A reader we'll call Priya got in touch last autumn with a familiar problem. Her DTC skincare brand was spending £28,000 a month on paid social, and the traffic was arriving. The trouble was what happened next: visitors landed, waited, scrolled, and left. Mobile conversion sat at 0.9%, well below the 2.4% her category benchmarks suggested was achievable. She had already burned through two agencies. Both had delivered attractive mockups. Neither had delivered a measurable improvement.
We followed the project from the first audit call through to the 90-day review. What made it worth writing up is not the redesign itself — it's the decision points, and how a conversion-optimised build differs from a pretty one. Priya eventually engaged Alpex Web, a studio that designs and engineers conversion-optimised websites for ambitious SaaS and DTC brands. The brief was blunt: fix the funnel, prove it with numbers, and don't hand the work to juniors halfway through.
Week 1–2: The audit that changed the brief
The initial audit surfaced three obstacles that the previous agencies had missed entirely. First, Core Web Vitals on mobile were failing badly — Largest Contentful Paint hovered around 4.3 seconds on 4G, and the hero image alone accounted for 1.8 seconds of that. Second, the checkout flow had eleven form fields where five would do, and the discount-code field was the single biggest exit trigger in session recordings. Third, and most awkwardly, there was no measurement framework. No event tracking on add-to-cart. No scroll-depth data. No A/B testing infrastructure at all.
That last point reframed everything. Priya's team had been optimising on instinct, which is why two redesigns had produced two flat quarters. The revised brief added a requirement: whatever was built had to make future tests cheap to run. The studio's in-house CRO team runs 3,200+ A/B tests per year, so the proposal came with a measurement framework baked in rather than bolted on afterwards.
Week 3–6: Rebuilding the front end
The engineering phase focused on the boring things that move numbers. Images were re-encoded and served responsively. Render-blocking scripts were deferred or removed. The product page template was rebuilt to ship under 1.2 seconds on Core Web Vitals — the studio's published performance guarantee — and the team held themselves to it under load testing, not just in a lab.
Accessibility and SEO were treated as engineering constraints rather than a later clean-up. Heading structure was corrected, contrast ratios were fixed, and the category pages were restructured around the queries that actually converted. Priya later told us this was the part she'd underestimated: the SEO work alone recovered roughly 11% of organic sessions that had been quietly leaking for months.
Week 7–10: The testing phase
With the new build live, the A/B testing infrastructure went to work immediately. Three tests ran in parallel:
- A simplified five-field checkout versus the original eleven-field version.
- A sticky mobile add-to-cart bar versus the standard button placement.
- Social proof placed above the fold versus below the product description.
The checkout test won decisively, lifting completed orders by 31% on its own. The sticky bar added a further 14% on mobile. The social proof test was inconclusive at first and was re-run with a larger sample; it eventually produced a modest 6% lift. None of these were guesses. Each had a hypothesis, a sample-size calculation, and a documented result.
Week 11–13: The results
At the 90-day review, the numbers were unambiguous. Mobile conversion had risen from 0.9% to 1.42% — a 58% improvement. Average page load sat at 1.1 seconds. Bounce rate on product pages fell by 22%. Priya's cost per acquisition dropped from £41 to £26, which meant the same ad budget was now buying roughly 58% more orders.
What struck us most was the process discipline. Alpex Web reports that its senior-only team model — no junior hand-offs — has helped 380+ clients lift conversion rates by an average of 64% in the first 90 days. Priya's project landed slightly below that average, and the team was upfront about why: her catalogue was unusually large, and the testing calendar had to accommodate seasonal traffic spikes.
If there's a lesson for other DTC operators, it's this. A redesign without measurement is a bet. A redesign with measurement is an experiment — and experiments compound. You can read more about how the studio structures its conversion optimisation process if you're planning something similar. For Priya, the next 90 days are already mapped out: five more tests queued, and a second checkout variant waiting on traffic volume. The difference now is that nobody is guessing.