Cutting lab checkout down to one screen

When a practitioner ordered blood work, the patient walked the full e-commerce funnel: email, consent, catalog, plan review, cart, address, payment. Seven screens and nine clicks — like being sent to pick up a prescription and having to walk the whole department store to reach the pharmacy counter.

Labs plans converted at roughly 57–65% from sent to paid, while Rupa Health, the competitor Fullscript had acquired, converted at about 79% on a much leaner flow. A provider-recommended test isn't a shopping decision; the patient already decided in the appointment. Every screen after the email asked them to decide again.

We didn't touch checkout. Engineering's insight: build one standalone screen that collects exactly what a labs order needs and bypasses cart and checkout completely. No dependency on the checkout migration running in parallel, and if the experiment failed we'd delete one page, not unwind a replatform.

A new patient has an account their practitioner created but never touched — they get every field, once, inline. A returning patient mostly confirms: saved card, pre-filled profile. The email link itself is the authentication, so there's no login wall and the email field renders read-only.

I tucked the practitioner's note behind a View message button: checkout first, everything else secondary. Our EM pushed back hard — this is healthcare, not Stripe, and the practitioner's voice probably drives purchases. We shipped it hidden, agreed to test open-by-default later, and I still don't know who's right.

It went live June 5 at 5%, then 20%, then 50%, and the first experiment fell apart. Two edge cases bounced treatment patients back into the standard flow, a parallel checkout migration muddied every comparison, and my own QA found multi-practice patients landing in the catalog instead. On July 10 we shut off the experiment — not the feature, the measurement.

We relaunched July 21 as a clean 50/50 and sat through three weeks of "it's not doing poorly, it's not doing well." Then one fix: the step for adding a card wasn't reading as a step, and every new patient stalled right at the money moment. Making it an explicit action moved conversion 3–4 points on its own.

Mobile is the queued half, and the platform has since sized it: roughly 4,500 lab orders a month still walk the seven screens, worth another 10–15% conversion if the same pattern lands there. The footnote I wrote at GA turned out to be the next bet.

Speed to the purchase moment is the lever, and the strongest move was subtraction. But months of flow architecture swung less than one button reading wrong — and we nearly judged the whole idea on a broken experiment. Hold your read loosely until the instrumentation deserves your confidence.