On this page
Case study
Dylan Karaitiana 15 Apr 2026 4 min read The operator ran a $4M GMV BigCommerce store. Cart-to-purchase conversion sat at 32% and had been static for three quarters. Paid acquisition was performing in line with category benchmarks. The market wasn’t getting harder; the surface was getting older.
The starting point
The operator came to us with a project brief: “redesign the checkout.” We pushed back and asked for a Map first. The answer to “redesign the checkout” might be “yes” — but the brief assumed the experience part of the operating system was the gap, and we wanted to verify that before pricing a build.
What the Map found
Three findings. The experience part of the operating system was underperforming, but not where the operator thought. The build pillar was fine — the Stripe integration, the inventory sync, the order fulfilment all worked. The demand pillar was strong; paid CAC was healthy and trending the right way.
The gaps were specific. Two of them.
First: the payment-method picker hid Apple Pay below the fold on mobile. 38% of the operator’s traffic was on iOS. Apple Pay conversion is meaningfully higher than card-entry conversion for that audience, and the user wasn’t seeing it without scrolling.
Second: shipping cost was revealed two clicks too late. The customer hit “checkout”, entered an address, hit “continue”, and only then saw shipping cost. By that point they’d invested four steps and were leaving anyway when shipping was higher than expected.
Both were experience-side gaps fixable in the build without a redesign.
The project
Eight-week project, fixed price, with the metric named: lift cart-to-purchase from 32% to 38%, sustained across three weekly cohorts.
Week one — Starting numbers & instrumentation
We set the starting numbers on the funnel rigorously before changing anything. Server-side events on every checkout step. Cohort definition: customers who hit the cart page and either completed or abandoned within 24 hours. We ran two weeks of pre-build measurement so the drift band was defensible. The 32% number had a ±2.1% week-over-week range; we needed to clear that band by enough margin to attribute the lift.
Weeks two–five — Build
Three changes delivered in sequence on a fortnightly go-live cadence.
The first go-live moved the payment-method picker above the fold on mobile, with Apple Pay as the first option for iOS traffic and Google Pay as the first option for Android. Card entry stayed available, just below the wallet options. Saw a 4% lift in mobile cart-to-purchase within the first cohort.
The second go-live moved shipping cost to the cart page. Customer sees the total before they enter checkout, not after. This was the change we were most nervous about — bringing a cost forward usually has a short-term abandonment cost before the long-term gain. The cohort data showed the opposite. Cart-to-purchase lifted 6%, presumably because the customers who abandoned at “shipping reveal” were going to abandon anyway, and the ones who stayed were no longer abandoning later in the funnel.
The third go-live added one-click reorder for repeat customers. A customer who’d purchased before could re-order their last basket from a single button on the cart page. This was a smaller cohort effect (2.5% lift) but compounding because repeat-customer share grew through the project.
Weeks six–eight — Verification & handoff
Three weekly cohorts after the last go-live, all clearing the 38% target. Week six: 38.4%. Week seven: 38.6%. Week eight: 38.4%. Drift band tightened to ±1.4%, well clear of the original starting range. We declared the metric moved, ran the handover, closed the project.
The result
The headline: $720K of annualised GMV recovered against a $14K project. The denominator on that calculation is the operator’s actual GMV trend, not a hypothetical. We checked again at the 90-day post-project mark — the lift held. Cart-to-purchase was 38.7% at day 90.
What we didn’t do
We didn’t redesign the checkout. The visual design didn’t change. The operator’s brand colours, fonts, photography, copy — all the same. The wins were structural, not aesthetic.
We didn’t run an A/B test. The operator’s traffic volume made A/B inconclusive within the project window; we delivered to 100% with rollback rehearsed and used pre/post cohorts for measurement. Some operators prefer A/B; we’ll do it when volume supports it. For this project it didn’t.
We didn’t recommend a full re-platform. BigCommerce was working; the build pillar was solid. Re-platforming would have been a 12-month project for marginal gain. Pillar thinking kept us focused on the actual gap.
What came next
The operator asked us to map the demand pillar next. CAC was healthy but they wanted a deeper attribution audit. That’s a separate project, in flight now. The experience-side go-live stays in production, the operator owns the launch plan for any future tweaks, and the relationship continues per the pillar-by-pillar model.
The takeaway
Sometimes the project the operator asks for isn’t the project they need. A Map up front saved this operator from buying a checkout redesign they didn’t need. Two surgical go-lives later, the experience part of the operating system holds, and the relationship is on pillar two of a multi-pillar cadence.