ProductAI
We tested BiteCal against a dietitian. Here are the numbers.
A 200-meal head-to-head between BiteCal and a registered dietitian, with break-downs by cuisine, complexity, and lighting.
Marketing pages love to claim “99% accuracy.” We think that number is meaningless without a study you can read. So we ran one.
Methodology
We recruited Anjali Deshpande (RD, 12 years practice) and had her log 200 meals using her usual method: weighing ingredients, cross-referencing IFCT 2017. Same meals were photographed and scanned by BiteCal. No retakes, no edits.
Results
- Mean absolute error on calories: 9.4%
- Mean absolute error on protein: 11.2%
- Mean absolute error on carbs: 8.1%
- Time per meal — dietitian: 4m 12s. BiteCal: 2.8s.
Where BiteCal lost
BiteCal underestimated oil in three dishes by 40-60%: dum aloo, pindi chana, and bhatura. All three hide oil in the gravy. We're adding an “oil-heavy” tag for users to toggle.