In this LAB note
Question · Hypothesis · Method · Results · What Failed · What We Learned · What Changed at MOMS
Question
If two of our shakes report the same calories, protein, carbohydrate, fat and fiber, how much of the same food is actually in them?
Hypothesis
We expected macro-matched recipes to converge on ingredients too. A 16 oz shake has six slots and a narrow target: roughly 350 to 400 calories carrying at least 30 grams of protein. We assumed the arithmetic would force similar solutions, because there are only so many foods that deliver protein without spending the calorie budget. Our prediction was that any two recipes agreeing on all five label numbers would share at least half their ingredients.
Method
Two parts: published research on whether a fixed nutrition panel fixes the outcome, and an internal search of our own recipe database. Hall and colleagues (2019) ran a 20-person inpatient randomized controlled trial at the NIH in which an ultra-processed and an unprocessed diet were matched for presented calories, sugar, fat, sodium, fiber and macronutrients. On paper the two arms were the same diet. Participants ate 508 plus or minus 106 kcal per day more on the ultra-processed arm, and gained 0.9 plus or minus 0.3 kg, against losing 0.9 plus or minus 0.3 kg on the unprocessed arm. Novotny, Gebauer and Baer (2012) attacked the same question from the other end, measuring the energy humans actually extract from almonds inside a mixed diet. They found 4.6 kcal per gram against the 6.0 to 6.1 kcal per gram the Atwater factors predict: a 32 percent overestimate, 129 kcal per 28 gram serving rather than 168 to 170. Internally, we took all 201 recipes in the MOMS database and searched every one of the 20,100 possible pairs for macro twins: pairs agreeing within 3 kcal, 0.6 g protein, 1.2 g carbohydrate and 0.6 g fat. For each twin pair we then parsed the six ingredient slots and measured how much actual food the two recipes had in common, counting a shared ingredient as the same food name appearing in both regardless of slot or grams.
Results
Twenty-one of the 20,100 pairs cleared the macro-twin threshold. Those 21 pairs share an average of 3.1 ingredients, and their median overlap - shared foods divided by combined foods - is 0.17. Four pairs share exactly one ingredient. The tightest pair. EY-W1-ENERGY-002-V1, Morning Ignite: Coconut Milk 195 ml plus Soy Milk 150 ml; Cashew 13.5 g plus Banana 40 g plus Dates 10 g; Chickpeas 30 g plus Egg White 195 g; Agave 5 g. It lands at 386.7 kcal, 31.9 g protein, 36.1 g carbohydrate, 13.3 g fat, 5.0 g fiber. WI-W4-RENEWAL-008-V1, Dawnlight Renew: Oat Milk 150 ml; Pumpkin 75.6 g; Apple 45 g plus Dates 8.4 g plus Mango 36 g; Chicken 60 g plus Greek Yogurt 96 g; Pecan 11.7 g. It lands at 384.4 kcal, 32.1 g protein, 37.0 g carbohydrate, 13.4 g fat, 4.9 g fiber. Every number on the label agrees: 2.3 calories apart, 0.2 g of protein, 0.9 g of carbohydrate, 0.1 g of fat, 0.1 g of fiber. Both are 16 oz. Both carry the Cal Label Standard. Between them they use 15 distinct ingredients and share exactly one, dates, at 10 g and 8.4 g. The food underneath is not close. One draws its protein from 225 g of chickpeas and egg white, the other from 156 g of chicken and Greek yogurt. One is 639 g of total food and the other 483 g - a 156 g gap, which happens to be the exact weight of the entire protein slot in the lighter of the two.
What Failed
We wanted our internal Score to be the instrument that catches what the label misses. On this pair it does separate them: Morning Ignite scores 16, Dawnlight Renew 13. That looked like the answer, so we tested it properly across all 201 recipes - and it does not hold up. The Score's correlation with calories is 0.02. With protein, 0.07. With fat, 0.21. With carbohydrate, -0.18. With fiber, -0.17. With ingredient count, 0.08. With total food weight, -0.25. Nothing clears 0.25 in either direction. Across the 21 twin pairs the Score differs by an average of 2.4 points on an 11 to 20 scale, with no pattern we can trace to anything we measure. So we are not claiming the Score sees the food that the label cannot. We do not know what the 3-point gap on this pair is tracking. The finding that two recipes can agree on every label number while sharing one ingredient in fifteen stands on its own; the Score's disagreement is, for now, unexplained, and calling it validation would be the easy and dishonest version of this post.
What We Learned
Macros are a projection, not a fingerprint. Five numbers cannot reconstruct 639 grams of food drawn from a 74-ingredient pantry; the mapping discards too much. In our own database, a matching nutrition panel predicts roughly one ingredient in six. That would be a curiosity if the panel were the thing that mattered. Both published results above describe outcomes moving while the panel stays fixed - absorbed energy in the almond work, spontaneous intake and body weight in the Hall trial. If you can hold the label constant and still move the result, then formulating to the label alone is optimizing a variable that is not the one doing the work. The practical version for anyone reading a nutrition panel: two products with the same numbers are not the same food, and the panel is not designed to tell you that.
What Changed at MOMS
We stopped treating macro equivalence as recipe equivalence in substitution work. Previously, when an allergen filter removed a recipe from someone's rotation, the replacement was chosen by finding the closest macro match within the same Cal Label. That rule is what produced the pair above: on macros alone, Dawnlight Renew is a legal substitute for Morning Ignite. In practice it would hand a customer chicken, pumpkin and pecans in place of chickpeas, egg white and cashews, and the swap would be invisible on the label. A substitute now has to clear an ingredient-overlap floor as well as the macro window: at least 40 percent shared foods with the recipe it replaces. Anything below that is presented as a new recipe with its own ingredient list rather than slipped in as a swap. Eighteen of our 21 macro-twin pairs fail the new floor, which is the point.
References
Hall KD, Ayuketah A, Brychta R, Cai H, Cassimatis T, Chen KY, et al. Ultra-Processed Diets Cause Excess Calorie Intake and Weight Gain: An Inpatient Randomized Controlled Trial of Ad Libitum Food Intake. Cell Metabolism, 30(1):67-77.e3, 2019. | Novotny JA, Gebauer SK, Baer DJ. Discrepancy between the Atwater factor predicted and empirically measured energy values of almonds in human diets. The American Journal of Clinical Nutrition, 96(2):296-301, 2012.
Keep reading
Previous experiment
Not sure which shake fits you? We match the formulation to your body type.
Find your body type