How are different meal ratios calculated for flights?

I can’t tell you what happens on an airline, but I was told by the head chef of a large cruise ship that he has on his computer how many of each meal have been served for the last three years, nicely separated by nationality, which lets him make very reliable forecasts. In case of a cruise ships things are even more complicated because behaviour changes throughout a cruise (people eat a lot more on the first day, and less on the last day).

As others have said, they use the average from many previous flights to work out the ratio. With large numbers (lets say 525 for an A380), you can tell that the actual number will come pretty close to the average.

I ran 2000 simulations in Excel, modelling a flight where they know, on average, 70% choose option A, 30% choose Option B on a 525 seat A380. The flight elects to take 404 of option A (10% more than they need on average) and 173 of option B (again, 10% spare).

The flight never ran out of option A in any of the 2000 ‘flights’. The flight ran out of option B 140 times (7%).

They can do better, by having a smaller % buffer on the more popular meal, due to how the law of large numbers works. Taking 15% spares on the less popular meal, and 7.8% spares on the more popular meal (which means taking 10% spare meals in total) reduces the number of times that the airline ran out of option B to 22 (1.1%), whilst still never running out of option A. There is likely further optimisation that could occur, 15% and 7.8% was the only number I tried other than the 10% for each.

This link here implies that the cabin crew (but presumably not the flight crew who have to eat different meals to each other) are made to eat whatever is not chosen by the passengers. On an A380, this is a 27 buffer out of 525 (just over 5%), so this is the minimum possible buffer an airline could use. But presumably they would at the very least have a few extra meals so that no-one goes hungry if one meal is dropped.

Assuming they bring 5 ‘spare meals’, plus 27 for the flight crew and the flight is at 80% occupancy, that puts the buffer on meals at 32/420. However, those 420 meals for customers won’t all be normal meals, but that 27 buffer is just for normal meals. Vegetarianism in the USA is 3.2%, and when you add in Gluten Free, Dairy Free and Kosher, we could easily reach 5%. That least us with a buffer of 32 out of 399, or 8%. With only 5 meals at risk of going in the bin each flight.

Results will be worse on smaller planes, which may be where conflicting anecdotal evidence from domestic short haul flights comes from.

The problem with topics like this is always that this kind of information has limited scientific use but a lot of commercial value. There will have been hundreds if not thousands of internal studies done on this since the beginning of commercial flight but the results of those will all be classified as trade secrets. That information is only intended for use by the airline that paid for it, compiled its own data and ran its own market tests. The formulas involved are meant to balance customer satisfaction against operational costs and there will be dozens of factors involved: origin, destination, passenger demographics, date/time/duration of the flight, what options are available and so on. Beyond just determining the ratio, the airline also has to decide how many options to offer which adds another layer of complexity.

For established routes the major driver will simply be historical data. After a couple of thousand flights you can fairly accurately predict the ratios and even determine the impact of certain events and holidays ahead of time. Post-Thanksgiving guilt or fresh New Year’s resolutions might drive up the vegetarian ratio for instance.

While it may seem a bit melodramatic to talk about this level of secrecy for something so simple, you have to remember that the airline industry is all about margins and lowering costs to remain competitive. Until an airline has sufficient information about a particular flight’s meal patterns it will be faced with either higher costs and wasted food, or unhappier customers.

I was recently on Philippine Airlines from Manila to Sydney. We were down the back and suspected they’d run out of what looked to be the most popular. Shrug, we took the second option. Man behind us – however, was fuming (he’d complained about everything already – the seat, the wine, the people around him) and had a good rant to the steward about it.

She explained the ratio was set at 70/30 for the two dishes, and unfortunately sometimes it just doesn’t work out if more people get it, obviously.

Anyway, the point was that for PR at least, it’s not 50/50 as someone in the comments alluded to, but I’m sure at least some airlines would go for a different ratio, especially if they have more than two options, as airlines like Emirates sometimes do.

Same way any restaurant catering is done: first you guess, then you iterate. Airlines frequently repeat the same meals over and over again, so it doesn’t take very long to get to a reasonably good balance, and of course the big carriers worked out their beef vs chicken ratios years ago.

Update: An interesting story about using AI to optimize this: https://skift.com/2018/11/05/airlines-hope-algorithms-can-finally-fix-their-drink-carts/