Average check is the most quoted number in a restaurant and one of the least informative, because a single figure is standing in for a distribution that is almost never shaped like a single figure.
It is not useless. It is fine for forecasting, fine for comparing this September to last September, and fine for a conversation with a landlord. It is bad at the job people mostly give it, which is describing guests.
What the number actually is
Total sales divided by something, and the something is where most of the disagreement starts.
Per check is sales divided by closed checks. Per cover is sales divided by guests seated. A four-top on one bill produces one check and four covers, and for that table the two numbers differ by a factor of four.
Both are legitimate. Mixing them across locations, or across years, produces a trend line that is a bookkeeping artefact rather than a fact about your business.
Average check = total sales ÷ number of closed checks Average cover = total sales ÷ number of covers A Tuesday dinner service: 60 covers · 22 checks · sales of 2,640 average check = 2,640 ÷ 22 = 120 average cover = 2,640 ÷ 60 = 44 Neither is wrong. Reporting one of them as the other is.
The bimodal room
Take a dining room with a twelve-seat bar doing snacks and a room doing a set menu. The bar spends one amount, the room spends another, and there is essentially nobody in between.
Your average lands precisely in that gap. It is the arithmetic mean of two populations and a description of neither, and every decision taken from it — the wine list, the staffing, the price of the set menu — is aimed at a guest who does not exist.
This is not an exotic case. Any restaurant with a bar, a lunch service, or a large-party business is at least bimodal, and most are messier than that.
Look at the distribution instead
Twenty minutes in a spreadsheet, and it is the single most useful thing in this article. You are not looking for a number. You are looking for a shape.
Export every closed check for a quarter. One column: the check total. Then bucket into bands and count. 0- 20 #### 20- 40 ############ 40- 60 ### 60- 80 ## 80-100 ########## 100-120 ##### Read the shape, not the mean: One hump the average means something. Two humps the average means nothing. Name the two groups and manage them separately. Long tail the average is being dragged by a small number of large parties, and it will move whenever they do.
Median, and the top band
Two numbers to keep beside the average, both cheap to produce once the checks are in a column.
The median is the check in the middle when you line them all up. It ignores the enormous December party that made last quarter look like a strategy, which is exactly what you want when you are deciding how to price a menu.
The top band is the share of your sales coming from your largest checks. If a small number of big parties carry the quarter, that is concentration risk — and it is completely invisible in an average, right up until the corporate account that booked all of them changes offices.
Average check versus what a guest is worth
The most expensive mistake this number causes is ranking guests by it.
A guest who comes twice a year and orders generously has a high average check. The one who eats at the bar every second Tuesday has a low one and is worth considerably more across the year. Rank by average check and you will build a VIP list composed largely of people who barely come.
The number that answers "who matters" is per-guest annual spend: everything they spent in twelve months, per identified guest. It requires guest identification to be working, which is a separate project — but it is the only version of this that has ever been worth acting on.
Cuts worth making
The same figure split four or five ways will usually tell you something the group number never could.
- By daypart. Lunch and dinner are two different businesses being reported as one.
- By party size. Spend per cover normally falls as party size rises. Knowing your own curve tells you what a large booking is genuinely worth before you hold the whole back room for it.
- By first visit versus repeat. If repeat guests spend less per visit, that is not a failure. It is what a regular looks like. Check the annual figure before anybody panics.
- By location — and resist averaging locations into a group figure. That is an average of averages, and by then it describes nobody at all.
- By booking channel. Walk-ins, platform bookings, and direct bookings rarely spend alike, though the difference is usually about who those guests are rather than how they arrived.