Grow

Wave 5

Grow.

Grow is the retention wave: the second purchase and every one after it. Acquisition gets the meetings and the budget, and retention quietly pays for the company. Move repeat rate and the economics change more than most acquisition wins ever manage, because the revenue arrives without an acquisition cost attached. This wave owns everything after the first order: cohorts, subscription mechanics, lifecycle timing, the cancel flow, winbacks, and the advocacy loop that turns kept customers into found ones.

The core problem is almost always timing. Churn tends to get defined after the customer has already gone: "cancelled" fires weeks after the real decision, and the winback goes out to someone who finished leaving a month ago. Meanwhile the signals were sitting in the order history the whole time, the stretched reorder gap, the skipped shipment, the streak of unopened emails. Most retention programs fail on timing, not persuasion.

The other problem is honesty. Almost any flow "works" if you credit it with every customer who was staying anyway, which is why this wave carries the method's hardest rule: every model and every flow that claims a lift gets measured against a holdout, a group of customers who didn't get the treatment. The lift is proven, or it isn't claimed.

Catch the signal while there's still a decision to influence.

Grow versus the waves next to it

What good looks like.

Grow ships six things, and each one moves a number a founder already watches: churn, repeat rate, what a customer is worth over time, and how much of your revenue comes from people who came back.

1. Churn that fires early

A churn definition built on behavior instead of paperwork: reorder gaps, skipped shipments, engagement fading, computed per customer on your own order history. The signal usually shows up weeks before the cancel button gets clicked, and that window is the difference between a conversation and a goodbye.

2. Lifecycle timed to usage, not the calendar

Replenishment emails get timed to how fast this customer actually goes through the product instead of a fixed monthly drum, post-purchase flows know a first-time buyer from a fifth-time one, and the subscription cadence can stretch without the customer leaving, because the real alternative to a flexible skip is usually a cancel. Fixed-calendar lifecycle work reads as spam precisely because it ignores the one thing you actually know: the customer's own rhythm.

3. A cancel flow that knows why

Most cancel flows treat every leaving subscriber as the same emergency, which is why most save offers feel like a toll booth. The rebuilt flow tells them apart: too much product gets a pause and a slower cadence, price sensitivity gets a right-sized plan, and a product complaint gets a human. Some cancels should just be cancels, handled gracefully, because a save rate built on trapped customers comes back later as refunds and reviews.

4. The second-purchase path

For brands that aren't subscription-led, the same discipline applies to the gap between order one and order two: cohort work to find where repeat actually dies, then fixes aimed at that exact spot. Sometimes it's timing, sometimes it's a category gap, sometimes the first product simply doesn't lead anywhere, and the cohort table will tell you which. When more customers come back, everything you spend on acquisition above this wave works harder.

5. An advocacy loop with one owner

Reviews captured from verified buyers while the experience is fresh, a referral program sized to what a cohort is actually worth, and all of it structured so machines can read it, because the same reviews that convince a human on the page are evidence to the engines assembling the shortlist. Advocacy usually fails as three orphan programs. It works as one loop with one owner.

6. Every claim against a holdout

Declared in Truth and enforced here, because this is the wave with the strongest temptation to take credit for customers who were staying anyway. Every flow, every model, every save offer reports its lift against a control group, and the number that survives is real: churn moved because of the thing we shipped, and the holdout shows it.

How it runs. Take a brand we'll call Marlowe, subscription-led, with a base that looked healthy until Truth redefined churn on order gaps and the cancels stopped looking sudden: the signal had been there for weeks each time. Grow put a score on every subscriber for how likely they were to drift, rebuilt the cancel flow around reasons, and re-timed replenishment to actual usage, all of it running against a holdout, so when the room reads the churn number it can also see what would have happened without the work. That comparison is what changes the conversation.

The threads, in this wave.

Woven in.

The models here are quiet and specific: how likely each subscriber is to leave, how fast each customer actually uses the product, what a cohort is likely to be worth, all trained on your own order history rather than someone else's benchmarks. The skip list is just as specific: no personalization platform before the basic flows exist, and no tool whose lift can't be measured against a holdout.

Your team.

The flows live in your email platform, the models train on your data and run in your accounts, and your lifecycle lead ends the engagement owning the score, the segments, and the cadence, with every definition documented back in the Truth page. Nothing routes through us to keep working.

The coda.

Keep them longer, prove it against a holdout, and the whole P&L above this wave gets easier.