Recipe/Example/Red Genius
Product
OFF: Red Genius
(0.8, 0.2, 14.0, 13.0, 1.9, 0.7, 0)
ingredients: 23.8 % purĆ©e de banane, 20 % jus de raisin, purĆ©e de pomme, jus de pomme, 7.5% purĆ©e de canneberge, jus d'orange, 3,2 % purĆ©e de grenade, 3 %, purĆ©e de baies d'aƧaĆÆ, jus d'aronia, jus de citron.
Ciqual
23.8 % purƩe de banane (13005) (0,0, 19.7, 15.6, 2.7, 1.06, 0)
20 % jus de raisin (2016) (0.057, 0.031, 16.3, 16, 0.14, 0.25, 0.0089)
purƩe de pomme (13038) (0.21, 0.017, 24.4, 20.7, 1.53, 0.23, 0)
jus de pomme (2074) (0.14, 0.046, 11.4, 10.9, 0.23, 0.17, 0.022)
7.5% purƩe de canneberge (13113) (0.13, 0.011, 7.6, 4.04, 5.13, 0.75, 0.02)
jus d'orange (2070) (0.11, 0.053, 9.61, 9.61, 0.28, 0.61, 0.012)
3,2 % purƩe de grenade (1.2, 0.12, 14.7, 13.7, 4, 1.7, 0.0075) (carbs corrected)
3 %, purĆ©e de baies d'aƧaĆÆ (5.3, 0.88, 5.3, 0, 5.3, 1.8, 0.065)
jus d'aronia (0, 0, 14, 9.7, 0, 0, 0)
jus de citron (2028) (0.29, 0.038, 5.7, 2.4, 0.4, 0.4, 0.053) (carbs corrected)
Results
Fit deviance 0.65, largest residual sugar. Many bound ingredients.
-
Recipe fit -
Nutritional fit -
Stretch
Claude solver
Carbs and sugar fit well (2% each) ā the expected result for a pure-fruit smoothie where CIQUAL fruit values are reliable. The three "large" errors are percentage artefacts of tiny absolute values: the entire fat gap is 0.62g per 100ml, which traces almost entirely to the aƧaĆÆ proxy. CIQUAL 2020 has no aƧaĆÆ entry; a realistic value (~10g fat/100g for frozen pulp) would account for most of the gap. The analysis section and CIQUAL table flag this explicitly.
One structural note worth highlighting: all 10 fruit ingredients score WEAK on identifiability. A pure-fruit smoothie is the hardest case for nutritional recipe solving ā every ingredient has the same macro signature (mostly sugar/water, negligible fat and protein), so the solver cannot separate them from nutrition alone and relies entirely on the 4 declared fractions.
Analysis
Carbs, sugar and protein fit well (all under 10%). This is the expected result for a pure fruit smoothie: all ingredients are fruit purĆ©es or juices with very similar macro profiles dominated by natural sugars, and the CIQUAL entries for common fruits are reliable. The yield of 96% (S = 104.0g per 100ml product) is essentially unity ā no cooking, no evaporation, just blending.
All 10 fruit ingredients flag as WEAK in the identifiability check. This is structurally unavoidable for a pure-fruit smoothie: banana, grape, apple, orange and pomegranate all have very similar fat/protein/carbs profiles, making them nutritionally indistinguishable from each other. Only the 4 declared percentages (banana, grape, cranberry, pomegranate) anchor the recipe; the remaining 6 ingredients are positioned by the ordering constraint alone. Aronia juice and lemon juice are driven to zero because the carbs/sugar budget is fully met by the other ingredients.
The three large errors (fat 77%, sat_fat 84%, salt 71%) are misleading in relative terms because the absolute values are tiny: the fat gap is only 0.62g, sat_fat 0.17g, salt 0.007g per 100ml. The fat and sat_fat errors trace entirely to the aƧaĆÆ proxy. AƧaĆÆ (Euterpe oleracea) is one of the most fat-rich fruits ā fresh frozen pulp carries 5ā15g fat per 100g ā yet the blueberry proxy (CIQUAL 13028) assigns only 0.33g fat per 100g. At 3% of the recipe, using a realistic aƧaĆÆ fat value of ~10g/100g would contribute 0.3g extra fat, accounting for roughly half the 0.62g gap. The remainder likely comes from trace fats in banana and grape that CIQUAL underestimates for the specific varieties used. CIQUAL 2020 does not have an aƧaĆÆ entry; adding one would substantially improve this fit.
The salt gap (0.007g absolute) is analytically negligible ā it is within the typical measurement uncertainty of sodium declarations on fruit-based products.