In this LAB note
Question · Hypothesis · Method · Results · What Failed · What We Learned · What Changed at MOMS
Question
Does our internal Score reward the Functional slot itself, or does the slot simply travel with something else the Score likes?
Hypothesis
We expected most of the gap to be an artifact. Recipes with a Functional add-in carry about 53 grams less produce (72 g against 125 g) and 38 grams less protein-slot food (209 g against 247 g), and across the full sheet, produce grams correlate negatively with Score (r = -0.18). If the Score quietly favors lighter recipes, Functional shakes would look better only because they are lighter, and the gap should vanish once body type and produce grams are held constant. There is also a reason to expect some real variety credit. McDonald et al. (2018) found that the number of different plant types people reported eating in a week was associated with the diversity of their gut microbiome, so a rubric built around variety would plausibly favor an extra ingredient. We treat the Score as a black box here and only measure what it moves with.
Method
We read all 201 recipes in the MOMS database. A recipe counts as having a Functional add-in when its Functional column lists at least one ingredient: 124 do, 77 do not. We compared mean Score overall, then inside each of the eight body-type codes (weighting each code by the smaller of its two groups), then inside four bands of produce grams (under 60, 60 to 120, 120 to 200, over 200), and finally across the number of Functional ingredients (0 to 3). Recipes in the database are variants that share ingredients, so they are not independent samples. Every figure below is descriptive, and we report no significance tests. The outside source is McDonald D, et al., mSystems 2018, an analysis of self-reported diet in the American Gut cohort that linked eating more than 30 plant types per week with greater gut microbial diversity. It describes people eating across a week, not a single shake, and we cite it only as the reason a variety-minded scoring rule is plausible.
Results
Overall: 124 recipes with a Functional add-in average a Score of 15.23. The 77 without average 14.57. The raw gap is 0.66 points. It rises with the count: zero Functional ingredients 14.57 (77 recipes), one 15.05 (78), two 15.51 (37), three 15.63 (8). Inside body types the weighted gap falls to 0.49. By code: EI +1.04, WI +1.05, MI +0.94, EY +0.87, MY +0.21, WAI +0.14, WY +0.04, WAY -1.33. Holding produce grams roughly constant, the Functional group scores higher in three of four bands: under 60 g, 15.27 against 14.50; 60 to 120 g, 15.44 against 15.00; over 200 g, 15.25 against 14.11 (only 4 and 9 recipes). In the 120 to 200 g band it is lower, 14.25 against 14.40. Produce does not explain the gap on its own, but the two groups differ in other ways. Functional recipes carry 11.7 g of fat against 7.8 g, nearly the same protein (33.2 g against 32.8 g), and slightly less fiber (7.5 g against 7.8 g). Two recipes show the spread. EI-W2-WARM-001 (Score 20) has Blueberry 55 g and Carrot 25 g, Egg White 195 g, and Cashew 19.5 g as its Functional ingredient: 387.8 kcal, 37.0 g protein, 14.9 g fat, 5.7 g fiber. EI-W4-RENEW-007 (Score 12) has no produce and no Functional add-in, just Chickpeas 30 g, Lentils 30 g, White Beans 30 g, and Egg White 195 g: 317.5 kcal, 32.3 g protein, 5.0 g fat, 9.6 g fiber. The lower-scoring recipe has the more fiber.
What Failed
The pattern is not universal. In WAY, the 7 recipes without a Functional add-in average 16.86 and the 19 with one average 15.53, a gap of -1.33. In WY (+0.04) and WAI (+0.14) the difference is effectively zero. Individual recipes also contradict it. EI-W2-WARM-013 has two Functional ingredients, Turmeric 0.2 g and Chia Seeds 6.3 g, and scores 13. EI-W1-CLEANSE-016 has none (Pineapple 75 g and Carrot 60 g, Lentils 55 g and Egg White 240 g) and scores 16. The 0.49 gap is also small next to the spread of Scores in the sheet, which run from 11 to 20. And the data cannot say which feature the Score responds to: the add-in itself, the extra fat it brings, or the extra ingredient on the list. Functional recipes differ from the others on all three at once, so none of this shows that adding an ingredient raises a Score.
What We Learned
The gap is real as a description of our own recipes. It survives adjustment for body type in seven of eight codes, and produce grams do not account for it. But it is about half a point on a scale where recipes span nine, and we cannot separate the slot from the fat or the ingredient count. The more useful finding is negative. The Score does not move with protein (r = 0.07), and it leans slightly against fiber (r = -0.17); the lowest-scoring recipe in our example has the most fiber. A Score difference under a point should not be read as a difference in quality. For the fat side of the Functional slot, see LAB #051.
What Changed at MOMS
We now list the Functional-slot count next to the Score whenever two recipes are compared in review, and we treat a Score difference under one point as noise. A Functional ingredient has to earn its place with flavor, texture, or a stated reason. We no longer add one to a recipe in order to raise its Score. If you want to see which body type your own menu is built from, the body type quiz at /body-type-quiz starts there.
References
McDonald D, Hyde E, Debelius JW, et al. American Gut: an Open Platform for Citizen Science Microbiome Research. mSystems. 2018;3(3):e00031-18.
Keep reading
Previous experiment
Not sure which shake fits you? We match the formulation to your body type.
Find your body type