How accurate is a photo, really?

The short version

For typical home-cooked and restaurant meals, VitaMenda is best thought of as a directional estimate, not a precise measurement. Calorie and macro estimates from a photo are commonly within 15–25% of a weighed reference; micronutrient estimates carry a wider band.

That accuracy is more than enough to surface patterns across two weeks — a chronic magnesium gap shows up clearly even with noisy per-meal data. It is not enough to make clinical decisions from a single meal.

Where the error comes from

• Portion estimation — the biggest single source of error. Plate geometry only tells you so much; a calorie-dense sauce can hide under vegetables.
• Hidden ingredients — oils, butter, sugar, and dressings used in cooking are often invisible in the photo.
• Food-composition variability — the same food (spinach, tomato, beef) varies in nutrient content depending on variety, soil, season, and cooking method.
• Mixed and processed dishes — curries, casseroles, packaged foods. Harder than a piece of grilled salmon.

Where we are most reliable

• Single-ingredient or simple plated meals.
• Trend signals over a 14-day window — the rolling average is robust to per-meal noise.
• Relative rankings (which nutrient is your biggest gap).

Where we are weakest

• Exact micronutrient values for any single meal — treat them as rough.
• Sugar and added oils in restaurant cooking.
• Drinks photographed without context.

The 2026 benchmark

We are preparing a public accuracy benchmark — VitaMenda vs. weighed reference, vs. manual logging in MyFitnessPal and Cronometer — on 50–100 real meals. We will publish the methodology, the errors, and the examples, including the meals we got wrong. Sign up for the waitlist to be notified when it's out.

What we are doing about the gaps

Manual portion correction is built in. Ingredient adjustments propagate to the rolling 14-day window. Future versions will add optional context (cooking method, sauce confirmation) to tighten the harder cases.