Five questions before a line goes public
- Technical truth
- What does the product do, and where does it stop?
- Buyer consequence
- What materially changes for the buyer?
- Evidence
- What is the denominator, comparison, sample, and material caveat?
- Category constraints
- Which safety, regulatory, workflow, or liability boundary governs the language?
- Competitive permission
- Can this company credibly own the line, or is it repeating a category claim?
Claim 1 of 4
- Developer infrastructure · called
- AI workflow · called
- Clinical operations · called
- Payments · called
Developer infrastructure
Most qualifying failed deployments roll back before an operator has to intervene.
- The product detects failed health checks and returns a service to its last healthy release.
- Across 4,800 qualifying failed deployments in a 12-customer beta, 93% rolled back without operator action.
- The definition of qualifying, the exceptions, and the beta sample can all appear in proof.
- In this fictional market, the closest alternatives alert but do not act.
My call: Ship
Same call. The boundary is clear enough to publish.
Here is the boundary I would not cross. I think this line stays on the right side of it.
The action sits inside the product's control. The word most keeps the claim within what the product can prove. The proof can carry the denominator and exceptions; the headline does not need the sample size.
Most qualifying failed deployments roll back before an operator has to intervene.
Lenses that decided itTechnical truth · Evidence · Competitive permission
West of Obvious · original public model
AI workflow
Automate vendor-risk decisions with AI.
- The system drafts assessments from submitted documents and policy controls.
- Sixty-one percent of low-risk supplier assessments were accepted with minor edits.
- Every draft links to its source material.
- A human reviewer owns the decision; high-risk suppliers were excluded.
My call: Rewrite
Same call. There is a useful claim here, but not this one.
The product has earned a claim. It has not earned this claim.
The original turns drafting into decisioning, hides the low-risk scope, and erases the human authority the product intentionally preserves. Narrower is not weaker here; it is the differentiator.
Give reviewers a source-linked first draft for routine supplier assessments, while every decision stays human.
Lenses that decided itTechnical truth · Buyer consequence · Category constraints
West of Obvious · original public model
Clinical operations
Prevent avoidable readmissions with predictive AI.
- The model ranks readmission risk from discharge data.
- Validation is retrospective.
- No prospective evidence shows that care teams change interventions or patient outcomes.
- The intended-use boundary and definition of avoidable are not yet settled.
My call: Escalate
Same call. This is not a copy problem yet.
Marketing cannot close this gap with a qualifier.
The product predicts risk. Prevention belongs to the care pathway, not the model alone. Copy cannot bridge an unresolved clinical and evidentiary boundary.
The clinical lead and data-science owner must define intended use, acceptable performance, and the intervention workflow before product marketing chooses a market claim.
Lenses that decided itEvidence · Category constraints · Technical truth
West of Obvious · original public model
Payments
Eliminate payment fraud.
- The product enforces approval paths, flags unusual routing, and records audit events.
- Banks and counterparties still settle transactions.
- The product cannot control social engineering or a compromised external account.
- No evidence supports a zero-fraud outcome, and every competitor makes some version of prevention.
My call: Kill
Same call. The line is not repairable.
A softer verb would not fix the strategic problem.
The claim is absolute, outside the product's control, unprovable, and not ownable. Adding help or reduce would soften the grammar, not fix the strategy.
Own approval integrity, control evidence, or investigation speed instead.
Lenses that decided itTechnical truth · Evidence · Competitive permission
West of Obvious · original public model
Each product needs its own answer. The same standard still applies.
Category fluency is not a vocabulary test. It is knowing what transfers, what must be learned, and when copy cannot close the gap.
A strong product marketer does not reuse the answer across categories. She reuses the discipline for earning the right to make one.
| Category | Candidate claim | Call | Deciding boundary |
|---|---|---|---|
| Developer infrastructure | Most qualifying failed deployments roll back before intervention | Ship | Product control and publishable proof align |
| AI workflow | Automate vendor-risk decisions with AI | Rewrite | Drafting is not decisioning |
| Clinical operations | Prevent avoidable readmissions with predictive AI | Escalate | Outcome depends on unresolved clinical workflow and validation |
| Payments | Eliminate payment fraud | Kill | Absolute, outside product control, unprovable, unownable |
West of Obvious · original public model
Design notes
The idea. A new category changes what must be learned. It does not remove the obligation to know where the product stops, what the evidence proves, and who has authority over the unresolved parts.
Escalation is a decision. Sending a claim back to a clinical, technical, legal, security, or financial expert is not indecision. It is recognizing that marketing cannot write its way around a missing product or evidence boundary.
The public boundary. This page shows the judgment calls. The deeper research, interview, validation, and category-entry system behind them is not published.
How it is built. One HTML file, vanilla JavaScript, no framework, no libraries, no build step. Native radio groups, visible progress, a screen-reader status line, reduced motion, and a no-JavaScript fallback.
Original public model · West of Obvious