Milk proteins do work in a frozen dessert that fat usually gets credit for

What non-fat milk solids actually consist of

Strip the fat and water out of milk and roughly a third of what remains is protein, mostly casein with some whey protein, while more than half is lactose and the rest is minerals. Frozen dessert formulations refer to this collectively as milk solids-not-fat, and it is treated as a variable in its own right, adjusted independently of the cream. Skimmed milk powder is the usual way of raising it, which is why it appears in so many careful recipes.

Raising the solids without raising the fat changes several things at once. There is more dissolved and dispersed material competing for the water, so less of that water is free to form large ice crystals. The mixture is more viscous before freezing and denser afterwards. And there is more protein available at the air interface, so the foam introduced during churning is better supported and holds up longer in storage.

Casein is the reason the foam survives churning

Casein does not exist in milk as isolated molecules but as micelles, clusters of protein held together with calcium phosphate. Those clusters adsorb at the surfaces of air bubbles and fat droplets, and they are what keeps a churned mixture from collapsing as soon as the machine stops. Whey proteins contribute alongside them, unfolding at interfaces in the same way egg white proteins do, though in smaller quantity and with different behaviour under heat.

This is why a frozen dessert built on water and fat alone is so much harder to stabilise than a dairy one. Non-dairy formulations have to supply the interfacial protein from somewhere else, whether that is soy, pea or nut protein, or replace its function entirely with emulsifiers and hydrocolloids. Some do this very well now. The point is that they are substituting for a job that milk protein was quietly performing, not simply swapping one liquid for another.

Lactose has a low solubility and a long memory

Lactose is much less soluble in water than sucrose is, and in a frozen dessert it sits in a highly concentrated syrup that is already close to saturation. Given time and temperature fluctuation, it can crystallise. Lactose crystals are hard, gritty and comparatively large, and the resulting defect is described in the trade as sandiness — a texture quite distinct from the coarse iciness caused by ice crystals, though the two are often confused by the eater.

This is the main reason milk solids cannot simply be raised without limit. Formulations usually cap the non-fat solids at a level that keeps lactose below its crystallisation threshold, and products stored badly or held for many months are the ones most likely to develop the fault. It is one of the clearest cases in dessert making where more of a beneficial ingredient produces a specific, nameable failure rather than a gradual decline.

Which is why concentrated dairy behaves differently

Evaporated and condensed milks carry a high solids load in a small volume, and recipes using them tend to produce a denser, chewier frozen texture with noticeably fewer large ice crystals. Sweetened condensed milk brings a very high sugar concentration alongside, which is why the popular no-churn method built on it works at all — between the sugar and the solids, little enough water is free to freeze that the mixture stays soft without any churning.

Cooked or caramelised dairy takes this further. Heating milk solids browns them through the same reactions that darken a crust, generating flavour compounds and altering the proteins. That is the basis of several frozen desserts built on slowly cooked milk, and the change is chemical rather than merely a matter of concentration. The resulting flavour cannot be reproduced by adding caramel to an uncooked base.

Balancing solids against everything else

A formulation is a balance of fat, sugar, non-fat solids, stabiliser and water, and moving any one of them displaces the others. Raising milk solids to fight iciness reduces the water available for sugar to dissolve in, which can push the mixture towards sandiness. Raising fat instead gives richness but at some point coats the palate enough to dull flavour. There is no single correct set of numbers, only a region within which the result behaves.

For a domestic cook the practical lever is usually a few tablespoons of skimmed milk powder added to a base that has been coming out icy. It is cheap, it dissolves readily in warm liquid, and it raises solids without raising fat or sweetness. Adding too much gives a chalky, slightly cooked flavour and a heavy body, which is a good reminder that this ingredient has an optimum rather than a direction.