Selling eleven memberships badly, then four well

The situation#

The network sold well in person and barely at all online. Studios had concluded the online channel did not work for their category and had stopped promoting it.

There were three separable problems and the brief only mentioned one.

The catalogue. Eleven membership packages, accumulated over four years of promotions that were never retired. Several overlapped. Two were strictly worse than a third for every customer. Studio staff could explain the differences in person; a pricing page could not.

The checkout. Five steps, an account creation requirement before any price was visible, and a mobile experience that had clearly been designed on a desktop.

The billing. Recurring mandates failing at a rate nobody was monitoring, with failures surfacing weeks later as lapsed memberships that studios then chased manually.

The brief we were given was to rebuild the checkout.

What we did#

Weeks 1–2 — analysed the catalogue before touching the code. We looked at what people actually bought, in studio and online, and what they bought after a first purchase. Four packages accounted for the overwhelming majority of revenue. Two of the remaining seven had never been sold online at all.

Recommending they cut the catalogue was not what the client expected and it was the highest-return item on the list. A comparison page with four options converts; one with eleven produces a decision people postpone.

Weeks 3–4 — repackaged what remained. Four tiers with a single axis of difference that could be stated in one line each, and the promotional variants moved to time-bound offers on top of a stable base rather than permanent additions to the catalogue.

Weeks 5–9 — rebuilt signup and checkout. Price visible before any account requirement, mobile-first, two steps, and the first class bookable immediately after payment rather than in a separate flow the next day.

Weeks 10–12 — fixed recurring billing. Mandate setup hardened, retry logic with sensible backoff, pre-emptive notification before a card expired, and a dunning path that recovers a lapse rather than discovering it. Monitoring on failure rate, which had not previously existed.

Weeks 13–14 — instrumented acquisition. Spend reconciled against actual activated memberships rather than signup events, which changed the picture of which channels worked.

The uncomfortable recommendation#

Cutting seven products from a catalogue feels like removing revenue. The network's concern was that customers on retired packages would leave.

They were grandfathered rather than migrated, which cost nothing, and the packages simply stopped being offered. Online conversion improved immediately and in-studio conversion did not fall, because studio staff had effectively been selling four packages anyway and using the other seven as anchoring.

What we would do differently#

We fixed recurring billing last, and it should have been first.

The failed-mandate rate was silently undermining everything upstream — memberships sold online were lapsing at a materially higher rate than in-studio ones, which was part of why studios distrusted the channel. For the first two months after the checkout rebuild we were reporting improved signup conversion into a leaky bucket, and the net revenue picture was better than before but worse than it looked.

Sequencing retention plumbing before acquisition improvements would have given a cleaner read and a faster net result. We now check the failure and churn path before optimising anything that feeds it.

Where it went next#

The network moved to a retainer covering the platform and the acquisition cadence. Provisioning automation for new studios followed, along the lines described in franchise and multi-site.

The offer structure is now reviewed quarterly with a rule the founder imposed after this engagement — a new promotional package has an expiry date at creation, or it does not launch.