You Designed the Mechanism. You Didn't Own the Meter.

The surface question: "BCG Digital Ventures — so that was consulting? Strategy engagements?"

The concern underneath: they are running you against a McKinsey template. Arrive, interview, recommend, present, leave. The deliverable is a slide. And behind that template sits a quieter question, which is the one that decides the interview: has she ever had to live with a decision of her own after it shipped.

The grain of truth: the quieter question lands. Four products, built inside a corporate venture-building firm, on engagements measured in months. You ran none of them afterward.

Most candidates at your level fight that sentence. Fighting it turns a fact into a wound. Say it first, say it tighter than the panel could have said it, and the concession starts reading as the judgment they were probing for. Then spend the remaining fifty minutes on what the concession does not entitle them to conclude.

The custody rule

Every outcome number in those four cases came off instrumentation somebody else owned: a finance model at Thermo Fisher, case reporting at the Red Cross, product analytics at Equinox and Allē. Not one of the four pages names the measurement owner, the baseline, or the calculation period, and inside a venture build with a corporate partner you rarely control any of the three. You designed the mechanism that moved the number. The meter belonged to the partner.

There is no weakness in that. It is what venture-build work looks like from the inside. But it sets your speaking rule: claim the mechanism, attribute the metric. Say a number, say whose number it is, same breath. When they ask you to prove impact, walk toward the mechanism.

The grades below tell you what to do with each claim under pressure.

The stipulation, in the first ninety seconds

Confidence: high. Every load-bearing claim in there is a duration, and durations you can defend line by line.

One sourcing correction before you rehearse it. Do not claim that handoff was BCG DV's universal designed endpoint. I couldn't verify that. The firm's own venture-building overview lays out several structures. Some the partner operates outright afterward. Some keep continuing equity involvement. Anchor on your four durations, not on a firm-wide rule you'd have to abandon under one follow-up.

Two axes, and you don't lose both

Build versus advisory: you win outright. Confidence: high. Live application endpoints, named corporate owners, two products with current App Store version histories, four zero-to-production launches. This is the highest-confidence element in the dossier and you should sound like it. If they keep pushing the consulting frame after your stipulation, they weren't listening, and that is worth noting about the panel.

Build versus operate: part of this you lose. Concede that part immediately and exactly. Same sorting logic as the org-scaling gap in Issue #5: better language doesn't shrink the gap, it sorts roles into the ones your evidence covers and the ones it doesn't.

Don't reach for TinyFish to patch it. Different evidence base, different provenance, and panels hear the reach as evasion.

The frame that carries your weight: handoff durability. Confidence: moderate, and moderate only because you have to show the artifacts rather than assert the property. Products built by teams that never expect to leave tend not to survive the leaving. Yours were built to be inherited, and the portfolio names the infrastructure. A design system and token architecture at Equinox+. At Thermo Fisher, five researched failure modes decomposed into five discrete modules. That is what lets a product outlive its founding team. Two of yours are still shipping under someone else's ownership, which is the test. One I can't verify, and I'll be specific about which one below.

The four cases, graded

Thermo Fisher mySupply

On the page. Twelve months, zero to live. Six pharma partners, nine sites, three continents. 42% overhead reduction and "$20M+ margin recovered annually." One designer, two engineers. You as product design director.

Provenance. Self-published, no independent delivery record. The $20M appears once, inside a case-level outcome band sitting beside business-case return figures. No module attribution. No calculation period. No baseline. No named owner of the model. Your most quotable number is also your least defensible one.

Grades.

• Build, scope, launch: moderate. The endpoint at mysupply.patheon.com still responds, which proves an artifact exists and tells you nothing about adoption.
• The $20M, and any personal attribution for it: use with caution.

Don't say: "My design recovered more than $20 million a year." One question about the baseline and you're improvising.

Say: "The program reported north of $20 million in annual margin recovery across the six partners. I own the design architecture, not the finance model. What I'll defend in detail is the mechanism. Thirty-two interviews surfaced five failure modes, and I turned each one into a module. Exception-first order management came first, because exceptions were surfacing two to four days after commitment, when rescheduling cost three to five times what prevention cost."

You trade away a headline and pick up the thing they are actually grading, which is always a favorable trade.

American Red Cross disaster relief platform

On the page. Six legacy systems into one. Zero to national deployment in six months. 1,689 cases in the first two weeks. $847K disbursed. 98.7% FEMA compliance. Team of eight, with you as design director and general manager.

Use the GM scope. When a panel probes whether you ever owned anything, this is the strongest asset you have: dual design and GM authority over an eight-person cross-functional team through a national deployment. Take it that far and stop. Don't let it drift toward implying P&L ownership across years, because a six-month engagement GM role is not that, and the follow-up will find the seam.

Provenance. The 1,689 figure is the strongest metric in your BCG DV set because it states a window, and a count with a timeframe is checkable. The $847K carries no date range, and the 98.7% has no denominator and no defined compliance test.

Grades.

• Six-month national deployment and six-into-one consolidation: moderate.
• Two-week case volume: moderate, and lead with it.
• $847K and 98.7%: use with caution.

Don't say: "I improved FEMA compliance to 98.7%." And don't say the platform is still running nationally.

Say: "Six caseworker systems consolidated into one, deployed nationally in six months, 1,689 cases in the first two weeks. That's the number I'd defend under scrutiny, because I can tell you exactly what window it covers."

Equinox+

On the page. Zero to MVP in three months across five brands and 600,000-plus premium members. Four designers and a researcher reporting to you. Information architecture, cross-brand navigation, design system and token architecture. 4.8 at launch. Plus the analytical hook: members who followed an instructor inside 30 days retained at close to twice the rate of those who didn't, which is what drove instructor-as-navigation and the decision to sequence three brand worlds ahead of the other two.

Provenance. The retention finding states its exposure window and nothing else. No outcome horizon. No cohort size. No measurement owner. The launch rating states a state with no count and no capture date.

Grades.

• Three-month MVP and design ownership: moderate.
• The 4.8 at launch: moderate.
• Nearly-doubled retention: use with caution.
• Current availability: high. The app is still sold and shipping, and Apple's app metadata shows a version released in July 2026 at roughly 4.7 across roughly 42,000 US ratings.

Don't say: "My navigation doubled retention." And don't offer today's rating as validation of your launch design. It belongs to everyone who has shipped since, which puts it on the operate side of the ledger.

Say: "The retention signal was a decision input, not an outcome I'm claiming. It told us the instructor was the navigational unit, and it told us which brands to sequence first."

Allē

On the page. Three designers under you on a two-sided loyalty product. 30 million-plus members, 40,000-plus provider practices, 3.2x redemption, 47% lapsed-member reactivation, CAC from $92 to $42, in-app planners converting at 2.4x front-desk enrollment.

Provenance. Your thinnest set. No cohort, denominator, sample size, or attribution method on any of it, and two unrelated 47% claims the page never distinguishes. The scale figures don't reconcile: 30 million-plus in one place, 18 million in the page's own narrative, and AbbVie's current help center saying more than seven million. Three population definitions across three dates, possibly. The page doesn't say.

Grades.

• Design scope and team role: moderate.
• Every outcome metric: use with caution.
• The 30-million figure: use with caution, and retire it from spoken answers until you reconcile the page.

Don't say: "I cut CAC from $92 to $42." Don't say Allē has 30 million active members.

Say: "The mechanism is what I'd walk you through. First-timer drop-off peaked in the six weeks after a first visit, so we triggered expiry notifications at 60 days and moved enrollment off the front desk into an in-app planner. The instrumentation wasn't mine, and I don't have the cohort methodology in front of me."

That last clause is free, and it buys credibility for everything around it.

Durability, corrected

In Issue #1 I handed you "Four products. Still live. Still adopted." It overstated the evidence. I'm retiring it.

Verified August 1, 2026:

• Equinox+ and Allē: publicly available and actively shipping.
• mySupply: the endpoint responds, which confirms an artifact and nothing further.
• Red Cross: cannot be identified in any public source. Not evidence it's gone. Not evidence it survived.

Two of four verified, a third still answering the door, one unknowable from outside. Two verified survivors years after the founding team walked out is a strong result on its own, and it's the version that holds up against a search bar. Confidence: moderate on the two verified survivors, use with caution on anything broader.

Say: "Two of the four are publicly available and still shipping. A third's application is still reachable. The fourth was an internal system I can't verify from outside, and I won't claim I can."

Four of four collapses under one search, and it takes every other answer you gave with it.

The break condition

One role profile defeats all of this. Learn the shape of it, and decline the frame instead of stretching to fit.

Wellsheet's Staff Product Designer posting, live as of August 1, 2026 and useful here as an archetype rather than a target, says it in plain words: "You've been the lead designer on a particular product for multiple years."

Read it precisely, because precision protects you in both directions. It asks for multi-year ownership of one product. It does not ask for multi-year tenure at one company. Concede what the sentence asks and not a syllable more. But if the panel means what the sentence says, no reframe closes that gap.

Say: "I've led several products from zero to production. I haven't been the in-house lead designer on a single product for multiple years. If that sustained ownership is the core of this role, my portfolio doesn't prove it, and I'd rather we both know now."

Confidence: high. That sentence loses you some searches. It wins the ones where the panel was checking whether you know the difference.