From zero to 100+ live experiments in four weeks
With over a million monthly sessions but no CRO team to act on them, Reima had upside it couldn't reach. Dalton changed that: more than 100 experiments live in four weeks across five markets, profitable at scale, without a single developer hour.
“It would be very challenging to make it profitable to run A/B testing with traditional methods. Now we can do it effectively at scale.”
The Challenge
Reima, a premium children’s outdoor clothing brand selling across 30+ countries, had a conversion problem and no way to fix it. With over a million monthly sessions, even marginal improvements to conversion rates meant significant revenue. But their six-person e-commerce team had no in-house CRO resources, no developer bandwidth, and no realistic path to running experiments.
"We have people who could do A/B testing, but don’t have any roles dedicated specifically to A/B testing or conversion rate optimization." - Jonni
Traditional A/B testing was a non-starter. Each test required separate development work, took months to reach statistical significance, and demanded heavy coordination across five markets. Worse, Reima operates on tight seasonal windows: winter gear launches in October, spring collections in February. By the time a conventional test concluded, the season was over.
The Solution
Reima implemented Dalton with a single JavaScript snippet, a setup comparable to installing Google Analytics, completed in one meeting. Within four weeks, the team had launched over 100 active experiments across 49 pages and five markets, with zero developer hours required.
Dalton’s workflow is built for speed: the platform scans pages and generates ready-to-launch experiment variants for headlines, CTAs, layouts, and product copy. The team reviews and approves in minutes. A multi-armed bandit algorithm then routes traffic dynamically toward winning variants in real time, with no waiting for statistical significance or manual analysis.
The Results
The impact was immediate and measurable. Optimising product information banners to emphasise safety and sustainability credentials delivered a +21.1% uplift. Rewriting category headline copy, shifting from "Children’s transitional season shoes" to benefit-led alternatives, produced lifts of +19.9% and +17.8% across variants. Simplifying filter layouts and adjusting trust signals by region drove consistent gains across markets.
What Dalton tested
Those headline wins were the standouts, but the bigger story is the pattern across more than a hundred experiments in five languages: again and again, the plainest, most literal version won.
- Size guidance copy on product pages was the single most repeatable win, the same insight replicating market after market.
- Across headlines, banners and product copy, benefit-led and understated versions beat the polished, on-brand ones. The subtle USP banner outperforming the bold one is a typical example.
- Not everything worked, and that is the point. Rewriting Reima’s already-optimised SEO page titles backfired, and Dalton sidelined those variants on its own.
Why it worked
None of this came from a single big idea. It came from doing something a six-person team structurally cannot: testing constantly, across every market, at once. Where a lean team might ship a handful of tests a quarter and wait months for each to read out, Dalton ran more than a hundred in four weeks and let real shopper behavior, not internal opinion, settle every debate. That is why the plain, literal copy kept winning. A brand team would have argued much of it out of existence, but Dalton carries no such bias and simply keeps whatever converts. And because the bandit shifts traffic toward winners and away from losers in real time, a weak idea costs a sliver of one experiment instead of a place on the roadmap. The payoff is leverage no lean team could buy with headcount: many cheap, fast, unbiased experiments compounding week after week, with the learnings from one seasonal collection carried straight into the next.
Reima went from talking about testing to running it at scale, across 30+ countries, with no developer dependency, in under a month.