Database analysis is usually enough for a product made from ordinary ingredients, where cooking mostly removes water you can weigh and the label makes no nutrient claims. Lab analysis is the better choice when processing changes the food in ways a recipe can't predict. That includes frying and fermentation, ingredients no database covers, nutrient claims that sit near a threshold, and buyers who ask for lab results.
FDA allows either method. Its Food Labeling Guide says the agency has not stated how a company should determine nutrient content for labeling, and that whatever the source, the company is responsible for the accuracy of the label.
Find your product:
- Baked goods, granola, nut butters, dressings, spice blends, simple sauces and jams: a database label usually holds up.
- Anything fried: the recipe can't tell you how much oil stayed in.
- Kombucha, long-fermented sourdough, yogurt, kimchi, fermented hot sauce: sugar changes during processing.
- Long cooking, draining or reducing, or declared vitamins: nutrients are lost, not just water.
- An ingredient that isn't in USDA's database: you need another source for it.
- A claim such as "good source of fiber" or "low fat": test if the number is close to the line.
- A buyer who asks for lab data: their spec sheet decides.
- Budgeting for a lab: a full label panel is listed at $1,656 per sample in one lab's 2026 catalog.
What does each method actually measure?
A database label is arithmetic. You weigh each ingredient, look up its nutrients per 100 grams in a source such as USDA FoodData Central, add them up for the batch, and divide by the finished weight. FDA's guide to nutrition labeling databases describes a recipe database as software that calculates label values from a product's ingredients, "taking into account nutrient losses during processing." The step-by-step guide to a label without a lab covers how to do it.
A lab label measures the finished food. When FDA checks a label, it analyzes a composite of 12 units taken from 12 different shipping cases using AOAC methods (standard lab test methods published by AOAC International, a scientific standards body), under the compliance section of FDA's nutrition labeling rule. A commercial lab analyzing your product uses the same kinds of methods, so its result is the closest thing to what an FDA test would find.
A lab result isn't automatically exact, either. It measures the sample you sent. FDA's Food Labeling Guide says the number of samples needed depends on how much each nutrient varies in your product, and that it's up to the manufacturer to work that out. A single sample from a single batch tells you about that batch.
FDA's own advice is to use both. The same guide says that companies using ingredient databases should validate the results by comparing them with lab values for the same foods.
When is a database label enough?
The database method works when two things are true. Each ingredient has a database entry that matches what you actually use, and the change from raw ingredients to finished product is mostly water leaving.
Baked goods, granola, nut butters, dressings and spice blends usually fit. Simple sauces and jams fit too, but they are reduced on the stove, so weigh the finished batch rather than estimating it. The post on cooking moisture loss explains how to account for the water. Getting the entries right is the bigger job; see choosing FoodData Central entries.
Utah State University Extension, in its guide to Nutrition Facts panels (a secondary source), says it is not always necessary to have a sample tested at a lab, and calls database calculation "a more economical way to get started, especially if you are selling primarily direct-to-consumer."
When is a lab test worth paying for?
Fried foods: the recipe can't tell you how much oil stayed in
Frying adds fat in an amount the recipe doesn't record. You know how much oil is in the fryer, not how much stayed in the food. A 2023 review in Heliyon, on frying oils and fat uptake, reports examples where fat content in french fries rose from 0.2% to 14% during frying, and potato chips reached as much as 40%. The review lists the factors that change uptake: oil temperature, frying time, the food's moisture content, its surface texture, and the cooling period after frying, when much of the absorption happens.
Weighing the finished product doesn't solve this. The weight change mixes water lost and oil gained, and you can't separate them on a scale. Fat and calories are among the nutrients where FDA's test fails a label that understates them by more than 20%. For fried products, a lab test is the more reliable way to get fat and calories right.
Fermented foods: sugar changes during processing
Fermentation can leave a product with less sugar than its ingredients put in, and FDA's rule recognizes this. When added sugars are reduced through fermentation or browning, a manufacturer must do one of three things:
- keep scientific data showing the amount of added sugars left in the finished food, specific to that type of food, with a written explanation of why it's sufficient
- keep records of the added sugars put in before and during processing, and declare that amount, never more than the total sugars on the label
- petition FDA for an alternative means of compliance
A database can't see any of this. A lab can measure the total sugars in the finished product, which is the number most likely to move. It can't tell added sugars from naturally occurring ones in the same food, so added sugars still come from your records or supporting data. The post on added sugars and sugar alcohols has more on that line.
Kombucha, sourdough with a long ferment, yogurt, kimchi, hot sauces fermented before bottling and similar products fit this category.
Heavy cooking: nutrients are lost, not just water
Weighing the finished batch fixes the water. It doesn't account for nutrients lost in other ways. Some vitamins break down with heat. Minerals are lost mainly when cooking liquid is drained away and takes them with it. USDA's Table of Nutrient Retention Factors (Release 6, 2007) gives retention factors for vitamins, minerals and alcohol for about 290 foods, across preparation methods that include baking, boiling and draining. A recipe calculation that doesn't apply factors like these assumes nothing is lost.
That matters most for vitamin D, calcium, iron and potassium, which are on every label, and for any vitamins you declare voluntarily. It also matters for long reductions, where the finished weight is hard to measure on a large batch, and for products where something leaves besides water, such as fat rendered off during cooking and poured away.
An ingredient no database covers
If an ingredient isn't in FoodData Central, the database method needs another source for it. A supplier's specification sheet with nutrient data can fill the gap. If the supplier can't give you one, or the ingredient is your own blend whose makeup you don't know, the only reliable number comes from analysis. That can mean testing the ingredient once and using the result as a custom ingredient, or testing the finished product.
Branded entries in FoodData Central come from the manufacturer's own label, according to USDA's data documentation. Those values are already rounded to label increments, which adds a little error when a branded ingredient makes up a large share of your product.
A claim close to the threshold
A claim such as "good source of fiber" or "low fat" requires a full Nutrition Facts panel, under FDA's general rules for nutrient content claims. The claim also has a definition the product has to meet. "Good source" means 10 to 19% of the Daily Value per reference amount. "Low fat" means 3 g of fat or less per reference amount, and per 50 g for foods with a small reference amount. Nutrient content claims for small food producers covers the rest of the claim words.
Example: a granola bar has a reference amount of 40 g, and the Daily Value for fiber is 28 g. "Good source of fiber" needs at least 10% of 28 g, which is 2.8 g per 40 g bar. If your calculation says 3.0 g, you have 0.2 g of room, about 7%, between the estimate and the claim. That is a thin margin for a number built from average ingredient values, so it's worth confirming with a lab before you print the claim. If the bar isn't "low fat," a fiber claim also requires you to state the total fat per serving next to the claim, so the fat figure needs to be right too.
A buyer who asks for lab results
Some retailers and distributors ask for a Nutrition Facts panel even when the law doesn't require one; who needs a Nutrition Facts label covers that. Whether they want lab results behind it is a separate question. Ask for the buyer's spec sheet or vendor requirements before you decide. If it says lab analysis, that settles it.
What does lab analysis cost?
Medallion Labs publishes its 2026 price catalog, with prices effective through December 2026. Its listed prices include:
| Test | Price per sample |
|---|---|
| Nutritional analysis, every nutrient the label requires | $1,656 |
| Nutritional analysis with a mock nutrition label | $1,714 |
| Basic composition: ash (minerals), moisture, protein, fat | $354 |
| Vitamin D | $300 |
| Sugar profile (each sugar measured separately) | $135 |
The nutritional analysis package covers every nutrient the Nutrition Facts label requires (the catalog calls them the "required NLEA nutrients," after the 1990 law that created the label); the separate tests are for checking one nutrient on its own. The catalog notes that nutritional analysis pricing may vary with the food matrix, and that rush work completed in five or fewer working days carries a 100% surcharge on the list fee.
Great Lakes Scientific posts a price list dated effective January 1, 2023, so check with the lab for current rates. On that list, nutritional analysis for an FDA nutrition label is $1,080 per sample for one to four samples and $1,026 per sample for five to fifteen, and the listed package includes vitamin D, calcium, iron and potassium. Whichever lab you use, confirm that the package covers every mandatory nutrient on the current label before you order.
Those are two labs' published prices, and other labs quote on request. What moves the total for a small producer:
- How many products and flavors you test. Each formula is its own sample, so a line of six flavors costs roughly six times as much as one, less any volume discount.
- How many samples per product. One sample is the least expensive, but several batches tell you more about variation.
- Which tests you add beyond the standard package, such as a specific fiber method or a sugar profile.
- Turnaround. Rush service can double the price.
- Re-testing. Changing the formula means the old result no longer applies.
Some universities calculate a Nutrition Facts panel from your formula for a fee. The University of Tennessee's Department of Food Science, for example, charges $150 per sample for a panel developed from ingredients and formulation, and says it doesn't analyze samples. That is still database analysis, done by someone else.
Label most products by database and test the riskiest
For most small producers, the sensible pattern is to label most products by database and send the riskiest ones to a lab. Riskiest means fried, fermented, carrying a claim, or headed to a buyer who asks for lab data. When you do get a lab result back, compare it with your calculation for the same product, as FDA's guide suggests. If they agree closely, that is good evidence your method works for similar products. If they don't, the gap usually points to a specific ingredient entry or a moisture assumption you can fix.
KitchenSync is recipe software for food businesses: recipes and their versions, costing from your supplier prices, and FDA nutrition labels with allergens. It handles the database side: nutrition from USDA FoodData Central entries, weighted by the gram amounts in your recipe and divided by the finished weight you enter. The nutrition label generator page has the details. A small snack or sauce line can label its straightforward products from the recipe and save the lab budget for the fried or fermented ones.
This is general information about federal labeling rules, not legal advice. Your product, your state and your buyers can change the answer, so check with your state agriculture or health department, or a labeling professional, about anything you're unsure of.
Sources
- 21 CFR 101.9, Nutrition labeling of food, Electronic Code of Federal Regulations
- 21 CFR 101.13, Nutrient content claims: general principles, Electronic Code of Federal Regulations
- 21 CFR 101.54, Nutrient content claims for "good source," "high," "more," and "high potency", Electronic Code of Federal Regulations
- 21 CFR 101.62, Nutrient content claims for fat, fatty acid, and cholesterol content of foods, Electronic Code of Federal Regulations
- A Food Labeling Guide: Guidance for Industry, U.S. Food and Drug Administration
- Guidance for Industry: Guide for Developing and Using Data Bases for Nutrition Labeling, U.S. Food and Drug Administration
- FoodData Central Data Documentation, USDA Agricultural Research Service
- USDA Table of Nutrient Retention Factors, Release 6 (2007), USDA Agricultural Research Service, via Data.gov
- A review of different frying oils and oleogels as alternative frying media for fat-uptake reduction in deep-fat fried foods, Mahmud et al., Heliyon, 2023
- 2026 Price Catalog, Medallion Labs
- Price List, Great Lakes Scientific
- Services, University of Tennessee Department of Food Science
- Nutrition Facts Panels, Utah State University Extension (secondary source)
