How to Rank Foods Honestly: A Transparent Scoring Method

Which food is the "best" food? People love a ranked list, and a number like 98 feels like knowledge. But there is no scientifically definitive ordering of individual foods. This page shows how to build a ranking you can defend: one that shows its weights, admits its limits, and is useful for shopping without pretending to be a clinical result.

This page is educational, not medical advice. It describes a way of thinking about food choices for a general adult reader. It does not diagnose or treat anything, and it does not replace advice from a qualified clinician or dietitian.

The problem: foods are not independent products

Nutrition science mostly studies dietary patterns, nutrients and biomarkers, not single ingredients competing head to head. A trial can show that a pattern lowers LDL cholesterol. It usually cannot say how much of the effect came from the oats, the nuts or the beans. Established The 2011 Portfolio trial by Jenkins and colleagues (see Further reading) tested a combination of plant protein, viscous fiber, nuts and plant sterols. That design supports the combination; it does not rank the parts.

So any ordered list of foods is an editorial synthesis: someone's weighted judgment, informed by evidence but not dictated by it. That is fine, as long as the list says so. A list that hides its method invites you to trust a number it has not earned. See Reading Nutrition Evidence: What a Study Can and Cannot Show for the difference between a trial, a biomarker and a mechanism.

A five-criterion rubric

One way to make the judgment visible is to write down the criteria and their approximate weights.

Provisional These weights are judgment, not a validated formula. No experiment produced "35, 25, 15, 15, 10". They state what the ranker cares about: evidence first, then whether the food helps a whole diet add up, then the practical matters that decide whether people actually eat it. If you care more about cost, you might shift ten points from density to affordability. That is legitimate, as long as you say so.

The category-level evidence matters because a single food inherits support from its group. A lentil variety borrows the evidence base for legumes in general. It does not earn a proven individual effect on mortality. See The Daily Dozen and Roles Over Ingredients for how a category checklist, a popular framework rather than an authority, fits this idea.

ANDI: what it captures and what it misses

ANDI stands for Aggregate Nutrient Density Index, popularized by Joel Fuhrman in Eat to Live (2003). In plain terms it asks how many nutrients a food delivers per calorie, and it scales the result so that the top foods score 1,000. Established (as a description of the method). Whether it predicts health outcomes is a separate question, and it has not been established by trials of the kind used to test dietary patterns.

What it captures well: leafy greens and cruciferous vegetables such as kale, collards, mustard greens and broccoli pack many micronutrients into few calories. That is a real and useful observation.

What it cannot see:

• Energy. A food that is low in calories can look wonderful while contributing little fuel to someone who needs it.
• Protein and fats. Lentils, tofu and nuts deliver protein and essential fats that a per-calorie nutrient score underweights.
• Serving size. A food can be dense but eaten by the teaspoon.
• Cost and waste. Watercress scores very high but is pricey, perishable and eaten in small amounts.
• Outcomes. A density number is not a clinical result.

Because of this, ANDI works best as one input among several, which is why it carries 15 points here rather than 100.

Why oats, lentils, tofu and walnuts rank high despite modest ANDI

Oats have a low ANDI figure in published tables, yet they stay near the top of a well-built ranking. The reason is the other criteria. Beta-glucan soluble fiber from oats and barley has an established basis for an authorized cholesterol-related health claim in U.S. regulation (21 CFR 101.81, which refers to 3 grams a day of beta-glucan; that is an amount of the fiber, not of total oats). Oats are also cheap, shelf-stable and easy to cook.

Lentils and tofu earn their places through protein, fiber, minerals and price. Walnuts supply the plant omega-3 ALA and unsaturated fats and fit well in heart-healthy patterns, even though a per-calorie density score sees them as calorie-rich. The lesson is general: when a method's results disagree with one of its inputs, check whether that input measures what you assumed. See Nutrient Coverage on a Plant-Based Diet for the nutrients a food list cannot guarantee on its own.

Ties and score inflation

Two habits make rankings misleading.

Treat 1 to 3 points as a tie. Your weights are rough, so small differences are noise. When foods tie, let ease of regular use and variety decide, and do not describe the order as a demonstrated difference.

Remember that scores are not percentages. A 90 is not 10 percent healthier than an 80. The scale is ordinal: it says "higher priority", not "this much more healthy". Treating it as a ratio scale is a common arithmetic mistake. See Mathematics for Better Decisions on scales and what operations they permit.

Counting honestly

A list of 300 foods does not contain 300 independent interventions. Many entries are separate shopping forms: beet root and beet greens, whole chickpeas and chickpea flour, fresh and dried fruit, grain and flour. They are alternatives in one role, not 300 separate targets for variety. Counting them as independent would inflate both the apparent diversity of a diet and the evidence behind the list.

The same honesty applies to authorities. If a list is inspired by many named experts, their work answers different questions, so counting their names as votes misleads. Analogy, not fact: a ranking built this way resembles a committee report whose members never met. The signatures do not add up to a consensus.

A sample, with reasons

Here is how a top tier and a low tier might look, and why. These illustrate the reasoning; they are not scientific orderings.

A lower rank does not mean a food is bad. It usually means the food costs more attention or money per unit of benefit, or fills a role the list already covers. For buying and storing the top tier cheaply, see The Practical Kitchen: Staples, Meal Prep and Food Safety.

Using a ranking for decisions

A good ranking is a stable shopping reference, not a command. Revise it when your budget, access, tolerances, training or nutritional needs change, not each time you read a new antioxidant claim. Headlines about a single compound usually describe cell or animal work, small trials or biomarkers, and Ioannidis (2018) argues that much nutritional epidemiology deserves cautious reading. Keep the method and change the inputs.

The arithmetic is simple enough to audit: multiply each criterion score by its weight and add. Writing the assumptions beside the number is the habit explained in Quantitative Reasoning with Stated Assumptions: a result is only as good as the assumptions you state.

Try this

1. Score ten staples. List ten foods you actually eat. Score each from 0 to 10 on the five criteria, multiply by the weights above, and total. Then change one weight and note whether the order shifts. Foods whose rank moves a lot are fragile choices.
2. Find the ties. Mark any pair within three points as a tie. Choose between tied foods on price and taste, and write down that this was the reason.
3. Audit one claim. Take a headline about a "superfood". Ask which of the five criteria it actually speaks to, and whether the evidence was a human trial, a biomarker or a lab result.

Further reading

• Fuhrman, J. Eat to Live (Little, Brown, 2003), for the original ANDI idea.
• Jenkins, D.J.A., et al. "Effect of a dietary portfolio of cholesterol-lowering foods given at 2 levels of intensity of dietary advice on serum lipids in hyperlipidemia: a randomized controlled trial." JAMA 306(8), 2011.
• U.S. Department of Agriculture and U.S. Department of Health and Human Services, Dietary Guidelines for Americans, 2025-2030.
• U.S. Food and Drug Administration, 21 CFR 101.81, soluble fiber and coronary heart disease health claim.
• Ioannidis, J.P.A. "The challenge of reforming nutritional epidemiologic research." JAMA 320(10), 2018.