Iron deficiency is one of the most common nutritional deficiencies in the world, and until now checking for it has meant the same three steps every time: book an appointment, sit for a blood draw, wait days for a number to come back.

Chromacare Labs set out to collapse that into ten minutes on a kitchen table. The test itself is a lateral flow assay — a strip that develops a coloured line in proportion to the ferritin in a single drop of blood. The hard part was never the chemistry. It was reading it. A faint line means nothing until something converts that colour into a defensible number.

247 Labs architected, designed and developed the Chromacare App® for iOS and Android to be that reader. The app guides the user through the test step by step, captures the developed cassette under controlled camera settings, calibrates the image against a printed colour reference card in the same frame, and returns a quantitative ferritin measurement — not a positive/negative line-read.

Every consumer smartphone camera lies, and it lies differently. Auto-exposure brightens a dim room. Auto-white-balance shifts warm kitchen light towards neutral. Two users photographing identical cassettes under a window and under an LED bulb produce two different images — and on a colourimetric assay, that difference is indistinguishable from a difference in the patient’s blood.

Chromacare needed a reader that would behave consistently across an unknown range of phones, lighting conditions and user technique, without shipping dedicated hardware to every customer. Three problems had to be solved at once:

• Capture control. Consumer camera apps optimise for a nice-looking photo, not a measurable one. Automatic adjustments destroy exactly the signal the assay depends on.
• Normalisation. Even with fixed capture settings, the app has to know what “true” colour looks like in this photo, in this room, on this phone.
• Signal extraction. The control and test lines on a developed strip are subtle gradients on a background that is itself uneven. A number pulled from raw pixel values without correction is noise dressed up as a result.

On top of the measurement problem, this was a regulated consumer health product. The guided test flow, the timing constraints, the failure states and the results history all had to be built to a protocol, not to a designer’s preference.

A guided test-taking flow. The app walks the user through the full procedure — kit acclimatisation, buffer and blood collection, applying the sample to the cassette well — with a built-in timer that gates the read. The developed strip is only valid inside a defined window, so the app enforces it rather than trusting the user to remember.

Deterministic RAW capture. When the timer completes, the app takes over the camera. It captures a RAW image at a fixed shutter speed and fixed ISO, with autofocus and white balance handled explicitly and the file saved at maximum fidelity. Nothing about the exposure is left to the phone’s opinion of a good photograph. The user simply frames the colour card; the app does the rest.

Colour-card calibration. The cassette is photographed sitting on a printed reference card carrying greyscale and colour swatch ladders plus machine-readable fiducial markers. The pipeline locates the card, reads the swatch ladders, and uses them to normalise the frame for that specific combination of light source, camera sensor and white balance. This is the piece that makes a $200 phone and a $1,400 phone agree on the same strip.

Signal extraction and baseline correction. With the frame normalised, the pipeline isolates the test window, verifies it is clean and readable, and builds an intensity profile across the strip. It identifies the control (C) and test (T) lines, samples the local background on both sides of each line, and corrects each peak against its own baseline before computing the T/C ratio — the value the concentration curve is built on. Replicates are averaged with their standard deviation reported, so the pipeline’s confidence in a reading is measurable rather than assumed.

Honest failure states. A test window that isn’t clean, a card that isn’t fully in frame, a photo taken outside the valid time window — each produces an explicit error and a clear instruction, not a silent bad number. In a diagnostic product, refusing to answer is a feature.

Results the user can act on. Readings are stored to the user’s profile, tracked over time, and shareable with a healthcare provider. 247 Labs also delivered Chromacare’s Shopify storefront so the kit and the app sit inside one coherent brand experience.