Ten seconds on the live Thermo Fisher teaser registers an execution story. The page isn't weak. The numbers are accurate, they're structured, and they clear the fold without help. The failure is one paragraph, and it is the cheapest item on this entire list.
Read the opening paragraph looking for a subject. There isn't one. A supply chain was broken. A platform now exists. No one in those sentences moves the first state into the second.
On a thinner page that would be a footnote. Here it costs you, because a scanner who can't locate the actor doesn't slow down and go looking for them. They fill the slot with whatever the category usually contains: competent designer, successful project, someone else's mandate.
What the ten-second scan sorts for
Above Director, shipped good work is table stakes. The sort runs on one variable: built the function, or joined it. CS-02 is the strongest leadership case in this portfolio. Zero to one in regulated pharma. Six partners, nine sites, three continents, twelve months, a margin number at the end of it. It is also the only case on the property that asks the reader to infer the leadership instead of handing it to them.
1. The title holds. One word downstream doesn't.
Metadata currently reads Thermo Fisher / mySupply — 0→1 Pharma Supply Chain, $20M+ Margin — Juno Chen. Run the altitude test. Does $20M+ Margin name someone who created the conditions or someone who executed inside them? Alone, it doesn't resolve. 0→1 resolves it, because it makes an agency claim no dollar figure can make on its own: nothing preceded you.
I drafted a replacement and killed it. Building the Supply Chain Function Six Pharma Partners Never Had swaps a verified number for an unverifiable one and shoves the outcome past the truncation point in search results. The July 24 audit said leave the title alone. Still the right call.
The number itself is another matter. It appears three times, phrased three ways.
| Surface | Phrasing |
|---|---|
| Title metadata | $20M+ Margin |
| Hero row | $20M margin opportunity |
| Full case | $20M+ margin recovered annually |
"Opportunity" is what consultants call a figure that lives in a deck. "Recovered annually" is what operators call a figure that happened. The altitude question sits inside that one noun, and you are currently running both sides of it.
Change the hero row to $20M+ margin recovered annually. That assumes the full case phrasing is the accurate one. If it isn't, downgrade the full case instead. Either resolution beats the version where a careful reader finds both.
2. Two of the three first-screen signals fire
At 1440 × 900 the fold sits below the hero, all six brief rows, and the top of the platform showcase. Scale: present. Outcome: present. Role altitude: missing.
You already solved this elsewhere. CS-01 opens on an abstract headline and then, in the next paragraph, states that you named the gap, built the research case, secured the mandate, and redesigned the platform. CS-03 runs the identical move. The pattern exists on the property. CS-02 kept the headline and dropped the paragraph.
Keep The supply chain six partners never had. It's on voice. Replace what sits under it:
No platform existed. Six pharmaceutical partners and nine manufacturing sites ran on spreadsheets and phone calls, and exceptions surfaced at the delivery gate, where they cost three to five times more to fix. I ran the 32 interviews that defined the problem, turned five failure modes into five modules, and led delivery as the only designer alongside two BCG Digital Ventures engineers. Zero to live across all six partners in twelve months.
Every clause there traces to published case content. One thing I cannot determine from the public page: whether you created the mandate for this work or received it. That distinction is worth more than any other sentence on the property. Created it, add the clause. Received it, leave it out. What makes the CS-01 formulation land is that it's specific enough for a reference call to break, which is exactly why it reads as true.
Then ship the row the last audit asked for:
Product Design Director · Solo design lead, 0→1 across six partners
The row already sits above Team and Scale, so placement was never the issue. Content was. It currently states a title. Make it state a mandate, and weight it so a scanner hits it before the numbers. As built, the team row says one designer and two engineers and leaves the arithmetic to the reader.
3. Five specimens sit on top of your best evidence
The craft grid opens on color tokens, status badges, buttons, type specimens, and a KPI card. The artifacts carrying leadership signal come after all of that.
A design-system specimen at the top of a scroll is a senior IC's opening move. It sets the frame for everything below it, and the frame it sets is execution.
The material underneath is strong. The exception card isolates LATE and AT RISK as the two states that cost money and makes the red segment the primary scan target. The partner scorecard is the first shared view of six partners' delivery performance, benchmarked weekly rather than after correction stopped being possible. The capacity card shows a production line at 107% and reports that it "prompted the conversation that recovered $2M."
That last string is the most valuable text on the page, and a reader has to work past five specimens to reach it.
The scorecard is doing something the page never takes credit for. Per the full case, the dashboard put Thermo Fisher's OTD figures next to each partner's, which turned previously invisible discrepancies into the opening question of strategic reviews. That is the closest thing to a published adoption mechanism: performance made legible between peers. If that's why six partners came on, say so, and 100% adoption converts from a number into a design outcome. Confirm it first. The case doesn't currently make that claim, and it is precisely the line an evaluator will ask you to defend.
The full case attributes no portion of the $20M+ to any individual module, so don't manufacture the connection. The bounded claims are enough.
Reorder to open on four items: regulatory constraints, exception queue, capacity heatmap, partner scorecard. Tokens and typography go into one strip at the bottom, or they go.
Then rewrite the regulatory card. It currently says, in passive voice, that the platform surfaces evidence for QA rather than replacing QA review. Correct thinking, no author attached to it. Trust Is the New Interface already runs compatible language, so nothing shows at the seam:
Five modules run continuously. One gate stays human. Batch QA release is the point where a wrong call is irreversible and personally attributable, so the platform assembles the evidence and the signature stays with the Qualified Person.
That paragraph does more work with the buyers currently hiring than anything else on the page, because what they are trying to figure out is who decides which decisions a machine isn't permitted to make.
4. The ask is fixed. The promise isn't.
Both CTAs read Request the full case study and point at the gated case, which bounces unauthenticated visitors to a password page. On July 24 that page explained nothing at all. It now explains the gate and carries a working email link. The broken-ask finding is closed. The redirect remains clumsy, but the conversion path exists.
What's left is the label. It promises a case study, delivers a wall, and says nothing about altitude. Choose by buyer:
| Use for | CTA copy |
|---|---|
| AI and trust-surface roles | "Request the full case: how five modules became five agents, and why batch release stayed human." |
| Platform and scale roles | "Request the full case: how six partners reached 100% adoption in twelve months." |
I verified the agent claim before recommending it. The full case states that the five modules become five continuously running agents with QA release as the single human gate, so the CTA promises something the case actually pays.
One reconciliation before either goes live. The teaser says "100% partner adoption." The full case says "6/6 pharma partners committed." Those are different claims, and the reader who gets past the gate is the same reader who notices.
Delta against July 24
| July 24 recommendation | Status |
|---|---|
| "Product Design Director · Solo design lead, 0→1." | Not implemented. The row still reads Product Design Director alone. |
| "Reduce the craft grid to four items: status badges, the AI-gate KPI card, regulatory constraints, and either the exception summary or partner scorecard." | Not implemented. The full token, component, analytical, and operational collection is still live. |
| "Link directly to mail or Calendly and label the action as a request for access." | Partially resolved by a different route. The label and link are unchanged, but the access page now supplies the email request the CTA should have carried. |
Also missing: the irreversibility sentence and the agentic-architecture promise. Neither one appears on the teaser in any form. On the credit side, the hero metrics were correctly left alone.
Still outstanding from the trust-calibration piece: the gated case HTML still transmits before the client-side check runs. On a portfolio whose thesis is trust, that remains the loudest contradiction in the build.
Order of operations
- One word in the outcome row. Five minutes.
- The intro paragraph. Fifteen minutes, and it is effectively the whole audit.
- Craft reorder plus regulatory card rewrite. One hour.
- CTA string, role row weighting, adopted versus committed. Under thirty minutes for all three.
This page is closer than it reads. It has a leadership story in hand and is telling it in the voice of a project retrospective.
-
Where the QA gate travels: The rewritten regulatory card is the paragraph to lead with for HackerOne's Director, AI Product Design, which reports to the CPO and names oversight, accountability, transparency, and risk visualization as the core of the mandate.
-
Adoption needs a mechanism: Before the six-partner story goes anywhere near Vanta's Head of Design search, which asks for quality mechanisms across a roughly 40-person org, decide what you can defend about why adoption happened.
-
The buyer's control vocabulary: "Auditable, attributable, reversible" is now enterprise product language, not UX theory, per Rubrik's Agent Cloud materials — worth borrowing when you describe the batch record.
-
Autonomy is not a staircase: Anthropic's measurement of agent autonomy found experienced users auto-approving more often while interrupting more often, which complicates any Watch → Verify → Delegate framing you bring into an interview.