Curl diagnosis is solved. Nobody's cross-referencing what's coming out of the tap.
Several tools already read curl pattern and porosity from a photo. None of them factor in the mineral load of your local water supply, the single biggest environmental variable affecting how Type 3–4 hair actually behaves week to week.
Run a sample diagnosis
The diagnosis half of this problem is already being solved well. The environmental half isn't being touched.
Photo diagnosis exists
Several tools already map a photo to curl pattern and porosity and route to a product recommendation. That layer is table stakes, not a differentiator on its own. Classification approach is grounded in academic precedent — Figaro-1k, a peer-reviewed open dataset of 1,050 annotated hair images, established multi-class hairstyle classification (straight/wavy/curly/kinky) as a validated computer-vision task.
Water data exists separately
UK postcode water-hardness checkers are common, and hard-water-vs-hair content is widely written about. But it lives in a different app, disconnected from any personalised diagnosis.
Nobody fuses both into one engine
Combining texture/porosity diagnosis with live local water chemistry into a single recommendation model is the actual white space, and the core technical claim this build is designed to prove out.