Recipe/Example/Poulain Grand Arome
Product
OFF: https://world.openfoodfacts.org/product/16256866/poulain-grand-arome
Ingredients: Cacao maigre en poudre (50%) (beurre de cacao (11%)), sucre, arôme
Nutrients: (5.8, 3.6, 55, 47, 17, 11, 0.02)
Ciqual
Cocoa powder, without sugar, powder, instant (18100) (20.6, 12.4, 11.6, 0.9, 29.5, 22.4, 0.11)
Sugar, white (31016) (0, 0, 99.8, 99.8, 0, 0.0054)
Aroma (-) () no data
Analysis
Base analysis
The base analysis uses the ingredients as found on the package. The deviance is very large (34.5).
The Fat, Saturated Fat seems overestimated, whereas fiber is underestimated. This might be due to the fact that NO fat-reduced cocoa powder has been used.
Secondary analysis
There is no reduced fat cocoa powder entry in Ciqual. Could the fat reduction be estimated?
If we split the cocoa powder in two components: Cocoa powder fat, which is responsible for fat and saturated fat, and fat free cocoa powder, which is responsible for the other nutrients.
Thus we have two new ingredients:
Cocoa powder fat (20.6, 12.4, 0, 0, 0, 0, 0)
and
Fat free cocoa powder (0, 0, 11.6, 0.9, 29.5, 22.4, 0.11)
As sugar is already a large ingredient, these two new ingredients are either rank 2 or 3. We can try to fit both possibilities to see shich one gets the best fit.
| Ranking | Deviance |
|---|---|
| Sugar / Fat / Cocoa | 34.4 |
| Sugar / Cocoa / Fat | 11.6 |
| Cocoa / Sugar / Fat | 4.58 |
| Fat /Sugar / Cocoa | 34.5 |
Thus the ranking of Fat-free cocoa powder, Sugar and Cocoa fat gives the best fit.
With the 3-ingredient fit, we can determine how much fat was removed from the cocoa. The amount of cocoa was fitted at 55% and the fat at 28%, a difference of 27%, which is the amount of fat removed.
Results
-
Blame graph original -
Blame graph deconstructed
Claude Solver
With only two nutritionally significant ingredients (cocoa powder and sugar), both having declared percentages, this product is analytically trivial for the recipe solver: the declared fractions completely determine the recipe, and the solver residual of 0.027 confirms the CIQUAL values are consistent with the declared 50% cocoa / 47.5% sugar split.
Fat proxy adjustment: the label states "beurre de cacao (11%)" — meaning the cocoa powder in this product retains only 11% fat, whereas CIQUAL 18100 has 20.6% fat. The solver uses the label-declared 11% fat content (sat_fat scaled proportionally from CIQUAL), giving 5.5g fat per 100g dry product. This is roughly half the CIQUAL default and illustrates why declared fat information on ingredient sub-components matters for accurate modelling.
The preparation analysis exposes a protein anomaly in the label. Using 27.2g powder and 72.8g whole milk per 100g prepared drink (ratio anchored on carbs), the model predicts 5.41g protein — but the label declares 8.40g. This 3g gap cannot be reconciled with any reasonable preparation ratio without breaking the carbs constraint. The declared protein value is likely either an error or represents a different preparation (e.g., fortified milk or a different beverage base).
Fibre is similarly anomalous: CIQUAL attributes 29.5g fibre per 100g cocoa powder, which predicts 4.0g fibre in the prepared drink, but the label declares 2.5g — suggesting the cocoa used has lower fibre than the CIQUAL average.