The Restaurant Labor Cost Benchmark Everyone Quotes Is Wrong (Here’s the Real One)

By The Pragmatic CFO. Last verified: 2026-09-03.

Someone told you the industry average for restaurant labor is around 30% and you nodded. That number is doing more damage than good. It lumps a Subway franchisee in with a white-tablecloth steakhouse, it is almost always quoted wages-only, and it comes with no comparison rules. Running your P&L against it, you are either patting yourself on the back for underperforming or losing sleep over a number that is fine for your concept.

Below: what labor % actually measures, why the “20% to 35%” range is misleading, how to compare your number to a public chain honestly, what twelve public operators disclosed for Q2 2026, and three tactical plays for closing a gap.

What “labor cost” actually means on your P&L

Labor cost as a percentage of net sales sounds like one number. It is really two.

Wages only. Hourly and salaried wages paid to restaurant employees, before employer costs. This is what an operator sees in the payroll register on Monday morning.

Fully loaded labor. Wages plus employer payroll taxes (FICA, FUTA, SUTA), workers’ comp premiums, group health and other benefits, PTO accrual, and formal training cost (recruiter fees, background checks, onboarding hours, uniforms). This is the number that belongs on your income statement.

The gap is not small. Depending on state (California SUTA and workers’ comp bite harder than Texas), benefits mix, and role mix, fully loaded labor runs 15% to 25% above wages only. That is the difference between quoting yourself a 27% labor number and reporting a 33% number to your CPA.

Compare your fully loaded number to a peer’s wages-only number and you will always look worse. That is not a performance problem. It is an accounting apples-and-oranges problem.

Why the “20% to 35%” range is misleading

Two reasons.

First, it collapses concept types that have structurally different labor models. A drive-thru QSR runs on eight to twelve labor hours per hundred transactions with mostly minimum-wage crew. A full-service polished-casual restaurant runs servers, bartenders, hosts, bussers, line cooks, prep cooks, dishwashers, and a manager on duty for the same hundred covers. They are not on the same benchmark.

Second, most quoted numbers are wages only, and the source is either an anonymous survey or a trade-press article that never defined its terms.

Here is what the concept-specific bands actually look like, on a fully loaded basis (wages + payroll taxes + workers’ comp + benefits + PTO + training):

• QSR: 25% to 30% healthy. Above 32% is a red flag.
• Fast casual: 25% to 30% healthy. Above 32% is a red flag.
• Full-service (casual, polished casual): 28% to 32% healthy. Above 35% is a red flag.
• Fine dining: 30% to 34% healthy. Above 36% is a red flag.

These bands assume tip-credit states and normalized benefits. Operate in California, Washington, New York, D.C., or one of the other high-wage jurisdictions, and add 200 to 400 basis points to each. Skip employer-sponsored health for hourlies, subtract 100 to 200 basis points.

Anything outside those bands means the benchmark is asking you a question. Not necessarily failing you.

How to compare your labor cost apples-to-apples

Four rules. Skip any of them and your peer comparison is decorative.

1. Strip the tip credit or stay within one regime. In tip-credit states the reported wage is the direct hourly the operator paid; tip income fills the gap to minimum. In no-tip-credit states (California, Alaska, Montana, Nevada, Oregon, Washington, Minnesota, Guam), the operator pays full minimum out of pocket. Comparing raw labor % across the two regimes without adjustment is nonsense.

2. Normalize to fully loaded on both sides. Add employer payroll taxes (usually 8% to 10% of gross wages), workers’ comp (California quick-service is 4% to 8% of payroll; Texas is often under 2%), benefits, PTO, and training. Do this to your number and to the peer’s. If the peer disclosure omits these, adjust the peer up rather than adjusting yourself down.

3. Separate FOH from BOH. A concept at 30% total might be 12% FOH / 18% BOH or the reverse. Different operating profiles, different fixes. Know which side is bleeding before you cut.

4. Use SPLH, not just labor %. Sales per labor hour (SPLH) is the operator’s real diagnostic. Labor % is a ratio: change the numerator (wage rate) or the denominator (check average, traffic) and the ratio moves without your labor model doing anything. SPLH tells you how much revenue your labor hours generated. Two operators with identical 29% labor can have SPLH of $58 and $92 and be running totally different businesses.

Worked example. Two fast casual operators, both reporting 29% fully loaded labor.

• Operator A: $85,000 weekly sales, 1,466 total labor hours, SPLH of $58. Fully loaded blended labor rate of $16.80. Labor % is $24,650 divided by $85,000, or 29%.
• Operator B: $105,000 weekly sales, 1,140 total labor hours, SPLH of $92. Fully loaded blended labor rate of $26.70. Labor % is $30,450 divided by $105,000, or 29%.

Same labor %. Operator B pays 60% more per hour (probably a high-wage state) but is doing 60% more revenue per hour of labor deployed. Operator B has a productive labor model. Operator A has a scheduling and throughput problem hiding behind a low wage rate. Take Operator A to a $17 minimum wage and their labor % jumps toward 33% overnight. Operator B absorbs the same wage move without moving the ratio.

Labor % alone cannot tell you which of these two P&Ls is healthier. SPLH can.

Q2 2026 labor cost disclosures: twelve public chains

Public chains disclose labor two ways. Company-operated chains (Chipotle, Texas Roadhouse, Shake Shack, Cheesecake Factory, Darden, Dutch Bros, Wingstop for its company stores, CAVA) report “labor costs” or “labor and related expenses” as a percentage of company-operated restaurant sales. Heavily franchised chains (McDonald’s, Domino’s, Yum, Restaurant Brands International) do not disclose a system-level labor number in a comparable way, because franchisee P&Ls do not flow through the parent 10-Q.

Below is what each chain disclosed for its most recently reported quarter (calendar Q2 2026 for most, fiscal periods noted per row where different).

A few reads on the table:

• The company-operated chains that report a clean labor % (CMG, TXRH, WING, SHAK, DRI, BROS) all sit inside the concept-specific ranges from earlier in this article. Fast casual clusters at 23% to 25%. Full-service casual sits at 31.7% to 32.5%. Coffee QSR (Dutch Bros) at 25.4%.
• Texas Roadhouse at 32.5% is at the upper edge of the healthy full-service band. Management commentary attributed the year-over-year decline of 40 basis points to a higher check average and productivity offsetting wage inflation of 3.9%. That is what a competently managed labor P&L looks like right now.
• Chipotle at 25.0% is at the low end of the fast casual band. That is a function of high average unit volumes (a Chipotle box does more revenue than almost any peer), not paying less. Spreading fixed labor over higher sales is the single biggest driver of a good fast casual labor %.
• Wingstop at 23.3% reflects a limited menu, minimal dine-in, and a small BOH footprint. If your concept is not built that way, do not expect to hit that number.

For MCD, DPZ, YUM, and QSR, do not read the parent P&L as a labor benchmark. The parent is a royalty and supply-chain business. The franchisee is the labor operator and does not publish.

Three plays for closing a labor gap

If your fully loaded labor is 100 to 300 basis points above your concept’s healthy band, here is the sequence that has worked for me.

Play 1: schedule to sales, not to habit

Most stores are still built around a Monday-through-Sunday shift template that has not been rebuilt in two years. Sales have moved. The template has not.

Pull four weeks of sales and labor by 15-minute increment. Overlay them. Look for two patterns:

• Blocks where you have three people on the clock and the store is doing $80 an hour. That is a scheduling error.
• Blocks where you have two people on the clock and the store is doing $340 an hour. That is a service-time problem quietly costing you check average.

Rebuild the template around the sales curve. Cut the overstaffed blocks first (they are the easier operator conversation), then add coverage to the underserved blocks. Expected recovery: 0.5 to 2.0 points of labor % in 30 to 60 days if you had real drift.

Cadence: run this exercise monthly. Sales curves shift with seasons, weather, school calendars, and marketing.

Play 2: attack turnover-driven training cost

Every replacement hire is fully loaded cost with zero productive labor for the training period. Published estimates for restaurants run about $2,706 for hourlies and up to $17,651 for general managers (Dailypay restaurant turnover center). Even at the low end, a store doing 30 hourly separations a year is burning about $80,000 on cost per hire alone, before you count the productivity loss during the ramp.

The tactic: measure 90-day retention, not annual turnover. 90-day retention is the leading indicator. If more than 40% of new hires walk before day 90, your onboarding is broken. Fix scheduling predictability for new hires (four-week posted schedules with no last-minute changes), assign a named trainer per new hire, and pay a retention bonus at day 60 or 90.

A 10-point reduction in annual turnover on a 30-employee store, at $2,700 per replacement, is about $8,100 back to your P&L per year. On a $1.5M store that is roughly 0.5 points of labor %. That drops straight to 4-wall EBITDA.

Play 3: reprice or repackage the items whose labor cost per sale is above concept average

You already do this on food cost. You almost certainly do not do it on labor.

Every menu item has a labor cost. It is not on your menu-mix report by default, but you can build it. Take the total labor minutes to prep, cook, plate, and clean each item. Multiply by fully loaded labor cost per minute (blended rate divided by 60). That is the labor cost of a sale of that item.

Now express it as contribution margin per labor minute. Item price minus food cost minus labor cost divided by labor minutes to produce.

You will discover that two or three items on your menu are contribution-margin-per-labor-minute dogs. They sell fine, they food-cost fine, and they eat your line cooks. Reprice them, simplify them, or 86 them.

Worked example: a full-service concept with a hand-formed lasagna at $22.95. Food cost 26% ($5.96). Labor minutes: 14. At a $28 fully loaded blended kitchen rate, that is $6.53 of labor per plate. Contribution margin: $22.95 – $5.96 – $6.53 = $10.46. Per labor minute: $0.75.

Compare to the pan-seared chicken at $19.95 that food-costs at 28% ($5.59), takes 6 labor minutes, and delivers $11.56 contribution margin per plate, or $1.93 per labor minute.

The lasagna is a labor sink. Either raise the price to $26.95, reduce the portion, or replace it with a preportioned pasta bake that hits eight labor minutes instead of fourteen. Expected recovery from repricing or re-engineering three or four items like this: 0.5 to 1.5 points of labor % once you get through a full menu-mix cycle.

Methodology footer

“Labor %” in this article means labor costs (however each chain reports the line) divided by the applicable revenue denominator (company-operated restaurant sales for company-operated chains; total revenue for Chipotle, which is how Chipotle discloses). Where a chain does not disclose a comparable single-line labor number, the cell reads “n/d.”

Chain figures were pulled from Q2 2026 SEC filings (10-Q or 8-K earnings release) or, for Darden which reports on a May fiscal year, from the FY26 10-K covering the full year ended 5/31/26. Starbucks reports on a September fiscal year and the cited quarter is Q3 FY26 (13 weeks ended 6/28/26).

Concept-specific labor bands referenced in this article match the RBL master benchmark table published 2026-08-22.

Last verified: 2026-09-03.