You log the same plate in two apps, get two different answers, and conclude that you must be logging it wrong. Mostly you are not. The reference data behind the two apps genuinely disagree, and the largest error sources on an Indian plate are structural rather than personal.

One thing to settle immediately: the 1,679 and the 1,069 in the title are illustrative. No study or institutional comparison reports that pair for any named Indian dish. They stand in for a disagreement whose scale is documented, which is what follows.

Key Takeaways:

• India's national food composition tables cover 528 raw foods and exclude composite cooked dishes such as chapatis and curries.
• Reported database-to-database differences for common single foods run roughly 6 to 25%.
• Rice roughly triples in weight when cooked, lentils and chickpeas rise about 2.4 to 2.5 times, spinach loses around 70%.
• Self-reported food records under-report energy by 11 to 41% against doubly labelled water, usually the larger error.
• No source reports a 1,679 against 1,069 comparison for any named Indian meal.

Both apps show a whole number, to the kilocalorie, with no interval around it, and that presentation produces the belief. A figure displayed as 1,679 reads as a measurement. It is an estimate, produced by matching what you typed to a row in a table compiled elsewhere, for a food possibly analysed decades ago, in a form you did not eat.

Energy values in composition tables are not measured in a calorimeter entry by entry. They are calculated from a food's proximate composition using standard conversion factors, so two tables analysing the same grain, with different sampling and methods, produce two defensible values. The disagreement was in the source data before either app rendered it.

The Indian Food Composition Tables 2017 are the national scientific reference, produced by the National Institute of Nutrition under the Indian Council of Medical Research. They cover 528 key foods, chosen because together they account for roughly 75% of population intake of energy, fat, protein and eight micronutrients, with values for 151 nutrient components.

Energy in IFCT 2017 is calculated rather than measured, applying 4 kcal/g for protein and carbohydrate and 9 kcal/g for fat. Coverage is raw, single ingredients: uncooked rice, wheat flour, raw tomatoes. Composite cooked dishes are explicitly excluded.

Two other lineages sit alongside it. The Gopalan-era tables, first issued in 1937 and revised repeatedly, remain embedded in many commercial tools because they circulate freely as PDFs, and where old and new overlap they differ through updated methods, sampling and unit conversion. USDA FoodData Central runs on different chemistry again, using general or food-specific Atwater factors and reporting carbohydrate by difference rather than by direct analysis, with almost no native coverage of Indian composite dishes.

How far apart those three lineages sit for one standard plate of anything has never been published. What exists instead is the 2024 build of the Indian Nutrient Databank, which had to blend sources to assemble a single Indian recipe database: IFCT 2017 for 528 raw foods, IFCT 2004 for another 369, the UK's Public Health England tables for a further 144, and USDA retention factors to estimate cooking losses.

That patchwork is close to what happens invisibly inside a commercial app's backend. Two apps can each use authoritative data and still diverge, because they patched the same holes differently.

The app layer adds its own spread. Validation against a national reference database found strong correlation for total energy in one widely used tracker, at r=0.96, alongside underestimation of protein by 7.8% and carbohydrate by 6.4% once implausible entries were removed, so the uncleaned catalogue performs worse than those figures suggest.

Serving standardisation decides what the number is attached to. ICMR-NIN's dietary guidance defines a medium katori as 200 ml, while other Indian teaching material in circulation defines one katori as 150 ml and one cup as 240 ml. Indian institutional sources do not agree with each other on the size of a katori, and most apps map their dropdowns to none of these, defaulting to grams or to an unspecified bowl.

Raw-versus-cooked weight is the largest single trap. IFCT values are per 100 g of the uncooked item, and people log a cooked, plated portion. Rice roughly triples in weight, lentils and chickpeas rise about 2.4 to 2.5 times, and spinach loses around 70% cooking down. Weighing 150 g of cooked rice and logging it against a raw-basis entry misstates the energy of that food by roughly two- to three-fold, and ICMR's practical manuals name the conversion precisely as the water absorption ratio.

Cooking fat is the error nobody weighs. The tadka is poured, not measured, and household consumption data show per-capita edible oil purchase rising from about 0.37 kg a month in rural India in 1993-94 to about 0.67 kg by 2011-12, against an ICMR recommendation of roughly 30 g per person per day.

Laboratory analysis of fried items shows the stakes: about 26 g of visible fat in one plate of pakodi, about 25 g in a large samosa. Most uptake happens after the pan, with 65 to 88% of absorbed oil in some products entering during post-fry cooling. The everyday case is least studied, since the laboratory data concentrate on fried snacks rather than routine dal and sabzi tempering.

Then there is who cooked it. One survey of street snacks in a single Indian city found a mean around 250 kcal per serving, with category means running from 239 to 311 and vendor-to-vendor variation wide enough that any one entry can only be an average.

Under-reporting usually dominates, and that deserves saying plainly. Measured against doubly labelled water, the closest available reference for true energy intake, a systematic review of dietary assessment methods found food records under-report energy by 11 to 41% and 24-hour recalls by 8 to 30%. One app-based validation classed 53% of participants as under-reporters, averaging 563 kcal a day below measured expenditure. Database-to-database discrepancies of roughly 6 to 25% for common single foods sit well inside that.

The exception matters most here. For high-oil or composite home-cooked Indian dishes, the raw-versus-cooked and oil-absorption errors can match or exceed typical under-reporting, because a two-fold error on the rice and 25 g of unlogged oil are not small percentages. Both halves hold at once.

Every app accuracy figure above comes from non-Indian populations checked against non-Indian reference databases, and no equivalent audit of the apps Indians use has been published. IFCT 2017 does not publish regional variance ranges for its 528 foods either, so cultivar and regional variation within India cannot be sized from public sources.

Comparing you against you is the job this number can do. An estimate carrying that much uncertainty is near useless as absolute truth about a meal and reasonably useful as a direction of travel, provided the method stays identical: same app, same entries, same basis, week after week. There is no official target to align to either, since no ICMR-NIN or FSSAI statement was located endorsing or rejecting a single reference database for consumer trackers.

Where the number is going to drive a decision about health rather than habits, the question belongs with a registered medical practitioner or a registered dietitian who can see what you actually eat.

Your calorie app gives you a direction, not a destination. A Voy clinician can read what you actually eat alongside your blood results, book a consultation.

This article is for general information and education only and is not medical advice, diagnosis, or treatment. GLP-1 and other medications referenced are prescription-only and are appropriate only for certain people under the supervision of a qualified clinician. Do not start, stop, or change any medication based on this article. Please consult a registered medical practitioner about your individual circumstances. Information reflects what was available at the time of review and may change.