One persona is out, a second one already sits in the data model.
Four user interviews later it was clear: one of the three persona hypotheses missed the target segment. What we changed, and why the second persona already sits in the data model even though the UI barely shows it today.
Struqo started with three persona hypotheses. Bernd, the DIY builder with timber experience. Winni, the ambitious self-builder without a carpentry background. A third persona for classic furniture self-build. Four interviews later, the furniture persona is out.
What the furniture persona showed
The hypothesis was that furniture self-build and timber construction in the garden or extension create the same tool expectations. The one out-of-target interview did not back that up. Someone building a bookshelf or an IKEA hack works with different tools, materials, and tolerances than someone planning a carport. When a person from the furniture segment says "the editor is too complicated for me", you learn nothing about Struqo, only about furniture assembly.
In the analysis we logged the furniture persona as out-of-target. Methodically clean, more honest than mixing its data in with Bernd and Winni. In the portfolio writeup it stands as its own section, with the reasoning for why it does not feed into the clustering.
Why Klaus already, now
Klaus is new on the persona list, yet shows up in no interview. Klaus is the pro: a small trade business, one to five employees, no Cadwork budget, planning today with pencil and Excel. He is a secondary persona, not a primary one. Still, he is in the data model from day one.
The reason is mundane and expensive at once. Once Struqo is built on a Bernd/Winni architecture and Phase 3 then wants to stack pro features on top for Klaus, migrations break open. Float32 instead of Float64. Snap fixed to 100 mm instead of scalable. Materials with no manufacturer field. A layer system that only knows "wall" and "beam", but no dimension layer.
So we pull the architecture decisions in now, even if Klaus makes not a single click today.
What changed in concrete terms
- Snap granularity is selectable (1, 5, 10, 50, 100 mm), not hardcoded to whatever value is convenient for Bernd right now
- Numeric input for every geometry as an equal path next to drag-and-drop, because Klaus types values instead of dragging
- The material record already carries fields for manufacturer, batch, strength class, even if the UI does not show them today
- Coordinates internally in millimetres with Float64
What we did not change
The landing page. The editor. The onboarding. Klaus lives in the data layer, not in the language. The moment we start addressing DIY builders and pros at the same time in our outward communication, we lose both. Bernd reads "DIN-compliant structural design" and closes the tab. Klaus reads "quick to your first plan" and does not take it seriously.
Two modes, one voice. That is the rule of thumb.