How accurate are calorie apps for Indian food? A test of five apps on 42 staples

1. Introduction

A calorie app is only as good as the entry a person picks. Most apps hold several entries for the same food: lab values, brand labels, recipes and entries typed in by other users. The search box decides which one people see first, and many people take the first one.

Earlier studies compared app totals with research software on Western diets and found agreement for energy that was often reasonable when the right foods were entered^{[3, 4, 5]}, but weaker when people chose entries themselves^[6]. A 2024 study of manual logging apps found energy was underestimated for Asian diets while it was overestimated for Western ones^[7]. We could not find a study that took plain Indian staples one at a time, searched them the way a user would, and checked the first result against India's national food tables.

Our question was simple: if you type an everyday Indian food into a calorie app, is the first thing you see the right food with the right calories?

2. Methods

2.1 Foods and search words

42 staples in seven groups, each in a plain form with one official value. Grains and pulses were tested dry, the form IFCT 2017 lists. We typed the same words in every app. Where the plain word usually means a cooked dish (dal, rice, meat), we added "raw", which applied to 11 of the 42 searches. The search words are in Table 3 and the CSV.

2.2 Apps, devices and dates

• HealthifyMe (version 31.5), Cronometer (4.59.5) and YourTrainer (1.0.3) on an Android emulator with a Pixel 8 profile, free accounts, in India.
• MyFitnessPal and FatSecret through their public food search websites, signed out, from India.
• Times in IST. Websites: 26 October, 12:06 to 12:15. HealthifyMe: 26 October, 12:03 to 12:49 (potato repeated at 22:34 because its detail pages had not loaded). Cronometer: 26 October, 13:16 to 14:09 (five searches repeated 22:24 to 22:32 because the list had not refreshed). YourTrainer: 27 October, 11:14 to 11:34.

2.3 Reference values

IFCT 2017, the Indian Food Composition Tables from the National Institute of Nutrition^[1]: energy per 100 g edible portion, converted from kJ by dividing by 4.184. USDA FoodData Central, SR Legacy^[2]. Five foods (poha, ragi flour, paneer, rohu and jaggery) have no close USDA match and use IFCT only. The food codes are in the CSV.

2.4 Reading each result

Each entry was converted to kcal per 100 g using the serving weight the entry itself showed. MyFitnessPal entries with a household measure and no grams used the USDA portion weight for that measure (one large egg is 50 g); with no USDA portion they were marked "no gram weight". FatSecret rows with no grams used FatSecret's own gram weight for that food, read from its database on the same day. Tiny servings ("1 g, 4 calories") are rounded by the app, so we took the value in the rounding range nearest the reference.

2.5 Grading

• Accurate: within 10% of IFCT 2017 or USDA, whichever is nearer.
• Off: 10 to 25% away. Wrong: more than 25% away.
• Different food: a cooked dish, snack, sweet, drink, another part of the plant or animal, or another food. The same name rules were applied to every app, then every accepted entry was checked by hand against its own screen. Entries that matched by name but showed another food on their panel are listed in the evidence browser below.
• First result is the top entry. First plain match is the first entry in the list that is the plain food, within the first 5 results in the three apps (each opened) or the first 10 on the two websites.

2.6 Our own app

The YourTrainer test account had logged foods before, and the app lists those under "Your foods" above the library. A new member does not see that section, so we graded the library results below it. YourTrainer's plain raw entries come from the same IFCT 2017 and USDA tables we grade against, which favours it on this test. See the disclosure under Declarations.

2.7 Statistics

We report counts with percentages and 95% confidence intervals by the Wilson score method^[8], and the median percentage gap for first results that were the right food. With 42 foods per app, differences of a few foods between apps are within the margin of error, so we do not rank the four established apps against each other.

2.8 Screenshots

Every search and opened entry was saved. Screens are unedited apart from resizing, cropping the two websites to their results column, and grey boxes over email addresses, phone notifications and FatSecret's "Recent Food Diary Entries" panel (other people's diaries). The numbered strip above each screen was added by us.

3. Results

3.1 What came up first, app by app

3.2 By food group

3.3 Typing "raw" did not keep cooked dishes out of first place

For "toor dal raw", HealthifyMe's first result was cooked toor dal by the katori. "Urad dal raw" gave Urad Dal Fry, "rajma raw" gave Rajma Curry and "kala chana raw" gave Kala Chana Masala. Cooked dal is mostly water, so Urad Dal Fry works out at 92 kcal per 100 g against 324 kcal for dry urad dal. Someone who weighs 60 g of dry dal and logs it against the cooked entry records about a quarter of what they ate. The plain raw entry was there, at position 3.

3.4 FatSecret's website matched the word, not the food

For six raw dal and bean searches and "mutton raw", the first result on foods.fatsecret.com was "Raw Vegetable" at 24 kcal a cup. "Maida" returned the Maid-Rite sandwich, and "white rice raw" and "brown rice raw" both returned Cream of Rice. In all, 17 of FatSecret's 42 first results were a different food, including 9 of the 11 searches that had the word "raw".

3.5 Cronometer had good entries, but often not first

Cronometer's first result was a different food 15 times out of 42: "Bananas, Raw" for "rajma raw", "Chana Dal, Samosa" for "chana dal raw", a fruit called poha berry for "poha" and Honey Maid graham crackers for "maida". Yet an accurate plain entry sat in its first five results for 34 of 42 foods, the most of the four established apps. The data was there; the ranking put something else on top.

3.6 Even the right food sometimes had the wrong number

FatSecret's first "Ragi Flour" works out at 523 kcal per 100 g against 321 in IFCT 2017. MyFitnessPal's "Generic, Mutton Raw" gives 234 kcal against 160 (IFCT) and 109 (USDA). HealthifyMe's raw potato gives 97 against 70 and 77, and both HealthifyMe and MyFitnessPal put paneer at 265 against 305. These are the right foods, so a user has no reason to doubt them.

3.7 The official tables disagree too

IFCT 2017 gives guava 32 kcal per 100 g; USDA gives 68. Papaya is 24 against 43, pomegranate 55 against 83, and raw chicken breast 168 against 120. Varieties, growing conditions and lab methods differ. That is why a value within 10% of either table counted as accurate, and no app was marked down for following one of them.

3.8 All 42 foods

4. Discussion

The numbers inside these apps were mostly fine. When the right food came first, the median gap to the reference was 2%. What went wrong was the choice of entry. A quarter of first results were a different food, and for dals and beans, the backbone of most Indian meals, it was more than half.

The errors also lean one way for the foods that matter most. A dry dal logged against a cooked dal entry, or raw rice against Cream of Rice, records far less energy than was eaten. That fits the 2024 finding that manual logging apps underestimated energy for Asian diets^[7], and the older finding that results get worse when people choose entries themselves^[6].

The fix sits with the apps: rank the plain ingredient above dishes when someone types "raw" or a flour name, and treat a search word as a food, not a string to match. Until then, the user has to check. The tips below take seconds.

What to check before you log

1. Open the entry, read the name. "Dal fry", "curry", "masala" and "by katori" are cooked dishes. If you weighed the dry food, scroll for the plain one.
2. Look for a gram weight. "1 serving" or "1 cup" with no grams cannot be checked.
3. Sense-check per 100 g. Dry dals in IFCT 2017 are 322 to 331 kcal per 100 g, atta 320, raw rice 356. A "raw dal" at 100 to 170 kcal is cooked dal. Our Indian food calorie calculator lists common foods per 100 g.
4. Weigh dry and log dry, or weigh cooked and log cooked. Never mix the two. For cooked food at home, the katori portion calculator turns katoris and rotis into grams.
5. Scroll past the first result. Taking the first plain match raised accuracy from 59% to 71% in this test.

More on this: how to log Indian food in a calorie app, Indian portion sizes in grams and dal nutrition compared. To see your own daily target, try the calorie calculator.

5. How our members log: a photo to the nutritionist

A food search cannot see the katori size, the ghee on the roti or the side dish that never got logged. That is why our members can also send a photo of the plate to their coach or nutritionist in the YourTrainer app, and why we do not rely on the search box alone. These are some of those photos, shared without names, food only.

Real member stories: the questions members ask their nutritionist in week one, what changed in three weeks of coached food logging, a weight plateau broken, week by week and case studies from coaching sessions. Why we do not trust photo calorie scanners on their own: a scan said 1,000 calories for an Indian plate. More results are on our results page.

6. Limitations

• A snapshot. Food databases and search ranking change often, and results can differ by account, history and location. These are the results on the dates shown.
• Websites for two apps. MyFitnessPal and FatSecret were tested through their public websites. Their phone apps may rank results differently.
• Scope. Android only, free accounts, English search words, one search per food. 42 staples are not a sample of everything people eat, and cooked dishes, where recipes vary, were not graded.
• Calories only. Protein was recorded but not scored.
• Small numbers per app. With 42 foods per app the confidence intervals are wide; read differences between the four established apps with care.
• Our own app. YourTrainer made one of the apps and ran the test. Its entries use the same references we graded against. We publish every screen so anyone can check the grades.

7. Conclusions

For plain Indian staples, the established calorie apps we tested usually hold a correct entry, but put it first only 59% of the time. The common failure is the wrong food, not the wrong number, and it hits dals hardest. Opening the entry and checking its name, grams and calories per 100 g catches most of it.

Declarations

Data availability

All 210 graded results are in the CSV file (food, search words, IFCT code, USDA id, reference values, first result, first plain match, kcal per 100 g, gap and grade). Every screenshot is in the evidence browser above. The full set of raw screens is available on request.

Conflict of interest

YourTrainer published this study and makes one of the five apps tested. During the study YourTrainer added and corrected entries in its own food library for foods in this test, using the same IFCT 2017 and USDA references, and changed how its search ranks plain foods. Its result shows the version released to members on 27 October 2026. Read it as the publisher's own app, not an independent rating. The other four apps were tested as found on 26 October 2026.

Funding

Funded by YourTrainer. No outside funding. No app maker saw the results before publication.

Corrections

If you make one of these apps and an entry has changed, or you think we misread a screen, write to yourtrainerbusiness@gmail.com. We will check it, publish your reply if you wish, and note any correction on this page with its date. Journalists can reach us through the press page. The medical guidelines we coach from are on Health Sources, and more of our original data is on Health Data.

References

1. Longvah T, Ananthan R, Bhaskarachary K, Venkaiah K. Indian Food Composition Tables 2017. Hyderabad: National Institute of Nutrition, Indian Council of Medical Research; 2017. www.nin.res.in/ebooks/IFCT2017.pdf
2. U.S. Department of Agriculture, Agricultural Research Service. FoodData Central: SR Legacy. 2019. fdc.nal.usda.gov/
3. Griffiths C, Harnack L, Pereira MA. Assessment of the accuracy of nutrient calculations of five popular nutrition tracking applications. Public Health Nutrition. 2018;21:1495 to 1502. doi.org/10.1017/S1368980018000393
4. Evenepoel C, Clevers E, Deroover L, Van Loo W, Matthys C, Verbeke K. Accuracy of nutrient calculations using the consumer-focused online app MyFitnessPal: validation study. Journal of Medical Internet Research. 2020;22(10):e18237. doi.org/10.2196/18237
5. Fallaize R, et al. Popular nutrition-related mobile apps: an agreement assessment against a UK reference method. JMIR mHealth and uHealth. 2019;7(2):e9838. doi.org/10.2196/mhealth.9838
6. Chen J, Berkman W, Bardouh M, Ng CYK, Allman-Farinelli M. The use of a food logging app in the naturalistic setting fails to provide accurate measurements of nutrients and poses usability challenges. Nutrition. 2019;57:208 to 216. doi.org/10.1016/j.nut.2018.05.003
7. Li X, Yin A, Choi HY, Chan V, Allman-Farinelli M, Chen J. Evaluating the quality and comparative validity of manual food logging and artificial intelligence-enabled food image recognition in apps for nutrition care. Nutrients. 2024;16(15):2573. doi.org/10.3390/nu16152573
8. Wilson EB. Probable inference, the law of succession, and statistical inference. Journal of the American Statistical Association. 1927;22(158):209 to 212. doi.org/10.1080/01621459.1927.10502953