The Joys of Catering, Part 1: Tickets Sold Does Not Equal Catering Count
The first challenge with conference catering is figuring out how many people you are actually going to feed. While ticket sales are part of that equation, there are actually a number of moving parts that will give you a final number that is unlikely to match the number of tickets sold.
While there is a certain mathematical component to figuring headcount, there’s also an art that doesn’t fit quite so neatly into a formula. Once you have done this a few times, or if you work with the same conference and/or caterer year to year, you develop an intuitive sense about how to adjust the headcount to fit the audience.
I’m going to use nice, round numbers for this example to keep the math simple. Substitute your own numbers as needed!
Adding up:

For our hypothetical conference, we’ve sold 450 tickets. Go, us!

If your conference sells tickets on the day of the event, add in your historical rate. If you don’t know that number, use 2% of Tickets Sold.

Next, look at your ticket sales: does that number include conference staff/volunteers, the outside A/V team you’ve hired, or other people who are working the conference? If not, count them up, and add that number.

Some conferences do a “plusone” ticket, which allows the spouse/partner/significant other of an attendee to partake of meals and breaks. If that number isn’t included in Tickets Sold, add that in as well. If you do not offer such a ticket, you need to decide how to enforce the “attendees only” rule at breaks and meals. For the conferences I work, we require everyone to wear their badge when entering the food service area—pleasantly uncomplicated and efficient.

Ask the caterer what number they plan for. Most caterers buffer 2%  3% so that they don’t run out of food, but also don’t end up with piles of leftovers.

Lastly, is there anyone else who is going to show up and expect to be fed? Perhaps the CEO of your SuperPremiumTopTierBiggest Sponsor. Who “forgot” to register, and the company used all their tickets already. Are you really going to tell them they can’t have coffee? (The answer here is “no”, in case you were wondering.) This is where the caterer’s overage comes in handy. You should have only one or two of these people. If you have more than that, take a look at your registration and sponsorship policies for next year—something is definitely out of alignment.
Subtracting out:

Most conferences have a 1%  3% noshow rate. I’ve worked conferences where the noshow rate has been as high as 8%  10%, but those are outliers. If you know the historical noshow rate for your conference, use that number; otherwise, use 3%.

There are always a few people who don’t eat the conference food, no matter what is on the menu. They just don’t. If you have historical numbers, go with that. Otherwise, 1%  2% is a workable estimate.

If the venue has other food options, for example, the venue has a casual restaurant, or a great sushi bar, or a line of food trucks outside the front door, you’re going to lose people to those other options. Additionally, if there’s a treasured local eatery within two blocks of the venue, some people will go there for lunch. If you have any of these situations, take off another 3%.
Here’s our hypothetical conference:
Tickets Sold: 450 WalkIns: 10 Staff and Volunteers: 5 PlusOne Tickets (if not included in Tickets Sold): 20 Unexpected Guests: 2 Total In: 487
Noshows at 3%: 15 Not eating at 3% 15 Other options at 3%: 15 Total out: 45
Total In  Total Out = Initial Adjusted Headcount: 487  45 = 442
If you quote the caterer a headcount of 442, and they plan a 3% overage, they will cater for 455.
If you’re concerned about costs, you can make an additional headcount adjustment. Most caterers allot a generous amount of food per person, so you could adjust your total downward between 3% and 5% and come out alright. I’ve adjusted as much as 10% for some events, based on a decrease in ticket sales from the prior year.
Additional 3%: 442  13 = 429 headcount, caterer sets for 442, your original headcount
Additional 5%: 442  22 = 420 headcount, caterer sets for 432, 10 fewer than your original headcount, which is still a workable margin
Additional 10%: 442  44 = 398 headcount; caterer sets for 410. We’re getting edgy here; I would go this low only if attendees had a number of other options, because you may end up taking a few people out for pizza or something if the food runs out.
For a real world example, a conference I ran a few years ago had a total adjusted headcount of 450. The caterer set a 3% buffer, so a headcount of 450 would be set for 463.
The conference had not kept records, so we had no historical numbers to compare. Purely on intuition, I gave the caterer a headcount of 420, which meant they set for 433. The actual count through the lunch area each day varied from 421  428; no one went hungry, and the caterer ran well under their 3% waste goal. So dropping the headcount that additional 6% worked out almost exactly. I don’t have a solid reason why I chose the 420 as the final number; it just seemed right, and, as it turned out, it was!
Again, though it’s not a hard science, the numbers are a solid start. The Initial Adjusted headcount can also be your end point as a more cautious approach, rather than going with the additional adjustments. If you’re not sure where to start, use the formula above to derive the Initial Adjusted Headcount for the caterer, and then make adjustments each subsequent year, based on experience.