I BUILD THE THING.

Problem Manufacturers that keep their machines across a whole life see how those machines wear and where they are hard to work on. Very little of that reaches the people designing the next one. Why?

What I did A qualitative embedded case study with Johan Karlström. 23 respondents, 6 workshops, and a mapping tool we built because nothing was capturing the comparison we needed. We collected the data together. I built the theory and the eight-condition framework. Johan built the methodology and categorised the data against it.

Result An eight-condition framework and a two-threshold model for diagnosing where end-of-use knowledge stops. Visibility and information flow rest on different organisational conditions, so a product-service business model can deliver the first without the second.

Why it matters now The EU's Ecodesign rules, the Digital Product Passport, and Right to Repair all land in July 2026. Each assumes manufacturers can get information from the installed base back into design. That is what this framework diagnoses.

What I'd do differently We stayed on information flow. Next to it I would want the economic models, because for this product the open question is whether circularity pays at all.

Why it exists I was turned down for a growth role for having B2C depth rather than B2B. Reading about it only teaches the vocabulary, so I wrote the judgement down instead: for each play, when it applies, what has to be true first, which number it should move, and how you would know afterwards.

Problem B2B funnel problems are easy to misread, and most advice jumps to tactics without saying when they apply.

What I built An engine that reads a funnel, names the biggest problem the evidence supports, recommends one play, and shows its reasoning. 31 research assets synthesised into 60 funnel signals, 33 play specifications, and 30 decision cases.

How A seven-stage TypeScript pipeline with 230 automated tests. Splitting architecture, implementation, and independent verification across three AI models cut the API-equivalent build cost by about 60% across six measured slices. Still building it.