Verdict: move the closing paragraph above the first section heading. One paste, this weekend, before anything else on this list.
Everything below explains why that paste outranks the rest, and what to do after it.
Nothing has shipped since July 24. I diffed the served strings against the July 24 capture and they match line for line, which I trust further than the July 17 Last-Modified header, because that header sits behind a CDN and tells you about cache state as much as content.
Context if you're coming to this cold. Allē is Allergan Aesthetics' loyalty program for cosmetic injectables, and the case covers both of its faces in one scope: the consumer iOS app members carry, and the backoffice dashboard practices run. Inside the portfolio, this case does two jobs no other case does. It's the multi-surface architecture credential and the consumer-scale credential, the only work you have that proves you can hold two audiences inside one model.
Six seconds on the page gets an evaluator the scope. It does not get them the system.
The track structure argues against your thesis
The spine as published:
eyebrow → H1 → intro → hero brief → Track A — Consumer iOS App (three decisions) → bridge → Track B — Provider Backoffice (three decisions) → analytics → close: Two surfaces. One loyalty system. Both live.
Two labeled tracks in sequence is the shape a portfolio uses for two projects. That's the read the structure teaches, screen by screen, and by the time the integration claim shows up it's arriving as an epilogue. Epilogues don't overturn a conclusion someone reached four screens back.
In July I called the H1 the principal problem. That was under-specified. You can swap a headline string and the altitude won't move, because the track structure is the skeleton.
The fix is already written and already on the page, in the wrong position.
Title and subtitle
Right on scope, wrong on proof: Allē · Dual-Surface Loyalty Redesign · 30M Members, 40K Practices · Juno Chen. Compare the siblings.
| Case | What the title tag leads with |
|---|---|
| Alibaba | $50B+ GMV |
| Thermo Fisher | $20M+ Margin |
| Allē | 30M Members |
Two business results and a population count. Swap it:
Allē · Dual-Surface Loyalty Redesign · 3.2× Redemption, $42 CAC · Juno Chen
The H1, When the product is someone's face, points aren't enough., isn't failing for being evocative. Your other H1s are evocative and they clear the bar, because each one names a structural absence (The supply chain six partners never had) or a mechanism (The signals before anyone touched a screen). This one names subject matter. Faces are the category, and the category was never the problem.
Take your own closing thought, move it up, put it in first person:
The program was built for the channel. I rebuilt it for the relationship.
Then demote the current intro to supporting copy and make it carry integration and outcome together:
Allē moved Allergan's product through providers. It was never designed for the
patient or the practice. I redesigned both surfaces, consumer iOS for members and
provider backoffice for 40K+ practices, and defined the single loyalty model
running both. Redemption 3.2×. Acquisition cost $92 to $42.Strategic premise
You have one and it's good: Built to move Allergan's products through providers, it was never designed for the patient or the practice. Real diagnosis of a real business failure, and it's the sentence that makes everything under it legible as systems work. Right now it's third in the intro paragraph, behind a member count and a line about member confusion, sitting under an aesthetic headline and fighting a six-row brief for attention.
Add this directly above the first section heading, before the surfaces split:
One loyalty model. Two role-specific views. The same loyalty event becomes a
member notification on one surface and a practice action on the other. Six
decisions, each one holding both sides of the appointment in the same frame.Rename the labels while you're in there. Track A and Track B read like rows on a Gantt chart:
Surface A — What the member sees
Surface B — What the practice seesOutcome proof
All three figures are already live, in the hero brief's outcome row: 3.2× redemption · 47% reactivation · $42 CAC. Position is the failure, not absence. That row is sixth of six, and on mobile the single-column stack drops it behind company, product, role, team, and scale. A six-second phone scan registers 3 product designers before it registers 3.2×.
Reorder the brief so outcome is row one. Add a metric bar above the H1:
3.2× redemption · $42 CAC (vs $92) · 47% reactivation
Hierarchy matters more here than placement does. Lead with redemption and CAC, not membership. Anyone can attach a visual refresh to a thirty-million-member brand; nobody triples redemption or cuts acquisition cost 54% with a visual refresh. Those two numbers are only explicable as systems work, so those two carry the altitude. Scale is context. Row two.
The full case says 30 million enrolled members in the hero and Eighteen million enrolled members in the problem section, while Allē's own public help material describes membership in the single-digit millions.
There's probably a defensible denominator here, cumulative enrollments across Brilliant Distinctions and Allē. It isn't on the page. Which means a hiring manager with ninety seconds and a search engine finds a public figure a fraction of the size, stops trusting your numbers, and the two numbers that carry the altitude go down with the one that didn't matter.
Name the denominator you can defend. If it's cumulative enrollments across both programs, put that on the page:
Scale — 30M+ cumulative program enrollments (Brilliant Distinctions + Allē) · 40K+ practices
If it counts something else, write that. If you can't reconstruct what it counts, cut it and lead with national-scale consumer and provider loyalty platform · 40K+ practices. An unsourceable number costs you more than no number.
Reconcile the 18M line the same day.
Craft signal
Strongest area on the page. Also the most underused. Six named decision blocks, each carrying a before-state, an after-state, and a reason, which is judgment rather than output.
Your best line sits in B·01: The default sort is the message. Five words, and they show a designer who understands that a system's defaults are its point of view. It's buried in a mid-page card. Pull it into the hero region as a design-principle callout, the treatment you already give Points aren't a relationship. Progress is.
What's missing everywhere is the rejected alternative. Six decisions, zero trade-offs. At Director+ altitude the panel is testing whether you can defend a choice against its most plausible competitor, and your B·00 copy already implies what you turned down. Say it:
The alternative was a configurable dashboard. Configuration is a tax on a front
desk that doesn't have time to pay it, so the system pre-decides the context and
lets the practice override it.Use a real turnaround-time number between patients if you have one you can source. Then add one trade-off on the member side. Two statements is enough to change the read.
The call to action
Request the full case study requests nothing, and both instances link straight to the full case. The copy promises a request; the link delivers a login wall. The wall fails in both directions. Medium confidence on the mechanism, high confidence on the effect: the full case returns a 200 with the whole document in the body, then a client-side check bounces a JavaScript-capable browser to login. Hiring manager in Chrome gets stopped. Anyone fetching the URL without JavaScript gets everything.
Pick a posture, match the copy. Gate stays:
Request access — full case: one loyalty model, two surfaces, six decisions
Gate goes:
Read the full case — how one loyalty model ran both sides of the appointment
Either way: move the check server-side or drop it. Shipping protected content in an unauthenticated response and hiding it behind a redirect is a craft signal too, and not the one you want a technical evaluator to find.
What changed since July 24, and the order to work in
No movement either way. Nothing shipped, and all five prior recommendations remain open. Everything that passed in July still passes: metadata scope, the desktop outcome row, the six decision cards, the closing integration line.
I'm sorting by one question. Does the fix change what an evaluator concludes on the scan, or only on the read?
- Move the closing system paragraph above Surface A. One paste. It is the difference between two projects and one system.
- New H1 and the title-tag swap. Scan layer, copy written above.
- Reorder the hero brief to outcome-first; add the mobile metric bar. Scan layer, phones especially.
- Reconcile 30M against 18M and define the denominator. Not altitude. Verification survival.
- Rename tracks to surfaces; elevate
The default sort is the message; add two trade-off lines. First-read layer. - CTA copy and gate mechanism, together or not at all. Last, because it changes nothing on the scan. It changes what happens after someone decides to click, and nobody is clicking yet.
One through three is a two-hour evening. Do them before your next outreach touch. The rest of this page is already strong enough that sequence is the only thing costing you.
- Maven's three-audience mandate: The VP of Design search at Maven Clinic spans member, employer, and health-plan audiences under one CPO, which makes Allē your closest structural precedent once the teaser reads as one system rather than two tracks.
- Stripe wants dual-surface commerce: The Senior Staff Agentic Commerce role names merchants, agents, and consumers as one connected problem, and it will inspect component-level craft, so the trade-off annotations you add here are the same evidence that role screens for.
- Equinox+ is still unlabeled: Your Equinox+ teaser remains titled generically, leaving the five-brand, three-month MVP story out of metadata entirely, which is a bigger scan-layer loss than anything currently wrong with Allē.
- Vanta buys standards, not projects: The Head of Design posting asks for authority over definitions of done and launch reviews, so a decision that became a reusable criterion, and the condition under which it stops applying, is the version of the Allē story that evaluator needs.