Restaurant KPIs: the six numbers to check every week

Short answer: you can steer a food business on six weekly numbers: sales against the same week last year, prime cost (food plus labour, target 55–65% of sales), average order value, table turnover (or orders per hour for counter service), labour percentage by daypart, and one customer number such as new-versus-returning or review rating. Put them on one page every Monday with their four-week trend, attach one decision to whichever moved, and skip the forty-tile dashboard entirely. Here is each number: the formula, the benchmark, and what to do when it drifts.

Why six numbers beat forty

Data does not run restaurants; decisions do. A page you actually read every week, where each number has a healthy range and a known lever, produces decisions. A dashboard with everything produces scrolling. The six below are chosen because together they cover demand (sales), efficiency (prime cost, labour), yield (AOV, turnover) and durability (customers), and because each one maps to a guide on this blog when it needs surgery.

The six, at a glance

1. Sales vs the same week last year

Raw weekly sales lie, weather, holidays and seasons swing them. The same week last year is the honest comparison. Two habits make it useful: split it by channel (dine-in, direct online, delivery apps, so you can see the profitable direct share growing) and by daypart, because “down 4%” usually turns out to be “Tuesday dinner down 30%”, which is a fixable, specific problem. If growth is the issue, the levers live in local SEO and the customer playbook.

2. Prime cost, the profit predictor

Food plus labour as a share of sales, the single best weekly health check. Inside 55–65% with sane rent, you are almost certainly profitable; above 65% for a few weeks, margin is leaking regardless of how busy it feels. The full weekly ritual, worked example and fix-lists live in the prime cost guide, with the food half detailed in food cost.

3. Average order value

Sales divided by orders, tracked per channel. Digital orders typically run 15–30% larger than counter orders because the screen never forgets to offer the add-on, so a rising direct-online share should pull AOV up with it. When AOV stalls, the levers are combos, one-tap add-ons and menu anchoring, all covered in the AOV guide. A $2 lift on 800 weekly orders is $83,000 a year; this number earns its slot.

4. Table turnover (or orders per hour)

Parties divided by tables per service for dine-in; orders per open hour for counter service. Full-service healthy range is roughly 1.5–3 turns depending on concept. The improvement order, kill the payment wait, then the ordering wait, then kitchen pacing, is in the turnover guide. For the ambitious: RevPASH (revenue ÷ seats ÷ hours) merges this number with AOV into one measure of how hard the room works.

5. Labour percentage, by daypart

Labour cost over sales, viewed per daypart rather than as one weekly blob. The overall band is typically 25–35%, but the daypart view is where the money is: a 45% Tuesday lunch and a 24% Friday dinner average out to “fine” while both are wrong. The fix is schedule shape, staffing to forecast demand, covered in scheduling, and chronic overtime is usually a hiring problem wearing a rostering costume.

6. One customer number

Pick the one that matches your strategy: new-versus-returning share if you are building regulars (returning customers are the cheap revenue, retention beats acquisition roughly five-to-one), or review rating and volume if reputation is the front. Watch it weekly the same as the money numbers, because it is the leading indicator the others lag: falling returning share this month is falling sales in three.

The Monday page (steal this)

One page, fifteen minutes, every Monday: the six numbers, each with last week, the four-week trend, and its healthy range. Circle the one that moved most. Write one sentence: what we will do about it this week. That sentence, “trim Tuesday close by one”, “push the direct-order flyer in every delivery bag”, is the entire point of measurement. A number without a decision attached is decoration.

Mistakes that turn KPIs into wallpaper

• Tracking twenty things monthly instead of six things weekly.
• Averaging away the signal, whole-week labour hides the daypart that is bleeding.
• Comparing against other people’s benchmarks instead of your own trend and band.
• Reading the page but never attaching a decision, measurement as ritual instead of steering.
• Ignoring the customer number because it is not in dollars, it is the earliest warning you get.

Frequently asked questions

What KPIs should a restaurant track weekly?
Six carry most of the signal: sales vs the same week last year, prime cost (food plus labour as % of sales, target 55–65%), average order value, table turnover or orders per hour, labour as % of sales by daypart, and one customer-side number (new vs returning customers, or review rating). Everything else is drill-down when one of those six moves.

What is a good average order value for a restaurant?
There is no universal number, a café’s $14 and a bistro’s $62 can both be healthy. What matters is your own trend and your channel mix: digital and QR orders typically run 15–30% larger than counter orders, so AOV rising as digital share grows is the system working. Benchmark against your own last eight weeks, not someone else’s concept.

How do I know if my restaurant is profitable before the accountant tells me?
Watch prime cost weekly. If food plus labour stays inside 55–65% of sales and your occupancy cost (rent) is under about 10% of sales, profit is very likely; if prime cost runs over 65% for a few weeks, you are burning margin no matter what the bank balance says today. The weekly prime-cost ritual is the closest thing to a live P&L an independent can run.

What is RevPASH?
Revenue per available seat hour: total revenue ÷ (seats × hours open). It is the honest measure of how hard your room works, because it combines spend and speed in one number. Two venues with identical sales can have very different RevPASH if one does it from half the seats. Track it by daypart to find the hours where the room is loafing.

How many numbers should I actually look at?
Six weekly, on one page, plus a monthly deep-dive. The failure mode is not too little data, it is dashboards with forty tiles nobody reads. One page, same six numbers, every Monday, each with its four-week trend and one decision attached (“labour crept to 34%, trim Tuesday close”), beats any amount of reporting theatre.

Do I need software for restaurant KPIs?
You need your POS data in one place; whether a spreadsheet or a dashboard assembles it is taste. The test is friction: if producing the weekly page takes more than fifteen minutes, it will quietly stop happening by August. Platforms that already hold your sales, labour and menu data can render the same page in a glance, which is the entire argument for them.