Estimate the container, then price the oil#
The move that makes takeout tractable is refusing to estimate "a chicken korma" at all. Estimate the container — how much food arrived, and what fraction of it you ate — then estimate the fat riding inside it, then everything else. Takeaway food is not a plate someone assembled in front of you; it is a mass of mixed food in a box, and the mass is the one thing you can actually observe.
That inversion matters because takeout differs from a sit-down restaurant meal in three specific, measured ways. The calorie number, where one exists at all, does not travel to the app you ordered from. The fat load is large enough that it dominates the estimate rather than decorating it. And the unit you receive is a package rather than a portion. The general biases in guessing a restaurant meal — how far low people run, and how to decompose a plate — are already worked out in estimating restaurant meal calories and apply here unchanged; this article is about the three things that don't carry over. The wider error stack sits in how accurate calorie counting is.
The number does not travel to the delivery app#
US menu labeling reaches only restaurants that are part of a chain of twenty or more locations trading under the same name, and 21 CFR 101.11 contains no provisions at all for third-party delivery platforms5. What that means in practice is not "independents are unlabeled" — that much is well known — but that the labeling which does exist was written for a menu board, and a menu board is not what you are looking at.
The scale of the gap has been mapped. Researchers systematically sampled the most popular outlets on Uber Eats across suburbs in Sydney and Auckland, capturing 1,074 unique outlets and 5,769 of their most popular menu items. Nutrient profiles for most popular menu items were simply not available on the platform. Of the items themselves, 84.3% in Sydney and 88.2% in Auckland were classified as discretionary rather than core foods, and roughly 73% of outlets in both cities fell into the least healthy category4.
Two things follow, and the second is the one people miss. First, you will usually be estimating from scratch. Second — and this is a genuine surprise from that dataset — the most popular delivery outlets were dominated by takeaway franchise stores, not independents. So the food you order most on a platform often does have a published calorie figure somewhere; it just isn't in the ordering flow, and a figure you have to leave the app to find is a figure most people never look up.
What a laboratory finds inside a takeaway portion#
Two lab measurements are worth carrying in your head, because they replace intuition with a prior.
The first is energy. When 157 meals — the 42 most frequently purchased dishes across nine cuisine categories, from randomly selected independent and small-chain restaurants near Boston — were burned in a bomb calorimeter, the mean was 1,327 kcal per meal (95% CI 1,248–1,406), equivalent to 66% of a typical day's energy requirement, with 7.6% of meals exceeding a full day's worth on their own1.
The second is fat, and it is the number that should change how you log. An accredited laboratory analyzed 74 meals and 26 side dishes from 27 takeaway outlets across eight London boroughs. The median standard meal carried 52.0 g of total fat per portion, with a range from 29.8 to 75.8 g; over 70% of all meals delivered more than 30% of a day's recommended fat intake in a single sitting, and the mean portion weighed 580 g3.
Do the multiplication yourself — this is my arithmetic on their median, not a figure the paper reports: 52 g of fat at 9 kcal/g is about 470 calories from fat alone, before a single gram of rice, bread or meat protein is counted. That is why a component-by-component estimate of a takeaway that "feels about right" is almost always short. The visible food was never the problem.
One more finding from that London study deserves a mention because it cuts against the obvious hope. Half the outlets sampled were participants in a Healthier Catering Commitment scheme. Their meals carried a median 43.6 g of fat against 52.0 g in standard outlets — and with one exception the difference was not statistically significant3. Scope that carefully: 27 outlets in one city, a cross-sectional snapshot, not a trial. It is a reason to keep your prior where it is rather than to conclude the scheme fails.
The same research group, two verdicts — and the reason#
Here the evidence genuinely splits, and the split is instructive rather than confusing.
Urban's 2013 analysis found independent and small-chain meals running 49% above popular meals from the largest national chains, and 19% above national food-database values1. Three years later the same group measured 364 meals from 123 non-chain restaurants across Boston, San Francisco and Little Rock, and found matching chain meals contained 68 kcal fewer — a difference that did not approach significance (P=0.41)2. Same team, same instrument, opposite conclusions about whether independents are bigger.
The design difference is named in the papers, and it is the whole explanation. The 2016 study ordered chain meals selected as the same entrées targeted in the non-chain restaurants — a matched, dish-for-dish comparison. The 2013 comparison was against the chains' popular meals, which is a comparison of what people order, not of what a kitchen puts in a given dish.
The two studies do not disagree about takeaway food. They disagree about what to compare it with — and once the dishes are matched, the gap closes.
The practical reading is precise. A chain's published figure for the dish you actually ordered is a reasonable starting estimate for the independent version of that dish. What the chain figure does not capture is that the dishes on offer at an independent kitchen skew larger and richer, so the meal you end up choosing runs heavier even when the dish-level numbers agree. Estimate the dish from a chain analogue; then correct for the fact that you didn't order a chain's dish. Whether the chain's own number deserves trust is a separate question, answered in how accurate restaurant calorie counts are.
A starting prior, and how to widen it#
| Measured quantity | Value | Source |
|---|---|---|
| Mean energy, independent/small-chain meal | 1,327 kcal (CI 1,248–1,406) | Urban 2013 |
| Same meals vs popular large-chain meals | +49% | Urban 2013 |
| Non-chain meals vs matched chain dishes | −68 kcal (P=0.41) | Urban 2016 |
| Median total fat per takeaway portion | 52.0 g (29.8–75.8) | Jaworowska 2025 |
| Mean takeaway portion mass | 580 g | Jaworowska 2025 |
| Popular delivery-platform items classed discretionary | 84–88% | Partridge 2020 |
Four moves, in the order they pay off:
Decide what fraction of the container you ate, first. At a mean 580 g, a takeaway box is routinely two portions of anything you would serve yourself. Halving a 1,300-calorie meal is a bigger correction than any refinement of the dish estimate, and it is the only input you can observe directly rather than infer.
Log the fat carrier as its own line. Sauce, oil and dressing are where the variance lives; giving them a separate entry means a later correction touches one number instead of forcing you to re-estimate the meal. The same logic applies to the free sides that arrive uninvited — a recurring theme in hidden calories stalling weight loss, and the reason the calories in cooking oil matter more per gram than anything else in the box.
Anchor on a chain analogue rather than a database dish. For the entrée itself, a published chain figure for the same dish is defensible: matched chain and non-chain dishes measured within 68 calories of each other2.
Record it as a band, and make the band wide. A single dish measured at one outlet is not the same dish at another, and the fat range alone spans 46 grams — over 400 calories — across the London sample. An entry of "900 to 1,400" carries information that "1,150" throws away. If takeaway is a regular fixture rather than an occasional one, the structural fixes are in tracking calories when you eat out often.
FAQ#
How many calories are in a typical takeaway meal?#
Measured rather than estimated: 157 meals from independent and small-chain restaurants averaged 1,327 kcal, about 66% of a typical day's energy requirement, and 7.6% of them exceeded a full day's requirement alone1. Treat 1,300 as a starting point for a full main-plus-sides order, not as a worst case.
Is a delivery order different from eating the same meal in the restaurant?#
The food is generally the same; the information is not. Menu labeling law covers chains of twenty or more locations and says nothing about third-party delivery platforms5, and a survey of 1,074 popular delivery outlets found nutrient profiles unavailable for most popular items4. Ordering in also removes the plate as a size reference, which is why the container becomes the unit.
Should I log a takeaway as one entry or split it into items?#
Split it, but not into many. Two or three lines — the entrée, the starch, and the fat carrier — capture nearly all the variance, and separating the fat means the number you will most likely need to correct is isolated. Splitting a curry into eleven ingredients adds precision the underlying evidence cannot support.
Sources#
- Urban LE, Lichtenstein AH, Gary CE, et al. The energy content of restaurant foods without stated calorie information. JAMA Intern Med. 2013;173(14):1292-9.
- Urban LE, Weber JL, Heyman MB, et al. Energy contents of frequently ordered restaurant meals and comparison with human energy requirements and USDA database information: a multisite randomized study. J Acad Nutr Diet. 2016;116(4):590-598.e6.
- Jaworowska A, Force S. Total Fat and Fatty Acid Content in Meals Served by Independent Takeaway Outlets Participating in the Healthier Catering Commitment Initiative in London, UK. Int J Environ Res Public Health. 2025;22(1):121.
- Partridge SR, Gibson AA, Roy R, et al. Junk Food on Demand: A Cross-Sectional Analysis of the Nutritional Quality of Popular Online Food Delivery Outlets in Australia and New Zealand. Nutrients. 2020;12(10):3107.
- 21 CFR 101.11 — Nutrition labeling of standard menu items in covered establishments. US Code of Federal Regulations (govinfo).
