480,000 sessions. Three friction points. Two sprints. Zero architecture changes. The checkout didn't need to be rebuilt — it needed to be measured.
Cart Abandonment
Before → After
Relative Reduction
Two-sprint improvement
Mobile CVR
Mobile checkout conversion
Sessions Analysed
GA4 + Clarity data set
Cart abandonment was sitting at 73.1% — a significant leak in the conversion funnel. The business knew the number but not the why. No funnel instrumentation existed below the product detail page. The default Shopify analytics showed a flat abandonment rate with no step-by-step breakdown, making it impossible to identify where users were dropping off or why.
Three months of GA4 data existed. Nobody had configured custom events for the checkout flow. The team was optimising based on aggregate numbers and anecdotal support tickets — trying to fix a leak they couldn't see.
Baymard Institute's large-scale checkout research identified 8 critical UX violations in the current flow — overlapping with the exact friction points customers were quietly abandoning over. The data existed. It just hadn't been connected to Sierra's specific implementation.
I built the measurement layer the checkout had always been missing.
Configured GA4 custom events across every checkout step — cart view, shipping information, payment information, order review, purchase. No checkout had ever been instrumented below the 'added to cart' event.
Layered Microsoft Clarity heatmaps and session recordings on top. Watched 50+ full checkout sessions across desktop and mobile. The patterns were consistent and unmistakable.
Cross-referenced Clarity rage-click data with GA4 event drop-off. Three distinct clusters emerged within two weeks of instrumentation: shipping estimate abandonment, promo code rage clicks, and payment error exits with no recovery path.
What looked like 'high cart abandonment' was actually three separate problems happening to different users at different points. Treating them as one problem would have fixed nothing. Each required a specific, independent fix — and they didn't need a checkout rebuild.
Three targeted fixes. Two sprints. Each validated independently.
Fix 1 — Shipping cost on the cart page. Users were reaching the shipping estimate step and seeing the full cost for the first time. Added transparent shipping cost display on the cart page so users saw the complete total before entering checkout.
Fix 2 — Promo code validation. Users were rage-clicking a non-functional promo code field. Fixed inline validation with error messaging that told users exactly what went wrong — wrong format, expired code, or ineligible category.
Fix 3 — Payment error recovery. Users were abandoning after a generic "something went wrong" message. Replaced with specific copy referencing card type, expiry, or network decline — and added a visible retry path.
Shipping Display
Before checkout entry
Promo Validation
Inline error messaging
Payment Recovery
Specific retry path
Cart Abandonment
73.1 → 53.9%
Mobile Conversion Rate
+47%
Checkout Completion
+49%
Sprints to Ship
2
No single-page checkout rebuild. No platform migration. Just instrumentation followed by targeted fixes.
The checkout template stayed the same. What changed was the execution quality of every step. The fixes outperformed Wayfair's checkout on mobile conversion within 30 days of the final sprint — a benchmark nobody had thought to measure against.
The checkout didn't need to be rebuilt. It needed to be understood.
The three problems looked like one because the data only showed the aggregate. Step-level instrumentation was the unlock — without it, we were trying to diagnose a three-part fracture with a single X-ray. Always instrument before you diagnose.
Session recordings should have started earlier. We had 480K sessions of data we couldn't use because nobody had turned on recordings. Two weeks of recordings gave us more actionable insight than three months of aggregate dashboards.
Baymard's research was sitting in a folder for months. Nobody had mapped their 8 critical violations to our specific checkout. External UX benchmarks are only useful when you test them against your own implementation. I'd do this mapping earlier next time.
Read the full breakdown
How I took checkout abandonment from 73.1% to 53.9% — a complete teardown with the numbers, the three friction points, and the fixes.
Read the blog post