[Product name]
[price]
12 people are viewing this now
Social proof
+2.50% RPV
- Trigger
- Real-time visitor count on the product page
- Shows
- One plain line of text near the price
- Hidden
- When the count is low enough to be meaningless
On this page
Evans: do small wins compound?
Six small experiences across the shopping journey, each tested on its own and judged on one metric: revenue per visitor.
+2.5% to +8.43%
revenue-per-visitor uplift, each experience in its own test
01 · The question
Many small bets, or one big redesign?
Evans didn’t need one dramatic fix. The question was what many small, honestly measured improvements would add up to.
1 · Conversion rate wasn’t enough
More orders can still mean less money if baskets shrink. Revenue per visitor (RPV) catches both.
2 · An audience that punishes complexity
Most Evans shoppers weren’t digitally native. Clutter, pop-ups and fake countdowns don’t just underperform here; they break trust.
3 · The big-bang temptation
One redesign concentrates risk and muddies measurement. Small experiences each carry their own case, test and number.
02 · The portfolio
Six experiences, each scored on its own
Every experience ran as its own test against a control and kept its place only if it earned an RPV uplift. There was no blended number to hide weak ideas behind strong ones.
each bar is its own test — these don’t add up
03 · Interaction patterns
Each pattern has a rule, not just a look
Every experience is a small, reusable component with three parts: a trigger, what it shows, and when it stays hidden. The last part mattered most. If a signal isn’t true, the component doesn’t render.
Selling fast
[Product name]
Size 18 selected
“Selling fast” badge
+3.34% RPV
- Trigger
- Stock in the selected size falls below a threshold
- Shows
- A text badge, no timer, no flashing
- Hidden
- When stock is healthy
Popular · top 200 today
[Product name]
[price]
Top-200 popularity
+8.43% RPV · best result
- Trigger
- Item is in the top 200 added to bag over the last 24 hours
- Shows
- A popularity label: local, recent and true
- Hidden
- When the item drops out of the top 200
Size 18DressesUnder 50
[Product name]
[price]
Mobile experience
+5.62% RPV
- Trigger
- Visitor on a mobile device
- Shows
- Simpler listings and filters, with large tap targets
- Hidden
- On desktop, where the full layout fits
Sizing welcome banner
+2.50% RPV
- Trigger
- First visit to the home page
- Shows
- A short banner explaining how sizing works
- Hidden
- For returning visitors, and once dismissed
Postcode
SW1AEnter a full UK postcode, like SW1A 1AA
Checkout friction removed
+3.19% RPV
- Trigger
- Every checkout session
- Shows
- Fewer fields, clear input rules, errors next to the field
- Removed
- Non-essential fields
Recreated specimens with placeholder products, built to show each rule. They are not the original Evans screens, and message wording is illustrative.
04 · Honest by design
Persuasion that’s true, and readable
For an audience that isn’t digitally native, manipulation doesn’t just fail; it costs trust. So the rules doubled as accessibility rules, checked against WCAG 2.2 AA.
| Rule | WCAG 2.2 | In practice |
|---|---|---|
| No countdown timers | 2.2.1 Timing adjustable | Urgency comes from real stock and demand, never a clock that pressures a slower shopper. |
| Badges are words, not colour | 1.4.1 Use of colour | “Selling fast” and “Popular” carry meaning in text; colour only supports it. |
| Live counts aren’t announced on every change | 4.1.3 Status messages | The visitor line updates quietly instead of interrupting a screen-reader user. |
| Errors next to the field, in plain words | 3.3.1 · 3.3.3 Error identification and suggestion | The checkout says what’s wrong and how to fix it, with an example. |
| Don’t ask twice | 3.3.7 Redundant entry | Removing non-essential fields also cut repeated information at checkout. |
| Big tap targets on mobile | 2.5.8 Target size (minimum) | Filter chips sized well above the 24px minimum. |
05 · Results
Small wins, honestly counted
+8.43%
Best single uplift: top-200 popularity messaging
6
Experiences, each measured on RPV in its own test
Millions
Several million pounds of added revenue, assessed at the revenue line (reported)
2.5 + 3.34 + 8.43 + 5.62 + 2.5 + 3.19 = 25.6% → not the program’s lift
The uplifts don’t add up. The experiences overlap in audience and page, their effects interact, and each was measured on a different group of visitors. The program’s contribution was measured at the revenue line instead.
Experimentation within the Arcadia Group for Evans. Proprietary metrics generalised for confidentiality. Arcadia’s measurement standard applied: investments are judged on incremental value, meaning revenue that wouldn’t have happened anyway.
06 · Reflection
What I’d tell the next team
Portfolios beat big bangs
Many modest, measured wins added up, with less risk and faster learning than one redesign. It took team alignment, governance and a steady cadence, not just code.
The metric is the strategy
Choosing RPV early changed what winning meant for every experience. Conversion can rise while revenue falls; RPV catches both.
Honest signals win twice
The best result came from the most truthful pattern: real, recent popularity. For this audience, truth was also the best persuasion.
Uplifts don’t sum
Knowing why you can’t add them up matters more than any single result.
If I ran it again
- Report each uplift with its confidence interval and sample size, not just the point estimate.
- Keep a global holdout, a small group who see none of the six, to measure the combined effect directly.
- Test the patterns with screen-reader and keyboard users, not just on revenue.