In this LAB note
Question · Hypothesis · Method · Results · What Failed · What We Learned · What Changed at MOMS
Question
Do MOMS's eight body-type profiles have an equal shot at a top-tier Score, and if not, is fat content the reason?
Hypothesis
We expected the 11-to-20 Score scale to rate recipes on their own merits, independent of which body-type profile they were built for. If certain body types clustered lower, we assumed it would be a fixable formulation gap — most likely under-dosed fat or protein, since both are commonly linked to satiety and diet-quality scoring in the nutrition literature — rather than something built into how the recipe system itself is structured.
Method
We pulled Score, Cal Label, Goal, and full macros for all 201 live MOMS recipes and grouped them by their eight body-type (BT) codes: EY, EI, WY, WI, MY, MI, WAY, WAI. For each BT we calculated the mean Score, the Score ceiling (max value reached), and the share of recipes landing in the top band (Score ≥18 of 20). We then ran Pearson correlations between Score and each macro (Protein, Fat, Carb, Fiber, Kcal) across all 201 recipes to test whether any single macro predicts Score well enough to act on. Published validation work on nutrient-profiling systems informed how we read a weak macro-Score correlation: Poon et al. (2018, British Journal of Nutrition) found that fat-weighted profiling models show their least reliable agreement precisely in high-fat food categories (κ=0.07, 33.8% discordant classifications for fats/oils under the FSANZ-vs-Ofcom comparison), and Drewnowski (2005, American Journal of Clinical Nutrition) noted that "healthful" is routinely defined by the absence of fat, sugar, and sodium rather than the presence of any benefit — a reminder that fat's relationship to a composite quality score is rarely linear or standalone. Last, we cross-tabbed Goal against BT to check whether body type and Goal move independently in our own system, or together.
Results
Mean Score by body type ranged from 13.75 (WY, n=28) to 16.13 (WAI, n=24) on the shared 11–20 scale. The share of recipes reaching the top band (Score ≥18) was 3.3% for EY, 6.7% for EI, 3.6% for WY, 4.0% for WI, 5.3% for MY, 5.3% for MI, 19.2% for WAY, and 16.7% for WAI — WAY and WAI (50 of 201 recipes) hit that band five to six times more often than the other six body types combined (151 recipes). Fat share of calories tracked the same split: WAY and WAI recipes average 29.7% and 30.7% of calories from fat, versus 18.9%–25.7% for the other six BTs. But across all 201 recipes, Fat (g) correlates with Score at only r=0.209 — the strongest of any macro we tested (Protein r=0.07, Carb r=-0.178, Fiber r=-0.171, Kcal r=0.022) — meaning fat share accounts for roughly 4% of the variance in Score. Individual recipes confirm the weak link: WY-W2-COOL-001-V1 ("Chill Balance," 27.3% fat) is WY's single highest-scoring recipe at 18 — a fat share close to WAI's average — yet it still can't cross into the 19–20 range that four other BTs reach routinely. Meanwhile WAI-W1-WARM-012-V1 ("Ember Rise") runs 38.3% fat, higher than any WY recipe in the set, and scores only 13. WAI's actual top scorers — WAI-W2-PROTEIN-001-V1 "Power Recover" (Score 20, 29.0% fat, 32.9g protein) and WAI-W3-NUTR-001-V1 "Antioxidant Glow" (Score 20, 32.5% fat, 36.5g protein) — pair fat with higher protein (32.9–36.5g) than WY-W3-BALANCE-002-V1 "Centered Glow" manages at a much lower 17.5% fat and still a respectable Score of 16. The cross-tab explains why fat alone can't be the lever: of our 28 named Goals, all but one ("Light & Vital Renewal," shared by WY/WI/WAY/WAI) belong to exactly one body type. Goal and BT are almost perfectly confounded by how the 201-recipe library was built — you cannot change a body type's fat share without also changing which Goal templates it draws from.
What Failed
Our working assumption — that a body type scoring lower could formulate its way up by adding fat — does not survive contact with the data. Fat is the best single predictor we have and it still explains only about 4% of Score's variance; WY's own highest-fat, highest-scoring recipe matches WAI's average fat share and still caps at 18, while a WAI recipe at 38.3% fat scores near the bottom of the whole set. We also cannot isolate "body type" as an independent variable at all: because Goals are assigned almost one-to-one to body types in this recipe library, the Score gap we're reporting is inseparable from which macro-template Goals a given body type happens to draw from. We do not have evidence that WY recipes are lower-quality food — only that they're structurally boxed into templates that this Score scale rates lower.
What We Learned
The Score gap between body types is real and reproducible — WAY and WAI land in the top band five to six times more often than the other six types — but it is a byproduct of which Goal-macro templates were assigned to which body types when the recipe library was designed, not a per-recipe formulation lever, and not proof that some body types get worse food. A composite score built for cross-recipe comparison can carry structural artifacts from upstream design choices, exactly the kind of discordance the nutrient-profiling validation literature has documented in other scoring systems.
What Changed at MOMS
We're adding a body-type-relative Score band alongside the shared 11–20 scale, so a WY recipe at 18 (its practical ceiling) reads as "top-tier for its template" rather than looking like a plain 2-point shortfall against a WAI recipe at 20. New Goal templates now get checked against the existing BT-to-Goal assignment table before launch, so any future disjointness between body types and macro targets is a deliberate design decision, not an artifact we discover a year later in a data pull.
References
Poon T, Labonté MÈ, Mulligan C, Ahmed M, Dickinson KM, L'Abbé MR. "Comparison of nutrient profiling models for assessing the nutritional quality of foods: a validation study." British Journal of Nutrition. 2018;120(5):567-582. | Drewnowski A. "Concept of a nutritious food: toward a nutrient density score." American Journal of Clinical Nutrition. 2005;82(4):721-732.
Keep reading
Previous experiment
Not sure which shake fits you? We match the formulation to your body type.
Find your body type