaks
- By the time a customer looks at risk, they're already gone
- What you get
- Models that see it coming, proven against a holdout
The shift
The next dollar moved to retention.
The old way
Retention meant hand-built segments and email flows, with at-risk defined only after a customer had already gone quiet.
The AI-era shift
Acquisition keeps getting more expensive, so the marginal dollar of growth now comes from the customers you already have. And the tooling caught up. Models trained on your own order history can flag a customer drifting away weeks before it shows in their behavior, and tell you what to send each one, and when. That used to take a data team. It doesn't anymore.
You can see a customer leaving weeks before they've left.
What we actually do
The work, made concrete.
Retention & LTV diagnostic
We read your order history the way an operator reads a P&L: how each cohort behaves over time, what a customer is worth after costs, where the retention revenue is going. It ends in a ranked list of what to fix first, and what to leave alone.
Predictive retention build
Models built on your own customer data that score who's likely to leave, what they're worth, and when they'll buy next. Deployed into the tools you already run, documented, and handed over. You own it, nothing's locked to us.
Lifecycle rebuilt around the predictions
Winbacks that go out before the customer's actually gone, replenishment reminders timed to how each person reorders, and a nudge toward the second product at the moment it's most likely to land.
An automation pilot, scoped honestly
Where it makes sense, we let the system run the routine decisions on its own. We start with the low-risk majority, and everything is measured against a control group, so you know the lift is real.
Proof
The number the work carries.
Darwin's · Pet & DTC Food (subscription)
A fresh pet food subscription doing real revenue, with churn defined only after a customer had already gone quiet. Winbacks went out weeks too late to matter.
A three-week retention and LTV diagnostic, then a churn model built on their own order history and deployed into their lifecycle flows, with a holdout so the lift was proven, not asserted.
Monthly subscription churn.
The number this engagement carried, read early enough for a save to matter and proven against a holdout instead of asserted.
MUD\WTRPer-subscriber churn propensity on the brand's own first-party data, with replenishment timed to actual usage instead of a fixed calendar.
DuckhornRepeat purchase within twelve months, from first bottle to club membership, timed to actual buying rhythm.
The first step
We start inside your own order history: how your customer cohorts actually behave over time, what a customer is really worth after costs, and where retention revenue is leaking. You get a prioritized read before you commit to building anything.
What we move
What we watch on retention.
Benchmarks and targets, not guarantees. We baseline yours first.
How we work
How the engagement runs.
- 01
Diagnose
We baseline your numbers and map the operation end to end, so the work targets a real leak, not a hunch.
- 02
Prioritize
We rank the opportunities by dollars of impact and effort, and agree on what to do first.
- 03
Build
We build the real thing in production (for you, or alongside your team) against a measured baseline.
- 04
Prove
We hold the work against a holdout or benchmark, so the lift is proven, not asserted.
- 05
Hand over
Documentation, dashboards, and an accountable owner on your team, so the work keeps running without us.
Where this connects
Start with your own order history.
The diagnostic is the honest first step. We read your numbers, show you where retention revenue is leaking, and rank what's worth fixing. Then you decide whether to build, and whether it's with us.