How the Density Score works.

Read this first: this scale starts at healthy

The Density Score compares produce with produce. Every item on this scale is a whole food from the produce department: fresh fruit and vegetables, and from version 1.2.3 the nuts and dried fruit that carry produce codes. A 30 here is a fresh fruit or vegetable with water, fiber, and vitamins in it. It is a stronger everyday choice than nearly anything sold in a wrapper.

Apps and labels that score packaged food rate a different universe. Their low scores flag additives, added sugar, and heavy processing. Produce has none of those, which is exactly why those systems give almost all fresh produce the same top grade and stop being useful in the produce aisle. Our score picks up where theirs run out of resolution: it ranks the aisle itself, so you can see that a bowl of raspberries carries more nutrition than a bowl of iceberg.

So a PLUed 40 and a packaged-food 40 are different animals. Ours means "a solid everyday fruit or vegetable, with denser picks available on the same shelf." It never means "avoid this." There is no red on this scale.

The score: four questions, one number

The Density Score is a weighted blend of four measurements, each answering a question a shopper actually has.

• Nutrient concentration, 50% of the score. How packed the food itself is with nutrients, once you set the water aside. We total fiber and seventeen vitamins and minerals (potassium, calcium, magnesium, iron, zinc, copper, selenium, vitamins A, C, E, K, B1, B2, B3, B5, B6, and folate) as a share of their FDA Daily Values, per gram of the food without its water. Sugar then reduces the result in proportion to how much of that it is. Grapes are mostly sugar once the water is gone, and the math says so.
• Nutrition per serving, 30% of the score. How much nutrition one typical serving actually gives you: the same Daily Value totals, measured for the FDA reference serving of that food (for example 85 g of leafy greens, 140 g of fruit, 30 g of nuts, and 40 g of dried fruit). This axis is why raspberries outrank iceberg: a serving of berries is food, a serving of iceberg is mostly crunch. Inside the score, one nutrient can count for at most 100% of its Daily Value, so a single standout nutrient cannot carry an otherwise ordinary food; the nutrition facts in the app show the true uncapped values (a serving of lychees really is 111% of a day's vitamin C). From version 1.2.3, every scored item carries all eighteen nutrients, per 100 g and per serving, and any value that is not a USDA laboratory measurement names its source (see Where the numbers come from).
• Low calories, 10% of the score. Fewer calories per bite means a higher score, water included. A small tiebreaker that honors "eat freely" foods without letting water alone win. Foods that get most of their calories from fat get one adjustment, explained next.
• Plant compounds, 10% of the score. The natural compounds beyond vitamins: the pigments in berries, tomatoes, and carrots, the broccoli-family compounds, the garlic-family compounds. We measure six families from USDA content databases (flavonoids, proanthocyanidins, carotenoids, glucosinolates, isoflavones, and sulfur compounds) and credit each food for its strongest one. No Daily Value exists for these compounds, so this axis is scaled against the produce aisle itself and capped at a tenth of the score.

Water is deliberately counted on both sides: removed from Nutrient concentration so filler cannot flatter a food's makeup, and included in Low calories because dilution is precisely what makes a food light. Tap any score in the app to see that item's own four sub-scores, its water share, its sugar share, and the full nutrient panel behind them. In the app the four bars are labeled Nutrients, Per serving, Low calories, and Plant power.

Healthy fats count fairly

Calories alone would mark down every nut and every avocado, because fat carries 9 calories per gram. In nuts and avocados most of that fat is unsaturated, the kind U.S. dietary advice asks people to choose. So from version 1.2.3 the score treats it fairly:

• When it applies. Only when fat gives at least half of a food's calories (9 times its grams of fat, divided by its calories, is 0.5 or more). That threshold keeps the rule off every ordinary fruit and vegetable. Today it applies to 23 items: the avocados, the two coconuts, and the fifteen nuts and seeds. Chestnuts (about a tenth of their calories from fat) and dried fruit are scored without it.
• What moves. The calories from unsaturated fat (total fat minus saturated fat) do not count against Low calories, and that unsaturated fat is set aside with the water when we measure Nutrient concentration, with the sugar share taken on what remains. Saturated fat still counts in full, and a food whose saturated fat USDA never measured gets no credit. The two coconuts barely move, because their fat is mostly saturated.
• What it does. A Hass avocado scores 70 with the rule and 54 without it. The item's nutrition facts say "Healthy fats are not counted against this score." wherever the rule applies.

Why: the FDA's 2024 update to the "healthy" claim replaced the 1994 definition's limit on total fat with a limit on saturated fat, so avocados, nuts, and seeds now qualify for it (final rule, Federal Register, December 27, 2024; FDA summary). The Dietary Guidelines for Americans, 2020-2025 ask people to limit saturated fat "by replacing them with unsaturated fats" and list nuts and seeds among the protein foods of a healthy pattern. And the PREDIMED trial found fewer major cardiovascular events with a Mediterranean diet supplemented with mixed nuts or extra-virgin olive oil than with a low-fat control diet (Estruch R, et al. N Engl J Med 2018;378:e34; this is the authors' 2018 reanalysis, which replaced their withdrawn 2013 report).

The data: USDA, item by item

Nutrient values come from the USDA's laboratory food-composition database (SR Legacy, public domain), through USDA FoodData Central. Each of PLUed's 891 scored kinds of produce is matched to its USDA food. The 876 kinds in the catalog before version 1.2.3 were verified four separate ways: an independent recomputation of every score, a row-by-row review of every mapping by five independent reviewers, external checks of every stand-in against published nutrition data, and a comparison against the FDA's own retail produce nutrition program values for the 40 most common fruits and vegetables. The 23 kinds added in version 1.2.3 (nuts, peanuts, sunflower seeds, Medjool and dried dates, prunes, fresh prunes, raisins, and dried apple slices) are each matched to the USDA entry for the form they are sold in, for example raw almonds and dry-roasted pistachios. The one exception is white walnuts, which are scored as English walnuts and labeled that way.

Some specialty items have no USDA entry of their own: 71 of the 891 scored items. Those are scored as their closest USDA relative and labeled that way in the app ("scored as: Grapes"), so nothing is presented as more precise than it is. Eight items carry no score at all, and the app says why on each one instead of showing a zero: decorations such as ornamental gourds and decorative corn, banana leaves (a cooking wrapper), and three with no reliable nutrition data (aloe vera leaves, sugar cane, and vanilla beans).

Serving sizes use the FDA's reference amounts customarily consumed (RACC), the same standard behind Nutrition Facts labels, rather than whatever cup measure happens to appear in a database.

Where the numbers come from

USDA's laboratory database does not measure every nutrient for every food. Vitamin K and vitamin E are by far the most common holes, then selenium, vitamins C and A, and pantothenic acid (B5). From version 1.2.3, no item shows "no data": every missing value is filled, and every filled value names its source on the item's nutrition facts. We take the first of these that exists:

1. USDA's laboratory value for the same food, in the form it is sold.
2. Another government's laboratory value for the same food.
3. USDA's value for the same food in another form (for example roasted chestnuts for raw ones).
4. A published laboratory study of the same food.
5. A laboratory value for a close variety or a closely related part of the plant.
6. Another food database's own estimate.
7. Last, an estimate from the closest related food USDA did measure. The app says "Estimated from" and names that food.

When several sources have a value at the same step, we use their median. A lab reading reported only as below detection or a trace is used only when no measured value exists for that nutrient. Every value is converted to the units of a U.S. label, and vitamin A is used only as retinol activity equivalents, as on a U.S. label. For five vitamin E values, the source table offers only a slightly broader measure (alpha-tocopherol equivalents); we use it as the closest measure available and check those first each month. A value more than one and a half times the highest USDA value for that nutrient among the catalog's fruits and vegetables is held back and the next step is used, because a value that far out needs a second look before it moves a score. Today 412 of the 16,038 nutrient values on scored items (891 items, eighteen nutrients each) come from these steps; the rest are USDA laboratory values (for an item scored as a stand-in, the stand-in's). We check the filled values again every month for better research.

On the item's nutrition facts, a filled value reads "Lab value:" with its source, "USDA," with the other form, "Lab value for a close variety:", "Estimate:", or "Estimated from" with the related food. For an item scored as a stand-in, a filled value is the stand-in's and reads "Estimated from" the stand-in, naming where that value came from, except where a study measured the item's own food (the fresh herbs' vitamin K), which shows as a lab value.

The bands

• 75 to 99: Powerhouse. The leafy greens and their neighbors. Watercress, spinach, kale, chard, and also broccoli, carrots, and bell peppers.
• 50 to 74: Strong. Berries, citrus, tomatoes, potatoes, sweet potatoes, avocados, and the nuts and seeds, from hazelnuts at 74 down to roasted peanuts at 53.
• 25 to 49: Everyday. Apples, bananas, grapes, pears, chestnuts, and prunes. Solid daily food.
• Under 25: Treat. Four items land here, all dried fruit: dried dates, golden raisins, black raisins, and dried pineapple, the fruit-stand candy. Fresh dates and Medjool dates sit right on the line at 25. Still fruit. Still fine.

Notice the floor: the lowest band is called a treat, because that is what it is. Nothing in the produce aisle earns a warning label from us.

The score in the app

The score is free in every version of PLUed, and it shows up in four places.

• On every item. An item's page shows its score and band. Tap the score for the four bars, the eighteen nutrients per 100 g and per serving with the source of any value that is not a USDA laboratory measurement, and the plant compounds. Scan a PLU sticker and you land on that page.
• Rankings. The Nutrition explorer ranks everything, fruit only, vegetables only, one store section, one recipe set, or your own store's list. Codes that share one USDA nutrition row are shown as one line, so every size of the same apple does not fill the screen. Items without a score are counted and left out. They are never ranked as a zero.
• Compare. Two to four items side by side: each score, and each nutrient as a share of its Daily Value in one serving.
• Build a plate. A plate's score is the average of its items' scores, weighted by how many grams of each are on the plate, so bigger portions count more. The app shows the plain average beside it, so the weighting is visible. The rings add up each nutrient across the plate. Calories are shown for the plate and never enter the score. Produce that is usually eaten cooked is marked as such.

The evidence: tested against the published research

Before shipping, we correlated the Density Score against the established academic measures of nutrient density, computed from the same USDA data:

• 0.88 rank correlation with the CDC-published "Powerhouse Fruits and Vegetables" nutrient-density ranking (Di Noia, Preventing Chronic Disease, 2014), the only peer-reviewed ranking built specifically for produce.
• 0.88 rank correlation with the Nutrient Rich Foods index (NRF9.3, Drewnowski), the validated measure behind most front-of-pack nutrition scoring, computed exactly across our full catalog of 891 scored items. Without the fair-fat rule it would be 0.91. The difference is by design: the NRF counts every calorie, so it marks nuts down for their fat, and we choose not to count unsaturated fat against them.
• 0.98 rank correlation when our eighteen nutrients, plus protein, are run through the CDC study's own per-calorie method. The two panels share fourteen nutrients: the CDC's seventeen add protein, vitamin D and B12, the last two essentially absent from fresh produce; our eighteen add magnesium, copper, selenium and B5.

The Plant compounds axis follows the same path the research is taking: the NRF's own author added flavonoids to fruit scoring in 2020 (NRF9f.3), and the USDA published its first glucosinolate database in 2026. Our score simply carries that direction across all six measurable families.

Where we knowingly differ from those indices, it is the serving-size question. Per-calorie scoring quietly assumes you eat 100 calories of everything, which flunks raspberries and blueberries as "not dense" while a real serving of strawberries delivers most of a day's vitamin C for about 45 calories. Our Nutrition per serving axis measures the serving you actually eat. That is the main place our ranking and the per-calorie literature part ways, and we consider it a correction rather than a disagreement.

References: Di Noia J., Defining Powerhouse Fruits and Vegetables, Prev Chronic Dis 2014;11:130390 · Drewnowski A., The Nutrient Rich Foods index, Am J Clin Nutr 2010 · USDA FoodData Central · FDA nutrition information for raw fruits and vegetables

Credits

The Density Score is built on public food-composition data and published research. We thank everyone who measured these foods. None of the organizations below produced, reviewed, or endorses PLUed. Where we used their values, we converted them to the units of a U.S. Nutrition Facts label and, where several tables cover a food, took the median.

Food-composition databases

• United States. U.S. Department of Agriculture, Agricultural Research Service. FoodData Central (SR Legacy, including its carotenoid values), and the USDA databases for the flavonoid, proanthocyanidin, isoflavone, and glucosinolate content of foods. Public domain.
• Japan. Ministry of Education, Culture, Sports, Science and Technology (MEXT). Standard Tables of Food Composition in Japan 2020 (Eighth Revised Edition), 2023 supplement (日本食品標準成分表（八訂）増補2023年). mext.go.jp
• Denmark. The Danish Food Composition Database, Frida version 5.5, National Food Institute, Technical University of Denmark (DTU). doi:10.11583/DTU.29500682, licensed under CC BY 4.0.
• United Kingdom. Public Health England (2021). McCance and Widdowson's The Composition of Foods Integrated Dataset 2021 (CoFID). Contains public sector information licensed under the Open Government Licence v3.0.
• Sweden. Livsmedelsverket (Swedish Food Agency), Livsmedelsdatabasen, licensed under CC BY 4.0.
• France. Anses. 2025. Table de composition nutritionnelle des aliments Ciqual. doi:10.57745/RDMHWY, licensed under the Licence Ouverte / Etalab 2.0.
• Germany. Max Rubner-Institut (2025): Bundeslebensmittelschlüssel (BLS), Version 4.0. blsdb.de, licensed under CC BY 4.0.
• Finland. Finnish Institute for Health and Welfare, Fineli, Release 20. fineli.fi, licensed under CC BY 4.0.

Published studies

• Presse N, Potvin S, Bertrand B, Calvo MS, Ferland G. Phylloquinone content of herbs, spices and seasonings. J Food Compos Anal 2015;41:15-20. doi:10.1016/j.jfca.2014.12.020 (vitamin K in fresh herbs)
• Dismore ML, Haytowitz DB, Gebhardt SE, Peterson JW, Booth SL. Vitamin K content of nuts and fruits in the US diet. J Am Diet Assoc 2003;103:1650-1652. doi:10.1016/j.jada.2003.09.028 (vitamin K in macadamia nuts)
• Kulczynski B, Gramza-Michalowska A. The profile of secondary metabolites and other bioactive compounds in Cucurbita pepo L. and Cucurbita moschata pumpkin cultivars. Molecules 2019;24(16):2945. doi:10.3390/molecules24162945
• Pinela J, Barros L, Carvalho AM, Ferreira ICFR. Nutritional composition and antioxidant activity of four tomato (Lycopersicon esculentum L.) farmer' varieties in Northeastern Portugal homegardens. Food Chem Toxicol 2012;50:829-834. doi:10.1016/j.fct.2011.11.045
• Vela-Hinojosa C, Escalona-Buendia HB, Mendoza-Espinoza JA, et al. Antioxidant balance and regulation in tomato genotypes of different color. J Amer Soc Hort Sci 2019;144(1):45-54. doi:10.21273/JASHS04525-18
• Tesoriere L, Butera D, Pintaudi AM, Allegra M, Livrea MA. Supplementation with cactus pear (Opuntia ficus-indica) fruit decreases oxidative stress in healthy humans. Am J Clin Nutr 2004;80:391-395. doi:10.1093/ajcn/80.2.391
• Gomez-Coronado DJM, Ibanez E, Ruperez FJ, Barbas C. Tocopherol measurement in edible products of vegetable origin. J Chromatogr A 2004;1054:227-233. doi:10.1016/j.chroma.2004.08.072
• Anibarro-Ortega M, Dias MI, Petrovic J, Mandim F, Nunez S, Sokovic M, Lopez V, Barros L, Pinela J. Nutrients, phytochemicals, and in vitro biological activities of goldenberry (Physalis peruviana L.) fruit and calyx. Plants 2025;14(3):327. doi:10.3390/plants14030327
• Pereira C, Barros L, Carvalho AM, Ferreira ICFR. Nutritional composition and bioactive properties of commonly consumed wild greens: potential sources for new trends in modern diets. Food Res Int 2011;44:2634-2640. doi:10.1016/j.foodres.2011.05.012
• Heleno SA, Stojkovic D, Barros L, et al. A comparative study of chemical composition, antioxidant and antimicrobial properties of Morchella esculenta (L.) Pers. from Portugal and Serbia. Food Res Int 2013;51:236-243. doi:10.1016/j.foodres.2012.12.020

Honest limits

• Bioactive compounds have no Daily Values. The Plant compounds axis uses measured content from USDA databases, never the withdrawn "antioxidant capacity" (ORAC) numbers, and is scaled against the produce aisle itself because no regulatory reference exists. Garlic-family sulfur compounds come from published laboratory studies rather than a USDA database, since those compounds only form when the food is cut or crushed. Fourteen items have no plant-compound measurement in any of these sources; they get the middle value of the produce catalog on this axis, neither a credit nor a penalty. This is why the axis is capped at a tenth of the score.
• Filled values are the best available, not the last word. A value from another country's table describes that country's samples, and an estimate from a related food is an estimate. That is why each one is labeled on the item and checked again every month.
• The score is informational, never medical. It ranks nutrient density. It does not know your diet, your needs, or your doctor's advice, and it makes no health claims.
• USDA values are averages. A specific melon's nutrition varies with variety, soil, season, and ripeness. The score describes the food, and no single piece of fruit signed a contract.
• Weights are choices. The 50/30/10/10 blend, the sugar penalty, and the fair-fat threshold are editorial decisions, published here so you can judge them. The correlations above show the outcome stays firmly anchored to the peer-reviewed consensus.

Questions about the methodology? Email [email protected].