Shell page. Structure, masthead and at-a-glance are final. Body copy is in the markdown draft; screens and the items below are still needed. Delete this block when the page is written.
- What the “smarter sticky banner” did differently, and what the product exploration component was
- What you were betting each of the three changes would fix
- Whether premium desktop losses concentrate by product
- Duration and team shape
- Screens: template + three tier applications, above-the-fold per tier, exploration component, sticky banner states
One template. Three verdicts.
| Tier | Direction | Probability read |
|---|---|---|
| Business | Up — best product a few points | 72% the redesign is best |
| Premium | Low single digits down, isolated to mobile | 99% control is best |
| Consumer | Down a few points | 97% control is best |
Read as a portfolio average, this is a wash with a slight negative tilt. Read as three results, it's the most useful thing I've worked on.
The hypothesis had an assumption buried in it
The stated hypothesis took for granted that “friction” and “appeal” mean the same thing on every product page in the portfolio. That assumption is the actual subject of this case study, and it was invisible to me when I wrote it.
Business shoppers convert when they scroll. Premium shoppers convert when they don't.
Business — the body is the dealbreaker. People who scrolled into the page body converted at roughly twice the rate of those who didn't, and the gap held in both control and test. A card journeys study independently found business shoppers naming lack of clarity and detail as a top pain point.
Premium — won or lost in the first viewport. Exactly inverted. The people who didn't scroll converted at more than twice the rate of those who did — in both control and test, so it's a property of the audience rather than an artifact of either design. And a premium landing page study found those shoppers reported their trust came from the brand, not the page.
Scrolling isn't engagement on a premium page — it's hesitation. The page's job is to confirm a decision in the first viewport and then get out of the way.
A template is the right abstraction for pages that share a reading behaviour
Card tiers don't share one. So the output shouldn't have been one template — it should have been a template family, with an explicit tier-specific contract about what lives above and below the fold.
Before unifying pages under a template, check whether the audiences read them the same way. If they don't, consistency of component is fine and consistency of architecture will cost you conversions.
What I'd flag
The limitations I'd raise before someone else did.
- Two tests stopped early and nobody wrote down why.
- My recollection is we had enough signal to make a call — a legitimate reason to stop — but it was never documented, and an undocumented pause is indistinguishable from a test that broke. That's the precise failure mode the catalog's Inconclusive-versus-Untested distinction exists to prevent, and this initiative is one of the cases that taught me it was necessary.
- Three changes shipped together on every tier.
- So I know the bundle's direction per tier and can't tell you which component drove which effect.
- I designed the assumption in.
- Nothing forced the portfolio-constant framing; I didn't interrogate it, and the scroll data was available in principle before the test ran.