In this LAB note
Question · Hypothesis · Method · Results · What Failed · What We Learned · What Changed at MOMS
Question
Eighty-nine of our 201 formulations carry a tree-nut declaration, and that label comes from one of two very different things: a base of almond milk, or a small handful of cashew, walnut, or pecan. When someone asks for a nut-free version, which of those two is actually load-bearing?
Hypothesis
We expected the almond milk to be the hard one. In any recipe that uses it, it is the single largest ingredient by mass — an average of 248 ml against a total blend weight of roughly 590 g — so removing it looked like a rebuild. The 5–20 g of whole nut sitting in the Functional column looked like a garnish: a texture-and-aroma note we could delete with one line in the recipe file and move on. We expected removing almond milk to cost a rebuild, and removing the nuts to cost nothing.
Method
We separated all 89 tree-nut-labeled recipes by where the nut actually enters the formula. Fifty-one are labeled because of whole nuts only, 22 because of almond milk only, and 6 because of both. For each of the 60 recipes containing a whole nut we computed the nut’s mass, its lipid contribution using USDA FoodData Central fat fractions (cashew 43.8%, walnut 65.2%, pecan 72.0%, almond 49.9%), and that contribution as a share of the recipe’s declared total fat. We then compared the 89 nut-labeled recipes against the 112 nut-free ones on Score, kcal, protein, fiber, and fat, and catalogued which fat-bearing ingredients the nut-free recipes use instead. Two published findings framed the interpretation. Baer and Novotny (Nutrients, 2018) measured metabolizable energy from cashews in a controlled human feeding study and found it lower than Atwater factors predict, because the intact nut cell wall limits lipid release — so a nut’s fat is real on the label but only partly absorbed. Brown and colleagues (American Journal of Clinical Nutrition, 2004) fed vegetable salads with fat-free, reduced-fat, and full-fat dressings and found carotenoid appearance in plasma chylomicrons was negligible in the fat-free condition and rose with dressing fat — so the fat sitting alongside carrot, beet, and spinach in a shake is not inert calories, it is the transport for the pigments we put there on purpose.
Results
The hypothesis was backwards. The almond milk is the easy removal. Its 248 ml carries roughly 1 g of fat, and the recipe database already contains five interchangeable bases — soy, oat, coconut, rice, and coconut water — that hit the same volume within the same calorie envelope. Swapping almond for oat is a base swap, not a reformulation. The handful is the hard one. Median whole-nut mass across the 51 whole-nut-only recipes is 8.5 g — about five cashews — and the range is 5 g to 20 g. That is under 2% of the blend by weight. But across the 60 recipes containing a whole nut, the nut supplies a median 39% of the recipe’s total fat, with a range of 12% to 69%. The specific cases are stark. Quiet Recovery (EY-W4-CALM-001-V1, Score 20, 386.3 kcal, 33.3 g protein) uses Cashew 15 g, contributing 6.6 g of fat against a declared total of 10.0 g — 66%. Slow Clean (EI-W1-CLEANSE-017-V1, Score 16, 372 kcal, 32.1 g protein, 4.5 g fiber) uses Cashew 18 g, contributing 7.9 g against 11.5 g total — 69%. Kinetic Bloom (WI-W1-VITAL-010-V1, Score 17, 396.2 kcal) uses Cashew 15 g plus Pecan 5 g, contributing 10.2 g against 14.9 g total — 68%. The population comparison confirms the same shape. Nut-labeled and nut-free recipes are nearly identical on everything except fat: Score 15.1 vs 14.9, protein 32.8 g vs 33.2 g, fiber 7.3 g vs 7.8 g, kcal 378.7 vs 372.0 — but fat 11.2 g vs 9.4 g. The nut is not adding protein and it is not adding fiber. It is the fat system, delivered in a package small enough to look decorative. The 112 nut-free recipes get their fat elsewhere, and the pattern is narrow: coconut milk (42 recipes, avg 231 ml), avocado (26 recipes, avg 27.3 g), chia seeds (15 recipes, avg 8.3 g), flaxseed (7 recipes, avg 3.6 g). Where a nut delivers roughly 4 g of fat in 8.5 g of material, avocado needs about 27 g to do the same job.
What Failed
Our first attempt at a fix was a blanket rule: replace any whole nut with an equal mass of chia seeds. It failed in both directions against the actual data. It undershoots at the top. Chia is about 31% fat against cashew’s 43.8%, so swapping 15 g for 15 g in Quiet Recovery delivers 4.7 g of fat instead of 6.6 g — a 1.9 g shortfall on a recipe that only had 10.0 g to begin with. Worse, that same 15 g of chia adds roughly 5 g of fiber, which would push Slow Clean from 4.5 g of fiber to about 9.5 g and change what the drink is. It overshoots at the bottom, because the 39% median hides a real spread. Soft Landing (EY-W4-CALM-007-V1) uses Cashew 5 g against a total fat of 17.9 g — the nut is 12% of its fat, and here the garnish framing was correct all along. Sunrise Surge (EY-W1-ENERGY-001-V1, Score 16, 393.5 kcal, 34.5 g protein, 9.4 g fiber) uses Cashew 5 g against 11.7 g of fat, or 19%, because its 195 ml of coconut milk is already doing the fat work. In both, a chia substitution adds unwanted fiber to solve a problem that does not exist. There is no single nut-replacement rule. The same 5 g of cashew is decorative in one recipe and structural in another, and the only way to tell is to look at what the rest of the formula is already carrying.
What We Learned
Ingredient mass is a bad proxy for ingredient function. The two ingredients that trigger the same three-word allergen declaration differ by a factor of roughly thirty in weight and invert completely in difficulty: the 248 ml one is interchangeable, the 8.5 g one is load-bearing. The useful number is not how much of an ingredient is in the shake. It is what share of a macro that ingredient is responsible for. Sorted by mass, the nut looks like the last thing you would worry about. Sorted by fat share, it is the first.
What Changed at MOMS
Every nut entry in the recipe database now carries a computed fat-share figure alongside its gram weight, and nut removals are routed by that number rather than by the ingredient name. Below roughly 30% fat share the nut is a straight deletion with no rebuild. At or above it, the recipe goes to a fat rebuild using the substitution ladder the nut-free formulas already established — avocado at roughly 3 g of material per 1 g of fat, chia where a small fiber increase is acceptable, or a base shift to coconut milk where the flavor allows. Almond milk is handled separately as a base swap with no macro rebuild at all. We also stopped writing “contains tree nuts” as a single fact about a recipe. The recipe file now records which nut, in which column, at what fat share — because those are three different formulation problems wearing one label.
References
Baer DJ, Novotny JA. Metabolizable Energy from Cashew Nuts is Less than that Predicted by Atwater Factors. Nutrients, 11(1):33, 2018. | Brown MJ, Ferruzzi MG, Nguyen ML, et al. Carotenoid bioavailability is higher from salads ingested with full-fat than with fat-reduced salad dressings as measured with electrochemical detection. American Journal of Clinical Nutrition, 80(2):396-403, 2004.
Keep reading
Previous experiment
Next experiment
Not sure which shake fits you? We match the formulation to your body type.
Find your body type