The gate is still broken. Fix it first.
A raw HTTP request to junochen.com/cs-01-alibaba returns HTTP 200, content-length: 119392, and the complete case study in the initial HTML payload. Every heading, every metric, every paragraph of the agentic sourcing vision. The pf_access_v1 script checks localStorage before redirecting unauthenticated browsers to /login.html. Content is already delivered by then. curl gets everything. So does a Slack unfurl. So does view-source.
Flagged in the site-risk audit three weeks ago. Same pattern holds for what-do-you-count.html, which ships the full private TinyFish/pricing essay before the overlay JS runs.
This matters more for the Alibaba case than for anything else on your site.
Your Trust essay argues that trust requires mechanisms that function under inspection. The Verify step of the Trust Ladder is where a platform buyer tests whether the system does what it claims. A gate that ships content before checking credentials is a Verify failure in your own framework. A CPO reads the essay, clicks through, and has an engineer on their team pull the source. They see every word behind the "protected" page sitting in the raw HTML. That's the same gap you diagnose in Alibaba's pre-redesign PDP: Trade Assurance sat below the fold while the platform claimed transaction confidence. Your portfolio is now doing what the old Alibaba did.
Server-side authentication or no gate at all. A cosmetic gate is worse than no gate because it makes a promise it doesn't keep. Everything else in this audit is secondary until this is resolved.
Where the case stands
The first full audit found strong raw material and five structural weaknesses: research read as a methods list, sprint sequencing had no stated rationale, IC+manager proof was thin past Sprint 01, tradeoff reasoning was absent, the agentic vision was implicit. The page has changed since then. The hero now claims you named the gap, built the research case, secured the mandate.
Language in a hero is a promise. The body has to deliver on it.
Six criteria, evaluated against the current live copy.
1. Problem diagnosis ownership
Passes.
The hero frames the 25%/80% gap (25% of traffic generating 80% of transaction value on desktop) as your diagnosis. The diagnostic section reinforces it with three named failures: first impression, search journey, transaction confidence. Each grounded in specific evidence. The procurement buyer quote comparing the page to a flea market works. The sign-in rate running 37 points below app works.
Most interview-ready criterion on the page. A panel member will see that you identified something the organization had not articulated, and that the diagnosis drove the work.
One refinement. The 25%/80% gap reads as a finding. What did the organization believe before you named it? Were they focused on mobile? Had they written off desktop as legacy? A finding means you measured something new. A reframe means you changed how the org thought about something it was already measuring. Reframes are harder. They prove more. Add one sentence about the prior assumption.
2. Research as mandate-building
Falls short.
The hero says you "built the research case, secured the mandate." The body lists 32 cross-functional interviews, a Baymard audit, gaze tracking on $500+ order sessions. These read as inputs to the diagnostic. They do not read as tools you used to build organizational permission for the work.
A Director+ panel will ask: Who did you have to convince? What was the resistance? Was this already prioritized, or did you create the priority? Thirty-two interviews is an impressive number. The case doesn't show who those interviews were with, what they revealed that the organization didn't already know, or how the findings changed resource allocation.
The teaser page metadata promises "executive buy-in." The full case body never uses that phrase and never identifies an executive by role or function. A buyer clicking from teaser to case study is making a trust-ladder transition. The teaser sets the expectation. The case has to meet it.
Add two to three sentences in the diagnostic section that name the organizational state before the research. What was the prevailing assumption? Who held it? What did the research reveal that contradicted it? What changed as a result? You don't need to name individuals. You need to show the research was a political instrument, not just an analytical one.
Use this structure:
"The 32 interviews surfaced X, which contradicted the prevailing assumption that Y, and gave the team the evidence to secure Z."
That proves mandate-building.
3. Sprint rationale
Falls short.
Three diagnostic failures map to three sprints: First Impression (Homepage), Search Journey, Transaction Confidence (PDP). Clean mapping. Stated order. No explanation for the order.
A senior evaluator will see the mapping and ask: why Homepage first? You could have started with PDP, where the money converts. You could have started with Search, where 73% of sessions died. You started with Homepage. Why?
Add one paragraph between the diagnostic section and Sprint 01. If first impression had to change before search improvements could be measured (because users were bouncing before reaching search), say that. If Homepage was the fastest win and you needed early momentum to maintain the mandate, say that. Either answer is valid. No answer reads as mid-level.
4. IC+manager proof
Falls short. Highest-risk gap for a CPO buyer.
Sprint 01 contains one named cross-functional moment: a PM worried that removing the sign-in wall would drop sign-in rates. The outcome (sign-up rose 4%) resolves the concern. Sprints 02 and 03 contain no named cross-functional interactions, no team-size details, no hiring or management actions, no decision-rights moments.
The metadata says Head of Design & Research. The body proves Head of Design. It does not prove Head of Research (where is the research team, how did you build or direct it?) or the leadership half of the title.
A CPO evaluating this case is asking one question: will this person reduce the number of design decisions I have to make, or generate more decisions that escalate to me? IC proof alone doesn't answer that. The prior audit flagged this. The page hasn't changed on this dimension.
You don't need a "leadership" section. Embed leadership proof inside the sprint narratives where it actually happened.
- Sprint 01: The PM objection is good. Expand by one sentence: what did you do with the objection? Did you run a test? Propose a compromise? Escalate? The resolution (sign-up rose 4%) is the outcome. The panel wants the decision process between objection and outcome.
- Sprint 02: Name one moment where you directed the research team, resolved a cross-functional disagreement, or made a call that required organizational authority rather than design judgment.
- Sprint 03: The -47% buyer-reported security concerns metric is your strongest outcome. Who measured it? If your research team ran the study, say so. That's leadership evidence disguised as a metric.
5. Tradeoff with reasoning
Falls short.
Each sprint follows the same pattern: problem statement, design approach, outcome metrics. The redesign reads as a sequence of correct moves. No sprint contains an explicit moment where you chose one path over another and explain why. Every redesign at this scale involved tradeoffs. The case presents outcomes as though the path was obvious.
Tradeoffs are the single clearest separator between a senior case and a mid-level case. A mid-level case shows what was designed. A senior case shows what was considered and rejected, and why.
Pick one decision from any sprint where you had a real alternative. Strongest candidate: Sprint 03. You put tier pricing inline and added a live order calculator. The alternative was presumably to keep pricing behind the inquiry wall, which protected supplier negotiation leverage. Why did you choose transparency over protection? What was the supplier-side risk? How did you mitigate it?
Three sentences. That moves the case from "here's what I designed" to "here's how I evaluate competing risks."
6. Agentic sourcing vision
Partially passes.
The closing section turns the three sprint surfaces into a sourcing loop: natural language in chat becomes a structured brief, traverses Search and PDP, evaluates trust signals, confirms pricing and lead time, reaches an order surface. The mock brief (500 custom water bottles, ≤20 days, ≤$10/unit, Trade Assurance required) is specific and grounded. Every surface in the vision derives from architecture you built in the case. That structural connection to the sprints is what makes it read as a real system rather than a concept deck.
But it reads as autonomous completion. The loop closes on-platform without naming where the procurement buyer decides whether to trust what the system produced. Your Trust essay calls this the commitment point. A procurement buyer committing six months of inventory spend on the strength of a screen. The case study vision shows the screen. It doesn't show the moment of commitment, or what the system provides at that moment to make the commitment rational.
Name the decision gate. Where in the sourcing loop does the procurement buyer review what the agent found and decide whether to proceed? If the agent surfaces three suppliers who meet the brief, what trust signals does the buyer see to choose between them? What is reversible at that point, and what isn't? One paragraph that makes the human judgment moment explicit. That turns the vision from a UX improvement pitch into a design-leadership position on how agentic commerce should work.
Trust essay connection
The Trust essay names Alibaba in §01. SMB owners committing six months of inventory spend through a screen-mediated manufacturer relationship. The §04 Trust Ladder (Watch → Verify → Delegate) is the essay's structural framework. The conclusion returns to the procurement buyer alongside pharma and disaster relief as cases where trust architecture determines whether humans can act on system outputs.
The Alibaba case study does not cross-link to the Trust essay. Does not use the words Watch, Verify, Delegate, or Trust Ladder. Does not reference the commitment-point framing.
For a buyer who hasn't read the essay, this doesn't matter. The case stands alone.
For a buyer who has read the essay, the connection requires them to do the mapping themselves. The essay makes the argument. The case study is the proof. But the proof doesn't know the argument exists.
The full structural option: reorganize the case around the Trust Ladder stages. The diagnostic section would frame the three failures as breakdowns at Watch (first impression drives bounce before engagement), Verify (search and PDP fail to provide the signals procurement buyers need to confirm supplier quality), and Delegate (the agentic vision as the architecture that earns delegation). The sprint sequence becomes a trust-recovery sequence. The agentic vision becomes explicitly a Delegate architecture, one that works only because the prior sprints repaired Watch and Verify.
Powerful for a buyer who has read the essay. Harder to read for a buyer who hasn't. The case needs to stand alone. Know the structural option exists. Don't execute it now.
What you should do instead: echo the framework at three natural contact points, so a reader who knows both pieces recognizes the architecture without being told to look for it.
- Diagnostic section, when you describe procurement buyers citing security fears at payment: that's a Verify failure. A phrase like "buyers had no way to verify what the platform claimed" echoes the essay without citing it.
- Agentic vision, when the sourcing loop reaches the order surface: that's the Delegate moment. A phrase like "the procurement buyer delegates the search but retains the commitment decision" connects to the ladder without naming it.
- The 40M buyers figure appears on your homepage case card but not in the case study or the Trust essay. If you use it, use it once, in the vision section, to convey the scale at which trust architecture operates.
Priority sequence
The case has improved since the first audit. Problem diagnosis ownership passes. The agentic vision is structurally grounded. The sprint structure is clean.
Four things need to change before this case is panel-ready. Ranked by speed-to-fix, not severity. IC+manager proof is the highest-risk gap for a CPO buyer, but it requires you to reconstruct specific cross-functional moments from memory. Can't be solved with a sentence template. The items you can fix tonight come first.
- Fix the gate (server-side auth or remove it). 2. Add tradeoff reasoning to one sprint. 3. Embed leadership proof in Sprints 02 and 03. 4. Name the human decision gate in the agentic vision.
- Fix the gate. Server-side auth or remove it. A cosmetic gate on a case study about trust is a contradiction a panel will notice.
- Add tradeoff reasoning to one sprint. Sprint 03 tier-pricing transparency is the strongest candidate. Three sentences.
- Embed leadership proof in Sprints 02 and 03. One cross-functional moment per sprint. Name the decision, not the org chart.
- Name the human decision gate in the agentic vision. Where does the procurement buyer decide whether to trust what the agent found?
Research-as-mandate-building and sprint rationale matter, but they carry lower interview risk. A panel can ask you about mandate-building and you can answer verbally. They cannot ask you about a tradeoff that isn't in the case, because they won't know it exists.
The Trust essay connection is positioning refinement, not structural repair. Do it after the four above are done. The essay is doing its job as a standalone piece. The case study's job is to prove the essay's thesis without the reader needing to be told that's what it's doing.
- Amplitude's Wave product mirrors the agentic sourcing vision's open question: its agent surfaces opportunities and routes work while a product team decides whether to act, which is the same commitment-point architecture the Alibaba vision needs to name explicitly.
- Rubrik's reversibility language offers a useful vocabulary model: their AI launch describes autonomous actions as auditable, attributable, and reversible, with human review required when an action cannot be undone.
- Brex's delegation framing is the closest live role language to the Alibaba vision's missing decision gate, asking designers to shape how users delegate financial work to AI, supervise work in flight, and review and correct outputs.
- OpenAI's confirmation-before-commitment pattern in their cloud browser documentation requires the system to pause and ask for confirmation before actions with financial or legal real-world consequences, which is the exact mechanism the Alibaba agentic vision should make visible at the order surface.