In this LAB note
Question · Hypothesis · Method · Results · What Failed · What We Learned · What Changed at MOMS
Question
Two of our shakes sit 4.3 calories apart on the label. One carries 2.5 g of fat and the other carries 22 g. Where did the other 175 calories go?
Hypothesis
Fat is the densest macronutrient we work with: 9 calories per gram, against 4 for protein and carbohydrate. We expected fat grams to be the strongest single predictor of a shake's calorie count, and we expected our highest-fat recipes - the ones built on cashews, avocado, and coconut milk - to sit at the top of our energy range. We expected the reverse to hold at the other end. A shake carrying 2.5 g of fat should be a noticeably lighter shake.
Method
Published research. Sacks and colleagues randomized 811 overweight adults to four reduced-calorie diets in which fat was targeted at either 20% or 40% of energy, with carbohydrate moving inversely from 65% down to 35%. At two years, weight loss was 3.3 kg in both the 20%-fat and the 40%-fat group, and reported satiety, hunger, and satisfaction with the diet were similar across all four arms. Gentilcore and colleagues looked at the same exchange at the level of a single meal: 30 ml of olive oil taken before a mashed-potato meal markedly slowed gastric emptying, delayed the postprandial glucose peak, and attenuated the insulin and GIP responses in six men with type 2 diabetes. Internal analysis. We took all 201 recipes in the MOMS database and converted every macro to calories - 4 per gram of protein and carbohydrate, 9 per gram of fat. We then compared the spread of each macro's calorie contribution against the spread of the shake's own total, computed the correlation between fat grams and carbohydrate grams across the full set, and pulled matched pairs: two recipes sharing a Goal and landing within 3 calories of each other but built on very different amounts of fat.
Results
The totals barely move. Across 201 recipes, calories run from 270.5 to 399.9 - a 129-calorie band with a standard deviation of 27.0 calories. Not one recipe crosses 400. The parts move more than the whole. Fat runs 2.5 g to 22.0 g, a spread of 175.5 calories, standard deviation 34.9. Carbohydrate spreads 206.0 calories, standard deviation 42.5. Each of those two components varies more than the number they add up to. Protein is the flat leg: the standard deviation of its calorie contribution is 9.4. That is only arithmetically possible if fat and carbohydrate move against each other, and they do: r = -0.708 across all 201 recipes. Fat's correlation with total calories is only +0.183. The matched pair. EY-W4-CALM-007-V1 "Soft Landing" - Coconut Milk 240 ml, Banana 14.4 g, Cacao 5 g + Cashew 5 g - is 386.4 kcal with 17.9 g fat, 27.4 g carbohydrate, 30.5 g protein. MI-W4-CALM-007-V1 "Serene Rebuild" - Rice Milk 110 ml, White Rice 33.1 g, nothing in the functional slot - is 389.2 kcal with 3.3 g fat, 56.4 g carbohydrate, 33.4 g protein. Same Goal, same week, 2.8 calories apart, and 14.6 g of fat apart. Those 14.6 g of fat are worth roughly 131 calories; the 29.0 g of carbohydrate running the other way are worth roughly 116. The extremes do the same thing, harder. MI-W3-VITAL-008-V1 "Bright Vitality" carries 2.5 g of fat and 56.3 g of carbohydrate at 387.0 kcal. WAY-W1-HYDRATE-011-V1 "Quenched Core" carries 22.0 g of fat and 19.8 g of carbohydrate at 391.3 kcal. That is 8.8 times the fat, 4.3 calories apart. The fat itself comes from where you would expect. The 156 recipes containing a nut, seed, avocado, or coconut source average 11.2 g of fat; the 45 without average 6.8 g.
What Failed
The trade is not universal. An r of -0.708 leaves about half the variance unexplained, and at least one recipe breaks the pattern outright. WAI-W4-RENEWAL-001-V1 "New Leaf" carries 21.1 g of fat and only 11.0 g of carbohydrate. The fat there is not being paid for in carbohydrate at all; it is paid for in protein, which climbs to 38.8 g, well above our usual band. The arithmetic does not close cleanly either. On the matched pair, the fat and carbohydrate offsets predict that Soft Landing should land a few calories above Serene Rebuild; the label has it 2.8 calories below. Fiber is counted at a reduced energy value and rounding runs through every field, so these reconciliations are approximate. We are describing a strong tendency, not an identity. Fiber does not ride along with fat. The correlation between fat grams and fiber grams is -0.216, essentially flat, and "New Leaf" pairs 21.1 g of fat with 2.9 g of fiber, near the bottom of our fiber range. A higher-fat shake is not automatically a more filling one by our own fiber measure. Our own Score barely registers the trade: the correlation between Score and fat grams is +0.209 across 201 recipes. So nothing here shows that the higher-fat version of a shake is the better one. It is a differently shaped shake at the same energy. That is the whole finding, and we are not going to stretch it further. The Gentilcore result also transfers less cleanly than we would like. The strong gastric-emptying effect came from oil taken 30 minutes before the meal; when the same oil was blended into the meal, the slowing was significantly weaker. A shake is the blended condition, not the preload. The study ran six men, all with type 2 diabetes.
What We Learned
Calories are a ceiling in our system, not a dial. With total energy held under 400 and protein pinned near 30 g, there is essentially one degree of freedom left in a MOMS recipe, and fat and carbohydrate share it. Which means "low fat" does not mean "lighter" here. It means "more carbohydrate." Serene Rebuild has less than a fifth of Soft Landing's fat and it is not a smaller shake. It is a rice shake instead of a cashew shake.
What Changed at MOMS
We stopped reading fat grams as a proxy for how heavy a recipe is, and we changed the order in which we size the slots. Fat is now set first and chosen for mouthfeel and for how slowly we want the shake to release - nuts, seeds, avocado, coconut milk. The carbohydrate slot (rice, banana, dates) is sized last, to bring the recipe back to its energy target rather than to hit a carbohydrate number of its own. We also added a check to the build sheet. Any recipe where fat rises without carbohydrate falling gets flagged for manual review, because it means a third macro is absorbing the swing and we want to see which one before the recipe ships. "New Leaf" is the recipe that taught us to look.
References
Sacks FM, Bray GA, Carey VJ, et al. Comparison of Weight-Loss Diets with Different Compositions of Fat, Protein, and Carbohydrates. New England Journal of Medicine, 360(9):859-873, 2009. | Gentilcore D, Chaikomin R, Jones KL, et al. Effects of Fat on Gastric Emptying of and the Glycemic, Insulin, and Incretin Responses to a Carbohydrate Meal in Type 2 Diabetes. Journal of Clinical Endocrinology & Metabolism, 91(6):2062-2067, 2006.
Keep reading
Previous experiment
Not sure which shake fits you? We match the formulation to your body type.
Find your body type