I've been running subscription P&Ls and looking at other people's subscription P&Ls for long enough now to have developed a fairly specific frustration: most operators benchmark their churn against the wrong number. They compare themselves to an industry average that lumps replenishment, curation, and content models together, then conclude they're either crushing it or doomed when they're really just looking at a category mismatch.

This is an operator-perspective piece. The data below comes from a mix of operators we've spoken to across categories, plus a sample we pulled from publicly available subscription-economy reports cross-referenced with what private operators told us looked right. Treat the numbers as directional benchmarks — the spread within each category is wider than the spread between categories.

The three models, briefly

For the purposes of churn benchmarking, the meaningful split isn't "DTC subscription versus everything else." It's the underlying model:

  • Replenishment — the customer subscribes to receive a consumable they were going to buy anyway. Razors, vitamins, pet food, coffee, contact lenses. The subscription is a convenience and price-lock wrapper on a product the customer has already validated.
  • Curation — the customer subscribes to receive a discovery experience. Beauty boxes, snack boxes, fashion-styling services. The product changes each cycle and part of the value is the surprise.
  • Content bundles — the customer subscribes to receive a recurring physical product tied to a content or community experience. Book clubs, hobbyist kits, fitness equipment with content layers.

These three models have structurally different churn curves. Treating them as one category is how you get benchmark numbers that nobody can actually use.

What the 2026 numbers look like

Monthly logo churn, after the first 90 days:

  • Replenishment: best quartile runs 3-5% monthly, median around 7-9%, bottom quartile north of 12%.
  • Curation: best quartile runs 6-9% monthly, median around 11-14%, bottom quartile above 18%.
  • Content bundles: best quartile runs 4-7% monthly, median around 9-12%, bottom quartile above 16%.

First-90-day churn (the killer number):

This is where the categories diverge most. Replenishment best-in-class operators report 15-25% cumulative churn through the first three billing cycles. Curation operators routinely see 30-45% in the same window. Content bundles fall in between, typically 25-35%.

The first-90-day number matters more than steady-state monthly churn for one specific reason: it determines whether your CAC payback model actually works. A subscription business with a $60 CAC and a $25 average monthly contribution margin needs the customer to make it through roughly three months to recover. If your first-90-day churn is 45%, your blended payback math is much worse than your monthly churn number suggests.

What separates the best quartile from the median

The interventions that consistently distinguish top-quartile churn performance, according to operators we spoke to, are not the obvious ones.

The obvious ones — cancellation flow optimization, save offers, win-back campaigns — do work, but they produce maybe 10-20% of the gap between median and top quartile. The bigger drivers cluster in the first 30 days:

  • Onboarding that sets the next-cycle expectation explicitly. Top operators tell customers, at the point of first purchase, exactly when the second box ships and what's in it (for curation) or exactly when the auto-renewal fires (for replenishment). The customers who churn fastest are the ones surprised by their second charge.
  • Skip-a-month as a default option, not a save offer. Replenishment brands that built skip-a-month into the standard flow report meaningfully lower involuntary-churn-leading-to-active-cancellation rates than brands that treat skip as a retention play.
  • Time-to-second-purchase optimization. For curation models specifically, the customers who engage with their first box within 7 days churn at roughly half the rate of customers who don't. Brands that optimize for that first-box engagement (push notifications, social sharing prompts, content tied to the box) outperform on retention even when their product mix is comparable.

Where the spread comes from

The honest answer to "why is the spread so wide" is that most subscription businesses are still running on assumptions baked in during 2020-2021, when the underlying customer behavior was structurally different. The customers acquired then had higher tolerance for the friction of a subscription relationship. The cohorts acquired since 2023 are noticeably less patient: faster to cancel, more sensitive to perceived value gaps, more likely to treat the subscription as a trial than a commitment.

Brands that updated their churn models for the new cohort behavior — typically by assuming first-90-day churn rates 5-10 points higher than their 2021 baseline — have stayed solvent. Brands that didn't are either now in trouble or have already restructured.

The takeaway

Benchmark against your model, not against "DTC subscription." Then benchmark against your model's top quartile, not its median, because the gap between best and median is where the actual money is. And update your assumptions every twelve months — the customer of 2026 is not the customer of 2021, and the model that worked then is not the model that works now.