Current-state summary, July 10, 2026. The full case returns HTTP 200, 213,788 bytes. Browser title: Project Horizon — mySupply — Juno Chen. Four content sections: hero (role, team, scale, outcome metrics), problem ("The world's largest CDMO, running on spreadsheets"), solution (five-module system with failure-to-module map), and agentic future ("If I were building this today"). No <img> tags anywhere. Every product surface is a coded HTML/CSS mockup, with four Chart.js canvases rendering live data visualizations. The teaser is ungated, 121,818 bytes.
Password gating. Same client-side localStorage guard as CS-01. Full content ships in the HTML before the JavaScript check fires. I flagged this in the Alibaba audit with the recommendation to fix both cases in one remediation pass. That recommendation stands. A candidate whose positioning rests on trust architecture cannot gate her most sensitive work with access theater. Fix it once, fix it everywhere.
Now the content.
The page documents what Juno built. The judgment behind each decision remains off-screen. At Director+, every candidate has a build record. Conviction comes from the thinking. Everything below follows from that gap.
1. 0→1 Mandate Framing
What the page does. The hero states "six pharma partners" were "running on spreadsheets and phone calls." The problem section reinforces with "the absence of a shared source of truth." Team: one designer, two BCG DV engineers. Timeline: twelve months. Metrics row: $20M+ annual margin recovered, 83% IRR, 6/6 partners committed, 42% overhead reduction.
Assessment: passes, but a careful reader has to assemble it. A Director+ evaluator scanning this page in three to five minutes between meetings will not do that assembly work. The phrase "zero to one" never appears in the full case hero, even though the teaser's meta description says it explicitly. The full case relies on implication where it should rely on declaration.
Fix. Add one sentence to the hero, before the team line:
"No platform existed. No design precedent. No internal design team."
Three declarations that make the mandate unmissable and set the scale before the reader hits the metrics. The metrics land harder when the starting point is already understood as zero.
The team composition is doing important work the page doesn't call out. "1 designer + 2 BCG DV engineers" means Juno was the entire design function for a platform serving six enterprise pharma partners across nine sites. Say it. "I was the design function" changes how every subsequent module is read. And the fact that this case study's product surfaces are hand-coded HTML/CSS mockups with interactive Chart.js canvases reinforces the same point: the person who designed the system also built the artifact that demonstrates it. That's 0→1 instinct expressed in the portfolio itself.
2. Research-to-Diagnosis Chain
What the page does. The problem section states thirty-two interviews traced the breakdown to the absence of a shared source of truth. Exceptions surfaced two to four days after commit dates. Five metric cards quantify the damage (50% short-notice orders, 5–10 hours status chasing per batch, zero partner visibility, OTD variance from 63% to 99% by site, 45% of data outside PRISM in Excel and PDFs). The solution section opens with "each research failure mode became a design brief" and presents a five-row failure-to-module map.
Assessment: the endpoints are strong, the middle is missing. I can see the input (32 interviews) and the output (5 failure modes). I cannot see the analytical step between them. How did thirty-two interviews become five failure modes and not seven? What was the clustering method? What did Juno hear across those conversations that made her frame the core problem as "exceptions discovered too late" rather than "data fragmentation" or "partner distrust"?
The research-to-diagnosis step is where Director-level judgment lives. Anyone can conduct interviews. Anyone can build modules. A CPO evaluates whether this person can look at messy qualitative data from thirty-two conversations across nine sites and extract the structural pattern that becomes the design strategy. The page asserts that this happened. The analytical step stays invisible.
Fix. Between the metric cards and the failure-to-module map, add a diagnosis block. Three to five sentences:
"Across thirty-two interviews at nine sites, the same pattern repeated. The data existed. The systems existed. But the information that would prevent an exception lived in one system while the person who needed it operated in another. Every failure mode traced back to the same structural gap: the absence of a shared, real-time view that both Thermo Fisher and the partner could see simultaneously."
Then:
"I clustered the interview findings into five failure modes, each a different manifestation of this gap."
This gives the evaluator the analytical move. It shows that the five modules are derived consequences of a single structural insight, not five separate feature ideas.
3. Module-to-Failure-Mode Mapping
What the page does. Five-row map before the module demos:
| Problem | Module |
|---|---|
| Flat list, buried exceptions | Exception-first orders |
| 5–10 hours of calls per batch | Live kanban |
| Different OTD definitions | Bidirectional KPIs |
| Emailed XLSX, filename versioning | Structured submission, SAP auto-load |
| Capacity constraints discovered post-commitment | 18-month utilization heatmap |
Assessment: the strongest structural element on the page. The one-to-one mapping is clean, legible, and does exactly what a Director+ evaluator needs. It shows design judgment as a traceable chain from problem to solution.
Module 01's framing is particularly effective. "The prior model treated 197 orders equally while research indicated managers needed to act on only three." That sentence communicates information architecture judgment, prioritization thinking, and user-research grounding all at once. A CPO reads that and understands Juno knows how to make a system opinionated about what matters.
One refinement. The map promises and the module details deliver, but on a page this long, scroll distance between the two creates friction in the conviction arc. The evaluator who reads Problem 03 in the map needs to believe Module 03 fulfills that promise. Right now they have to scroll through Modules 01 and 02 to find out. Anchor links from each map row to its corresponding module section would keep the evaluator in the belief chain rather than losing them to page fatigue. Small move, but conviction accumulates, and scroll distance interrupts the accumulation.
4. Tradeoff Evidence
What the page does. Module 01 contains one implicit tradeoff: showing all 197 orders versus surfacing only exceptions. Module 04 contains another: email-based workflow versus structured submission. Both are framed as before/after comparisons, presented as improvements over a prior state, with no indication that alternatives were evaluated and rejected.
Assessment: the biggest gap on the page. A Director+ evaluator distinguishes between a designer who built the obvious solution and a designer who chose the right solution from a set of plausible alternatives. Every module currently presents its solution as the natural, inevitable response to the problem. Clean storytelling. But it strips out the judgment evidence that hiring committees at this level are specifically scanning for.
Fix. You don't need tradeoffs in every module. Pick two.
Module 03 (Dashboard). The bidirectional KPI design is the most interesting decision on the page, and the tradeoff behind it is invisible. Why show both Thermo Fisher OTD and Customer OTD side by side? Based on the design as shipped, the likely rejected alternative was a single normalized OTD metric. The reason to reject it: a single metric would have hidden the 3-point gap between TF's 97% and the customer's 94%, and that gap was the conversation starter that made the dashboard a strategic tool rather than a reporting surface. If that reasoning matches what actually happened, say it. Two sentences. It transforms Module 03 from "we built a dashboard" to "we made a specific information-architecture decision that changed how partners used the platform." [Confidence: inferred from the design's structure, not stated on the page. Juno should confirm or correct the rejected alternative before publishing.]
Module 05 (Capacity). The heatmap distinguishes "true constraints" from "manageable overcapacity." That distinction is a design decision. The design suggests the rejected alternative was treating all overcapacity uniformly, likely as red or high-alert states, which would have triggered false alarms and eroded planner trust in the tool over time. The choice to use two visual states reduced alert fatigue and kept planners engaged with the heatmap across an eighteen-month horizon. A CPO will recognize that as sophisticated systems thinking. Surface it. [Confidence: same caveat. The visual distinction exists in the coded mockup, but the reasoning behind it is my inference.]
In both cases, the tradeoff block needs only two to three sentences. The structure: The alternative was X. We chose Y because Z. That's what judgment evidence looks like at this level.
5. Partner Adoption as Design Outcome
What the page does. The hero reports 6/6 pharma partners committed. The problem section establishes that partners had zero visibility into Thermo Fisher's performance. Module 03 shows bidirectional KPIs. The page never explicitly connects adoption to design decisions.
Assessment: adoption is stated as a business outcome, unattributed to design. A VP Design reading this page could reasonably conclude that partner adoption was driven by sales, executive relationships, or contractual obligation. That conclusion would be wrong, but the page doesn't prevent it.
The design decisions on this page are directly responsible for adoption in ways that take three sentences to articulate.
Fix. Add a short block after the five modules, before the agentic section:
"The bidirectional dashboard gave partners something they'd never had: visibility into Thermo Fisher's performance against the partner's own metrics. Partners adopted because the platform gave them leverage they didn't previously have. The forecast portal eliminated thirty minutes of manual SAP loading per file, which meant the Thermo Fisher operations team had an internal incentive to push partners onto the platform rather than accepting emailed spreadsheets. Adoption was designed into the system's incentive structure. The platform's own mechanics drove it."
Three to four sentences. They reframe 6/6 from a metric into evidence of systems-level design thinking.
6. Agentic Vision Placement
What the page does. The final section, "If I were building this today," reframes the five modules as five continuously running agents. It preserves one human gate: batch QA release requiring regulatory signature. The mockup shows a live activity feed with three active agents, 195 orders processed, and notification cards for human gates, exception monitoring, document compliance, forecast deviation, and order scanning.
Assessment: placement is correct, framing is strong, bridge is thin. The agentic section earns its position because it follows five modules of demonstrated domain expertise. The human-gate preservation is the smartest move on the page. It signals that Juno understands where automation should stop in regulated environments, which is the exact judgment call that companies building AI into consequential workflows need from a design leader right now.
The weakness is the transition. The page jumps from Module 05 to "If I were building this today" without explaining why the current architecture has a ceiling. The section mentions "mySupply was pull-based because someone still had to open it to find problems," but that observation is buried in the middle of the section where it should be the opening argument.
Fix. Before the agentic section, add two sentences as a bridge:
"mySupply solved the visibility problem. But visibility is only valuable if someone is looking. Across nine sites and six partners, the assumption that the right person would open the right screen at the right moment was the system's remaining single point of failure."
Then the agentic reframe lands as the logical next step, continuous with everything above.
Teaser-to-Full Continuity
The teaser gives away an extraordinary amount of product-system detail before the gate: color tokens, status taxonomy, nine data visualizations, partner scorecards, capacity heatmaps, regulatory rationale, a fully rendered interactive dashboard embedded via iframe. A hiring evaluator who sees the teaser may feel they've already seen the product. And if they feel that, they may never request the password. The full case becomes invisible to the evaluator it was built for.
The full case's unique value should be the judgment chain: research → diagnosis → module mapping → tradeoffs → adoption attribution → agentic vision. If you implement the fixes above, the full case delivers something the teaser structurally cannot: evidence of how you think, beyond what you shipped.
Right now, significant portions of the full case read as a more detailed version of the teaser, a difference of degree where there should be a difference of kind. The tradeoff gap and the adoption-attribution gap are the reason. Close them and the teaser-to-full progression becomes a deliberate escalation from impressive product to impressive judgment. That escalation is what makes the password worth requesting, and the full case worth reading.
Priority Stack
Add the two tradeoff blocks (Module 03 and Module 05). Confirm or correct my inferred rejected alternatives, then write two to three sentences each. Tradeoff evidence is the single highest-signal element for Director+ evaluation and the single biggest gap on this page.
If you have time for two: Add the adoption-as-design-outcome block after the modules.
If you have time for three: Add the research-to-diagnosis bridge between the metric cards and the failure-to-module map.
The 0→1 language fix and the agentic bridge are each one to two sentences. Do them regardless of everything else.
- Ramp's AI-native design language: Their current Director, Product Design posting describes designing memory, shipping PRs, and defining how design works in an AI-native world, which maps directly to the agentic vision section's strongest moves.
- Headway's consequence-weighted workflows: The Design Director, Provider Experience role owns clinical documentation, insurance verification, and AI-centered workflows where the cost of a confident wrong action is high, making Thermo Fisher the natural lead case.
- Trust essay as the agentic bridge: Juno's public Trust Is the New Interface essay explicitly ties the decision-gate argument to Thermo Fisher by naming five agents running continuously while QA regulatory release remains the non-negotiable human gate.
- NIST's expanding trust framework: The AI Risk Management Framework released a Trustworthy AI in Critical Infrastructure concept note in April 2026, which gives the agentic section's human-gate argument an external policy anchor for enterprise and regulated-platform outreach.