Nutrient density scores promise a single number expressing how much nutrition a food provides for its energy. The scores exist in several incompatible versions, and the disagreements come from choices made in their design.
Every score begins with a selection
A scoring system must decide which nutrients count, and the usual approach includes a set of beneficial ones alongside a set to be limited, such as saturated fat, added sugar and sodium.
Adding or removing a single nutrient from either list reorders the results, because foods that are strong in one area are frequently weak in another.
There is no scientifically neutral selection available, since the choice reflects which deficiencies and excesses are considered important in a given population at a given time.
The denominator changes the answer
Nutrients can be expressed per hundred grams, per unit of energy, or per typical serving, and the three produce markedly different rankings.
Per hundred grams favours foods with high water content, since water dilutes energy but not necessarily nutrients, while per calorie favours foods that are low in energy regardless of portion.
Per serving depends on how a serving is defined, which is set by convention and sometimes by manufacturers, importing an additional judgement into what looks like arithmetic.
Capping and weighting hide decisions
Most systems cap the contribution of any single nutrient so that a food fortified heavily with one vitamin cannot dominate the score.
Where the cap is set determines whether fortified products score close to whole foods, which is a consequential choice presented as a technical detail.
Weighting nutrients unequally has the same character: it is defensible, it is necessary, and it is a value judgement embedded in a number.
Bioavailability is largely absent
Scores are calculated from nutrient content as analysed, not from the amount the body can absorb and use.
Iron from plant sources is absorbed far less readily than iron from meat, and some minerals are bound by compounds in the same food that reduce their availability.
Because absorption depends on the whole meal and on the individual, it resists inclusion in a per-food score, and its absence flatters some foods and penalises others.
What the scores are useful for
Within a single category the comparisons are usually sound: one breakfast cereal against another is a fair use of the tool.
Across categories the results become strange, because a food is being compared with something that occupies a different role in a meal entirely.
Diets are assembled from combinations, and no per-food score can capture how foods complement each other, which is why dietary pattern rather than food ranking is what nutrition guidance is built on.