How every score is worked out.
No black boxes. Here is what each number means, what it is based on, and where it can be wrong.
Nutri-Score
A letter from a to e summarising nutritional quality, following the European Nutri-Score system. Vegetables, fruit, legumes, plain fish and whole grains sit at the top; sugary drinks, confectionery and deep-fried snacks at the bottom.
Eco-Score
The same a–e scale applied to environmental impact. Local plant foods score best; red meat, shellfish and heavily packaged industrial products score worst. It is an estimate of production impact, not a measure of your individual footprint.
NOVA group
A one-to-four classification of how processed something is. 1 is whole or minimally processed — raw fruit, plain meat, milk, eggs. 2 is culinary ingredients like oil, butter and sugar. 3 is processed food such as cheese, fresh bread and cured meat. 4 is ultra-processed: packaged snacks, soft drinks, ready meals.
Glycemic index
A 0–100 estimate of how quickly something raises blood sugar. Anything without meaningful carbohydrate — plain meat, fish, eggs, cheese, oil, tea, coffee — is zero. Below 55 is low, 55–69 medium, 70 and up high. Shipshape also totals a glycemic load across your day, which accounts for portion size as well as index.
Fibre
Grams per 100g, flagged as high-fibre at 6g or above. Lentils, beans, oats, chia, whole-grain bread, raspberries, broccoli and almonds typically clear it; meat, dairy, white rice and most desserts do not.
The sleep score
A single number out of 100 built from three things: how long you slept, how consolidated it was, and how the stages balanced out. Duration alone is measured against your effective sleep — time in stretches long enough to be restorative — rather than everything the watch logged.
Continuity covers how long you took to fall asleep, how much you were awake afterwards, how many times you properly woke, your longest unbroken stretch, and how dense the night was overall.
Where a measurement genuinely is not available — no in-bed data means sleep latency cannot be known — its weight is redistributed across the others rather than scored as zero, so a gap in your watch's data does not quietly cost you points.
Photo and description estimates are exactly that — estimates. They are good enough to spot patterns over weeks and not precise enough to weigh a bodybuilding cut by. Barcode data comes from public product databases and can be out of date. Sleep staging is only as good as the device that recorded it.
See it on your own food.
The scores make more sense against things you actually eat.
Download for iPhone