How accurate is Cronometer? Its named databases and the studies that tested it

Is Cronometer accurate enough to trust its food numbers?

Cronometer says it draws on more than 10 nutrition sources, with NCCDB supplying most foods and a curation team checking user submissions for its community database. In Morello's athlete records, mean energy was 2502.4 kcal in Cronometer against 2513.4 in the reference and protein was 121.5 g against 119.9, while Gioia found no significant calorie or macronutrient differences for a three day record. Yet Bayram and Ozturkcan placed Cronometer among the apps with the greatest discrepancies.

Morello concluded that validity was good apart from fibre and vitamins A and D. These studies do not supply a single accuracy figure, and we give no figure of our own. The five detailed study accounts below draw on three full papers, one abstract and one meeting abstract. Company material and research were checked October 13, 2026.

Ten database names and what Cronometer says each contains

The company's overview lists these names in this order. It supplies no database makers, food counts, nutrient counts or checking methods for the individual sources. The descriptions and counts below instead come from its Data Sources help article, checked October 13, 2026.

CRDB is absent from the overview's ten names. The help article describes eleven named sources, including CRDB, plus custom foods. Cronometer describes CRDB as its own collection of foods submitted by users, usually from product labels, and reviewed by its curation team. Its CRDB food count is Not stated.

Its overview says nutrition comes from lab-analyzed data and not crowd-sourced guesses. Its help article describes reviewed user submissions in CRDB. The same company makes both statements. Custom foods, the help article says, are private to their creator and friends with whom they are shared.

The overview says the industry recognizes Cronometer as the most accurate nutrition tracker. The overview names no supporting study, test or ranking and does not identify the industry behind that claim.

Source letters, food symbols and submission review

Cronometer's getting started post says letters beside search results identify the nutrition source, with NCCDB, USDA and CRDB as examples. Its food search post describes a source and nutrient count on the food's own screen. The web help page also describes viewing the source and nutrient count before adding a food. The exact field label on that food screen is Not stated in the help center.

Food search symbols distinguish kinds of data. The Data Confidence article says the red laboratory beaker marks what Cronometer calls high-quality data items and the barcode marks data from a product label. It says Data Confidence assesses each nutrient value using the food type and source to indicate how complete the data is.

The web filter is called Source, and Common Foods is described as drawing from the company's highest quality databases. On mobile, the Database filter can select one source or several, with the selection shown above the results.

Cronometer says every food submitted by a user receives curation review before public inclusion. Submitted branded products contain only information from the package nutrition table or the brand's official website. That differs from a private custom food, visible only to its creator and friends receiving it through sharing.

For an unmatched barcode, the mobile scanning article describes creating a custom food that is automatically submitted to the public database. It says the food is added after review and that the submitted photos must be readable and complete so the team can verify the product.

The CRDB publication article limits submissions to common packaged, store-bought or restaurant foods with nutrition information on the package or official brand website. The custom food article requires clear photos of the package front, including brand and product name, and the nutrition panel. Cronometer's mobile publication article says hundreds of foods arrive daily and warns that adding them may take time.

Across these data source, scanning and submission pages, the reviewers' identities and number, review duration, the share of rejected submissions, how often existing community foods are checked against current labels, and a measured database error rate are Not stated.

Research records behind the Cronometer findings

Research checked October 13, 2026. Of the five detailed studies, Morello, Gioia and Li were read in full through PubMed Central. Bayram and Ozturkcan was read only as a PubMed abstract. Evenson and Berens was read as a published meeting abstract, which is the whole publication.

Morello's athlete records and shared Canadian data

Morello and colleagues, 2025, Journal of Human Nutrition and Dietetics, began with 50 Canadian endurance athletes aged 40 to 65 who completed three day food records. Seven records were incomplete, leaving 43 athletes, 27 men and 16 women. Two independent researchers entered the records in 2020 using free Cronometer version 2.18.6. Updates were disabled and they used no barcode scans.

The reference was ESHA Food Processor with the 2015 Canadian Nutrient File, entered by a trained researcher. Cronometer selections also used the 2015 Canadian Nutrient File whenever possible. Otherwise, the raters chose lab-analyzed sources such as NCCDB and USDA. The Cronometer entries and the reference came from the same Canadian database wherever the raters could pick it.

The comparison measured three day mean daily energy and the nutrients in Table 4 below. For Cronometer it also assessed vitamins A, D, B3, B6 and B12, calcium, potassium, iron, magnesium and selenium.

Morello used p below 0.025 for the follow-up comparisons. Where the table says none printed, the overall test across the three tools was not significant, so no follow-up p value appears. Sugar and fibre differed significantly, with the fibre difference also described as clinically meaningful.

Table 6 also reports significant and clinically meaningful differences for vitamins A and D. Vitamin D was 178.1 IU in the reference and 229.6 IU in Cronometer, with p 0.013. Vitamins B3, B6 and B12, calcium, potassium, iron, magnesium and selenium did not differ significantly at the paper's threshold.

For energy, Cronometer's intraclass correlation between raters was 0.966. That measures agreement between people entering the same records, not accuracy against the reference. The authors judged validity good except for fibre and vitamins A and D, and agreement between raters excellent except for iron, with no differences between genders.

Morello's MyFitnessPal results differed significantly from the Canadian reference for cholesterol, sugar and fibre, with trends toward differences in energy and protein. The MyFitnessPal research article covers that app's published tests.

The authors point to different fibre fields and differences in fortification between countries and brands as likely reasons for the fibre and vitamin gaps. They note that the Canadian reference is updated infrequently, sometimes contains outdated values and sometimes uses generic averages instead of brand data. Entry order across the tools was not randomized. They report no specific funding and declare no conflicts of interest.

Gioia's three day record and four sample foods

Gioia and colleagues, 2023, JMIR Formative Research, had no study participants. One researcher supplied a food record spanning two weekdays and one weekend day. The separate sample foods were a medium orange, three Oreo cookies, a peanut butter and jelly sandwich and a Burger King Whopper. Eleven apps were reviewed for meal timing research, with nutrient comparisons covering five apps.

A single researcher entered foods through text search, choosing the first result, or the first verified result where that option existed, with identical portion sizes across apps. A registered dietitian nutritionist used the same record and sample foods in the Nutrition Data System for Research, or NDSR, to produce the reference values.

The measurements were calories, carbohydrates, fat and protein. Vitamins and minerals were not compared. Table 6 gives these totals across the three days.

Gioia reports that Cronometer alone had no significant calorie or macronutrient differences for the three day record. Bitesnap, Macros and MyFitnessPal had significant calorie differences from NDSR. Across the four separate sample foods, Cronometer had an average difference of 0.21 kcal. Carbohydrates differed significantly for those foods, with p 0.02, but calories, fat and protein did not.

The authors say a second reviewer did not verify the entries, so input errors were possible. They also note that the selected foods may not represent the wider population's diet and that the accuracy assessment used text search only. Funding came from National Institutes of Health grants and hospital awards. The conflicts statement reports one author's consulting income from other companies and states no link to Cronometer.

Li's diet plans and nutrient table values

Li and colleagues, 2024, Nutrients, used diet plans rather than participants. Dietitians created three day records for Western, Asian and Recommended diets. Three researchers rotated entry duties. Sixteen apps from Australian app stores supplied nutrition data, including Cronometer version 4.19.5.

Foodworks.online supplied the benchmark, using the three day averages for each diet. Measurements covered energy, macronutrients as a share of energy, and fibre, calcium, iron and sodium per megajoule. Energy findings are pooled in the text, with no Cronometer energy figure in the text or tables.

Li's Western diet paragraph says Cronometer's carbohydrate share was higher than the reference and describes it as 7% over the reference. The paper's Table 6 values below are shares of energy in percent.

For carbohydrate share in the other Table 6 diets, the Asian reference was 46 and Cronometer was 51. The Recommended diet reference was 48 and Cronometer was 58. These are also percent of energy.

Table 7 lists sodium in mg per megajoule. Western diet values were 278 for the reference and 876 for Cronometer. Asian diet values were 402 and 875. Recommended diet values were 201 and 159. The paper makes no statement about Cronometer's sodium figures.

Li notes that funding constraints restricted the work to free apps and features. The researchers' nutrition knowledge and dietetics training likely made their food matching more accurate than public users' matching. The authors report no external funding and no conflicts of interest.

Bayram and Ozturkcan's abstract on standardized menus

Bayram and Ozturkcan, 2026, Nutrition and Health, was read as an abstract only. It describes standardized food records and seventeen evaluated apps from the App Store in Türkiye, with no participants described. Who entered the foods is Not stated in the abstract, and it does not identify the software used for the reference records.

The abstract compares energy and macronutrients with those reference records. It names Cronometer among the apps with the greatest discrepancies, but supplies no Cronometer figure. The authors say the findings are limited by the use of standardized menus. Funding and conflicts are Not stated in the abstract.

Evenson and Berens' meeting abstract on student records

Evenson and Berens, 2021, Current Developments in Nutrition, is a meeting abstract, not a full paper. It uses one day food records from 30 introductory nutrition students. Two researchers entered the records after demonstrating agreement between raters above 0.90. They compared app nutrient values with ESHA Food Processor.

The abstract reports Cronometer energy at p above 0.05 and fat at p 0.249, neither significantly lower than ESHA. Carbohydrates were significantly lower, with p 0.026, and sodium was lower, with p 0.019. Protein and fibre did not differ from ESHA across the apps, with p above .05.

Only p values are supplied for these results, without intake amounts or the size of the differences. No funding was provided.

Outside those five accounts, Chen, Cade and Allman-Farinelli, 2015, JMIR mHealth and uHealth, compares app energy totals with a weighed food record, pools its results and names Cronometer only in a quality table, without a Cronometer figure.

Cronometer presents its own research list as studies in which the app was used. Most recorded or analyzed diets with it rather than testing its numbers. The list includes Morello and the Evenson and Berens meeting abstract as named tests, along with Chen's pooled paper.

The chosen entry and the entered portion still matter

Which food record supplies the values

Morello preferred Canadian Nutrient File entries in both Cronometer and the reference. Gioia instead selected first text search results, or first verified results where available. Li says trained researchers likely matched foods more accurately than public users. These are the papers' descriptions of entry selection.

Cronometer directs people seeking detailed nutrient information to NCCDB or USDA SR28 in Common Foods. For matching package numbers, it points to mobile barcode scanning or entering the barcode in web search. It says label entries can contain fewer nutrients than generic USDA or NCCDB foods.

Cronometer's completeness article says not every food has all the nutrient data points the app displays. Branded, packaged and chain restaurant items often have gaps. It says American labels usually report about 15 nutrients. Its potassium example explains that a missing database value makes the daily total undercount potassium even when the food contains it.

The company's Brazil nut example gives 17 nutrients for a scanned Trader Joe's product and 76 for a generic unsalted NCCDB entry. The same post describes nutrition values as averages of several laboratory tests and cautions that they are not precision truths.

Cronometer's reporting article says the company is not a laboratory and does not test foods for nutrient content. It says product listings can provide only the manufacturer's label information. It also describes NCCDB and USDA records as compiled from laboratory analyses and research papers, with rounding differences that may prevent an exact package match.

How much food the entry represents

Cronometer's accuracy tips say some serving choices are subjective and recommend weighing ingredients and entering their gram weight. Its reporting article points out that the package and database can use different serving sizes. Its raw or cooked help article uses broccoli to illustrate measuring a raw entry before cooking and a cooked entry afterward.

For photos, Cronometer describes AI identification of ingredients and quantities as an educated guess from the image. It says the identified ingredients are then matched to its existing verified database, with nutrition values from lab-analyzed sources rather than generated by AI.

What can be inspected for one food in the log

1. The letters next to a search entry identify its nutrition source. Cronometer gives NCCDB, USDA and CRDB as examples.
2. The food screen lists its data source and nutrient count. The web help article describes inspecting both before adding the food.
3. A red beaker marks what Cronometer calls high-quality data items and a barcode symbol marks data from a product label, according to Cronometer's icon explanation.
4. The package serving size can be compared with the entry's serving size. Cronometer notes that these amounts can differ.
5. The mobile report route begins with the three dots in the top right corner of the Nutrition Summary. Cronometer asks for reports when a confirmed database value differs from the package nutrition table. Its mobile article says issues are typically addressed within 1 business day. The web report article states 2 business days.

Other trackers' sources and tests on our other pages

• MyFitnessPal research, including differences between study designs.
• Cal AI accuracy claims and the research available for photo logging.
• A survey of tracker data sources and published evidence.
• The source information and review marks described by calorie trackers.
• A broader Cronometer review covering logging, nutrient detail and plans.
• Cronometer subscription costs, renewal terms and discounts.
• MyFitnessPal and Cronometer plan access, scanning and food databases.
• Cronometer and MacroFactor subscriptions and ways to enter food.
• Cronometer and Lose It free tools and paid logging features.
• Cronometer and MyNetDiary free access, nutrient fields and paid plans.
• TrueCal and Cronometer pricing, nutrient detail and logging options.

Where TrueCal's nutrition estimates begin

The sourcing rule requires every nutrition number to come from a source or the user. Homemade meals use sourced ingredients. Sources include USDA FoodData Central, FatSecret, Open Food Facts, and published restaurant and brand information. Barcodes use nutrition databases, while label photos read the package directly.

Nutrition results are estimates and can be wrong. They depend on the meal details supplied, including portions and ingredients, as explained in the product reference.

Questions about Cronometer food data