Delta since July 24
Redeployed this morning. Three entries for the log.
- One fix. Act 05 now names a decision outright: "Urgency bars replaced sorting, supervisors moved to the highest-priority case automatically." That closes the craft gap I flagged last month.
- One break. The declared share image,
og-cs-03.jpg, returns a 404. I can't tell you whether it broke this month or was never uploaded. It's live and it's yours. - Everything else held. Title and metadata still lead with mandate and scale. The humanitarian H1 is untouched. Nobody added a subhead. The decision-architecture line is still absent, and both buttons still aim at the gated full case.
Eight seconds still buys a humanitarian design project. Every piece of evidence for a GM-scope consolidation exists on this page. None of it lands in the first two beats.
One withdrawal before the audit. In the July pass I told you to trade "Designing for the worst day of someone's life" for "One platform. Six legacy systems. National deployment in six months." Wrong call, and it's why nothing moved: I asked you to swap a live sentence for a spec sheet. You declined. You were right to.
1. Mandate scope vs. humanitarian framing
Currently: the scan opens on "Designing for the worst day of someone's life." The intro then puts displaced families in shelters. Ownership shows up in sentence four.
Verdict: fails. Two beats of empathy before the reader learns you ran anything, and in an eight-second scan two beats is the whole scan.
Keep the stakes. Move them one position right. Disaster stakes are the cheapest scope-compression device on this page: a reader who registers that families were still in shelters understands, in one beat, why six months wasn't negotiable. What they can't do is sit where the scope belongs.
Replace the H1:
"Six legacy systems consolidated into one national platform in six months. Families were still in shelters."
"Legacy" earns its place in front of a GM-level reader. And the shelters hit harder second, because now they explain the compression instead of substituting for it.
Then add the subhead you skipped last month, rewritten:
"Product Design Director and GM on a seven-person team, with federal oversight live from launch day. Every choice an untrained volunteer made at intake set the eligibility tier, the assistance cap, and the audit trail for that family's entire case."
That second sentence is the thing this page has never once said. Right now the teaser tells a reader how big the problem was and never tells them what kind of problem it was, and the kind is worth more. You designed the point where a non-expert, under pressure, makes a call the system then enforces downstream, irreversibly, in front of an auditor. That's the highest-value asset in the case, and it appears nowhere above the fold.
2. GM scope before craft
Currently: the right-side brief carries the load, and it's the best-built element on the page.
| Role | Product Design Director / GM · 0→1 Build |
| Scale | 6 months · National deployment · Replacing 6 legacy systems |
| Outcome | $847K disbursed · 6→1 system switches · 1,689 cases in 2 weeks |
Verdict: works on desktop, fails on phones. In the markup, headline and intro come before the brief, so a phone reader gets a full screen of mission framing before "GM" exists anywhere. I have no analytics on your traffic, so treat that as inference. Recruiter shares default to mobile. Act as if it's true.
Reorder so the brief precedes the intro at mobile widths, then promote a single stat rail directly under the subhead so scope survives any layout collapse:
"6 legacy systems → 1 platform · National deployment in 6 months · 1,689 cases in the first two weeks"
The dollar figure stays out of the rail deliberately. Section 3.
Then put ownership into prose. A "Role" row reads as metadata, and metadata doesn't persuade anyone. Open the intro:
"I owned the whole thing: the platform, the rollout, and the six systems it replaced."
Your Thermo Fisher teaser already runs scope through title, headline, and first sentence. Steal that opening. Its conversion path breaks the same way yours does; see section 5.
3. Outcome proof, and one figure to pull
Currently: the metrics live in the brief's Scale and Outcome rows and in the closing line of the intro.
Verdict: works. They're visible, and they're positioned as evidence for the scope claim rather than as decoration. Leave them where they are. Three cautions.
Never let $847K sit next to a time window. The full case labels it "disbursed across active events" and states no period. It's a stock figure, not a rate. Put it beside "in two weeks" in a headline or a recruiter email and the first finance-literate interviewer asks for the window you don't have.
On the 98.7% FEMA compliance figure, I was wrong. I told you to promote it above the fold. Withdraw that. You can't say what the denominator was, who measured it, against what standard, or over what reporting window, and it isn't in your own canonical outcome list either. A compliance percentage is the single most interrogated class of number in a portfolio review. Keep it off the teaser until you can say what was measured against what.
"National deployment" is asserted, not shown. Nothing on the page names a region count, a chapter count, or a rollout sequence. The claim is probably sound, but load a one-sentence footprint answer anyway, because anyone who has run a multi-site rollout asks inside ten minutes. If that sentence doesn't exist, fall back to what the case actually supports: "Deployed across three concurrent disaster events." It's a smaller claim, and nobody can take it apart.
4. Craft signal above the fold
Currently: the urgency-bar line is the only design decision the page states as a decision, and it sits five acts deep. The intake logic (three fields instead of eight, driving eligibility, identity verification, and the FEMA audit trail) surfaces only in the closing section, as a number rather than as a choice you made. Nothing in the first screen separates design work from program management. Six acts and eighteen interface demos scan as inventory.
Verdict: improved, still failing where it counts. Pull the intake decision up to the second screen, under the stat rail, not above it:
"Intake ran on three fields instead of eight. One of them, disaster type, set the hardship codes, assistance caps, and eligibility tier for the rest of the case. So the form preselected it and hid the rest until needed."
Thirty-eight words, a decision with its consequence attached, no internal vocabulary for a hiring manager to decode.
One thing to skip, one to fix. Skip the offline-mode line: the page shows a "Low connectivity area" label on two mockups and documents no sync behavior anywhere, and a label is not a capability. Fix the six illustrations shipping with empty alt attributes, which is a bad detail to leave broken on a disaster-relief case. The pattern is this simple:
alt="Field volunteer interviewing a displaced family beside a damaged home."
5. The ask
Currently: both buttons read "Request the full case study" and link straight into the full case, which serves roughly 249KB of complete HTML and then bounces unauthenticated visitors to a login page via a client-side check.
Verdict: fails twice. The label promises a request process that doesn't exist, and the gate hands the entire case to anyone who views source. Friction for real buyers, zero protection against anyone who cares.
Pick one and commit.
- Keep the gate and make the ask real: label it "Request access — full case study," point it at a mailto with subject prefilled to "Red Cross case study access" and body to "I'd like access to the American Red Cross disaster relief case. I'm reviewing it for [role]."
- Open it and say so: "Read the full case study," and strip the client-side check.
Open it. Nobody reading your portfolio at 10 PM before an interview emails you for a password.
Order of operations
- Upload
og-cs-03.jpg. Until it lands, every share renders a gray box under a headline about the worst day of someone's life. - H1 and subhead. This is where the eight seconds go.
- The CTA.
- Mobile reorder and alt text. A week out is fine.
- The compliance figure. Holds until you can source it.
- The Red Cross buyer for this case: TRM Labs' current Head of Product Design posting describes a five-person team doing high-stakes investigative work and explicitly rejects management-only profiles, which makes it the single best-matched destination for a Red Cross teaser fixed the way this audit describes.
- The private-essay exposure is the same bug:
what-do-you-count.htmlis still linked from the homepage and still delivers its full text to an unauthenticated browser before the password check runs, which means the client-side gating problem behind the Red Cross CTA is live on at least six surfaces and undercuts a trust-and-authority thesis while it sits there. - Equinox+ metadata is the cheapest remaining teaser fix: while Red Cross now carries a mandate-forward title, the Equinox+ page still ships as the generic "Equinox+ · Teaser · Juno Chen", leaving its five-brand, three-month MVP story invisible to anyone scanning a search result or a shared link.
- Why the intake decision travels further than you think: Maven Clinic's Senior Staff Care Delivery posting names confidence indicators, AI-human handoffs, and progressive disclosure as design-system patterns — the same mechanism as three-fields-not-eight, which makes the Red Cross intake decision reusable evidence in clinical AI interviews, not just nonprofit ones.