Step One Life  ·  US · AU · UK  ·  Dec 2025 – Jan 2026

One message,
three markets

The same PDP banner treatment — “X+ Million Sold” replacing a generic USP strip — run as three parallel tests to find out whether a social-proof claim travels. US: +28.1% conversion at >99.99% confidence.

Role

CRO Lead + Ecom Ops — sole owner: hypothesis → build → run → analyse → ship

Programme scale

3 tests, one treatment, three storefronts, ~110k visitors

Primary metric owned

Conversion Rate, with RPV as the commercial guardrail

Status

US rolled out on a definitive result. AU shipped mobile-first with a desktop watch-list. UK held as directional.

01The Problem

Step One had no social proof on the product detail page above the fold. The brand had sold millions of pairs, and none of that scale was visible at the moment a customer was deciding whether to buy. For a category where the purchase is a considered, somewhat private decision — fit, quality, and "will this actually be good" are real objections — the absence of any validation signal was a gap.

The hypothesis was deliberately simple:

If we brag about how many pairs of underwear we have sold, then we will increase conversion rate, because it will help validate the customer's purchase.

The treatment was a single banner on the PDP: "X+ Million Sold" replacing a generic USP banner. No layout change, no imagery change, no pricing change — pure messaging. That simplicity is what made it worth running in three markets: if a social-proof claim travels, it should travel everywhere.

---

02Method

Platform: Shoplift, 50/50 storefront tests. Statistics: two-proportion significance test on conversion rate and add-to-cart rate; zero-inflated test on revenue per visitor. 95% confidence threshold. Segmentation every read: device, new vs returning, channel. Design decision: the same treatment, translated but not redesigned, ran in all three markets — so the comparison is as clean as a cross-market comparison can be.

The three tests ran in overlapping windows (US from 11 Dec, AU and UK from 10 Dec), both through the Christmas trading period. That overlap matters for interpretation and I address it below.

---

03The Results — Same Treatment, Three Different Answers

US — US_Theme_CVR_251211 | 11 Dec 2025 – 5 Jan 2026

This is the definitive result. A +28.1% conversion lift at >99.99% confidence, with ATC rate up 23.3% and clickthrough up 10.2%. The treatment didn't just convert better — it moved people through the funnel earlier, which is the signature of a message that lands on arrival rather than one that rescues a hesitant shopper at the end.

AOV compressed 3.4%. That's the honest caveat: RPV still rose 26% because the conversion gain dwarfed it, but the treatment did pull in slightly lower-value orders. Worth noting that AOV compression alongside a big CVR lift is a pattern I watch for specifically — on paid channels it's often a warning sign rather than a win.

The moderate traffic-imbalance flag (51.3/48.7, >95% confidence) is real but immaterial: a 2.6% imbalance cannot produce a 28.1% effect at >99.99% confidence.

AU — AU_Theme_CVR_251210 | 10–23 Dec 2025

The aggregate hides the story. On the surface this is a mild directional win: +4.4% CVR, +3.3% RPV, revenue essentially flat. But the device split is decisive:

Mobile won on every metric. Desktop lost on the two that matter. The same banner that lifted mobile conversion 7% and mobile RPV 7.2% depressed desktop conversion 5.2% and desktop RPV 9.7%.

That divergence is the finding, and it shaped the recommendation: ship mobile globally, hold desktop pending a follow-up on banner size and placement. The most likely explanation is that on a small screen the banner reads as a compact, credible credibility marker, while on desktop it competes with existing page hierarchy and crowds information the customer was already using.

UK — UK_Theme_CVR_251210 | 10–23 Dec 2025

A directional result I reported as directional. The variant generated more orders from fewer visitors and 4.2% more revenue, with RPV up 5.3% and AOV up 3.0%. But CVR moved only 1.5% at a 68.7% probability of improvement — roughly a one-in-three chance the lift is noise — and the test never reached significance.

The revenue lift substantially outpaced the conversion lift, which suggests the treatment influenced higher-value segments or encouraged larger baskets rather than simply converting more people. Mobile again outperformed (CVR +2.0%, RPV +5.5%, AOV +2.9%) while desktop was flat-to-negative (CVR −0.2%, AOV +3.6%).

---

04The Cross-Market Comparison

The same banner produced a 28.1% lift in the US and a 1.5% lift in the UK. That spread is the case study.

Three factors plausibly explain it, and I'd want to test rather than assert:

1. Baseline conversion differs enormously. US baseline CVR was less than half that of AU and the UK, which were close to each other. A social-proof claim has far more room to move a low baseline than a high one — the US site was converting half as well to begin with, which suggests a less-optimised experience with more headroom.
2. The credibility bar differs. "X+ million sold" is a stronger claim in a market where the brand is less established. In AU, where Step One is a well-known domestic brand, the claim may be closer to assumed than persuasive.
3. Christmas trading was in all three windows, so seasonality is a shared factor rather than an explanation for the divergence — but it does mean none of these results should be treated as clean baseline behaviour.

The consistent thread across all three markets was mobile outperforming desktop, which held in the US, AU and UK reads independently.

---

05Key Insights

1. Cross-market replication is not confirmation — it's a second test with its own answer. +28.1% (US), +4.4% (AU), +1.5% (UK) from an identical treatment. If I had run only the US test I would have concluded social proof is a major lever; if only the UK, that it's noise. The honest reading is that the mechanic works and its magnitude is market-dependent — and the magnitude correlates with how much headroom the baseline has.

2. The aggregate can hide a segment-level failure. AU's +4.4% CVR aggregate contains mobile at +7.0% and desktop at −5.2%. Reading the aggregate would have produced a clean "mild win, ship it" — and shipped a change that degraded desktop RPV by 9.7%. Device segmentation changed the recommendation from "roll out" to "roll out mobile, hold desktop."

3. AOV compression is a flag, not a footnote. The US lifted CVR 28.1% while AOV fell 3.4%. RPV still rose 26%, so it was the right call — but a social-proof claim that pulls in lower-value orders is doing something different from one that converts the same customer. On paid channels I treat that pattern as a warning; here, with RPV decisively up, it was acceptable.

4. Statistical honesty is a position, not a hedge. The UK result had a real revenue story (+4.2%) and I could have presented it as a win on revenue. It didn't reach significance on its primary metric, so I called it directional with the 68.7% probability of improvement stated plainly. The US result — >99.99% confidence — needs no such caveat, and the contrast between the two is more convincing than either alone.

5. A one-element test is worth running in three markets precisely because it's one element. No layout, imagery, or pricing confound. Whatever difference emerged is attributable to the message and the market, which is why the 28.1%/1.5% spread is a genuine finding rather than an artefact of three differently-built pages.

---

06Business Impact

In plain terms: a single messaging change — replacing a generic USP banner with "X+ Million Sold" — delivered a statistically definitive +28.1% conversion rate and +26% RPV in the US, and a directional mobile win in AU that I shipped mobile-only after the device split showed desktop regressing. The UK result was a real revenue improvement that never reached significance, and I reported it as directional rather than dressing it up.

The programme's most useful output is the cross-market spread itself: it tells the business that social proof is worth deploying everywhere but is a major lever only where baseline conversion has headroom — which reframes the US opportunity as an experience gap, not a messaging win.

---

07What I'd Do Differently / Next

• Run the desktop follow-up in AU. The −9.7% desktop RPV is unexplained and unresolved. Test banner size and placement specifically on desktop to establish whether the problem is visual weight or message scepticism.
• Test the credibility threshold directly. The US/AU/UK spread correlates with baseline CVR, but I never tested the claim's form across markets. A localised proof point ("Australia's most worn underwear" vs "X+ million sold") would test whether the mechanism is scale-credibility or local relevance.
• Avoid Christmas windows for messaging tests. All three ran through December trading. The divergence is too large to be explained by seasonality, but a clean-window replication in at least one market would remove the doubt entirely.
• Pre-register the device hypothesis. Mobile outperformed desktop in all three markets independently — which I only noticed after the fact. That's a pattern I should have predicted, stated as a hypothesis, and tested rather than discovered.
• Segment the rollout decision from the first read. The AU "roll out mobile, hold desktop" recommendation was correct but arrived at the end. Device-level guardrails should be built into the decision framework before the test runs, not applied after the split appears.

---

Sources: the Step One experiment log and Shoplift test exports. Revenue, traffic and baseline figures have been generalised; relative changes and test statistics are as measured. Programme IDs: US_Theme_CVR_251211 · AU_Theme_CVR_251210 · UK_Theme_CVR_251210. Note: US_Theme_CVR_251211 carries a moderate traffic-imbalance flag; AU_Theme_CVR_251210 a minor one. Neither is material to the effect sizes reported.