Why subscription brands overspend when they optimise for ROAS

A coffee subscription is sold on the promise of the tenth box, but ROAS only sees the first. A campaign wins a subscriber, Google Ads records one order, and every renewal after that arrives without an ad click. The bid that won the subscriber was set on an assumption about how long they'll stay. The gap between ROAS and POAS is the gap between that assumption and what actually happens.

The figure above shows what actually happens to renewals. In Recharge's subscription data, covering July 2025 to June 2026, 20.7% of coffee and tea subscription renewal cycles ended in active churn: the subscriber cancelled. That was the joint highest of the five consumable categories Recharge compared, level with beauty. This figure covers coffee and tea subscriptions, not curated subscription boxes, which Recharge didn't break out.

How the gap forms

The first box is not the customer

Coffee's early numbers look excellent. In the same data, 99.3% of first-time coffee and tea subscribers completed a first reorder, the highest of the five categories. But only 9.3% of the cohort reached a sixth reorder, and 1.3% a twelfth. Coffee subscribers don't leave at the first box; they leave over the months that follow, which is exactly the period a payback calculation depends on.

Discounted first boxes

A heavily discounted or free first box wins subscribers cheaply, and ROAS records the win. But the discount means the first order earns little or nothing, so the whole acquisition cost has to be paid back by renewals. If subscribers attracted by a deep discount cancel faster, the payback never comes.

Renewals counted as new conversions

If your subscription app sends renewal orders to Google Ads as conversions, the account can credit campaigns for revenue they didn't win this month. ROAS rises, and the bidding learns from customers who were already subscribed.

Working out your own gap

• Send only first orders to Google Ads as conversions, and track renewals in your subscription app.
• Build a cohort curve for each month's new subscribers: the share still active after one, three and six renewals.
• Compare cohorts by first-box offer. Subscribers won with a deep discount can behave differently from those who paid full price.
• Read cancellations by renewal number. If cancellations cluster right after a discounted first box, the offer is attracting the wrong subscribers. If they come later, the coffee, the price or the delivery cadence is the problem, and no bid change will fix it.

How to close it

1. Set acquisition targets from your own cohort payback, not from an assumed lifetime value.
2. Limit the first-box discount to what your cohorts show you can earn back.
3. Offer skip and pause at cancellation. Recharge's data shows coffee subscribers skip 4.0% of renewal cycles; every skip that replaces a cancellation keeps the payback going.