Book delivery at net and your food cost percentage is wrong

I've written here about reading a P&L and about which metrics matter. This post is about the layer underneath both: the single journal entry your POS posts every night. If that entry is wrong, every ratio built on top of it is wrong in ways that look completely plausible.

Start with the one that costs the most.

The net-booking error

A delivery platform deposits money in your account. It is not your sales figure.

Under the revenue recognition standard, the test is control. A restaurant that sets its own menu and prices, prepares the food, and is responsible for the quality of what arrives is generally the principal in the transaction — which means it records revenue at the gross menu price and the platform's commission as an operating expense.

Take a $10 item with a $3 commission:

Booking the deposit understates revenue by 30% and understates expense by 100%. And because sales is the denominator of nearly everything, it then corrupts food cost percentage, labor percentage, prime cost, and rent as a percentage of sales — all at once, all in the direction that makes them look better than they are.

For scale on the commission itself: New York City's cap, made permanent in 2021, allows 15% for delivery, 3% for credit card processing, and 5% for other services — with a 2026 settlement permitting an additional 20% for "enhanced services." That is the size of the gap between your gross sales and your deposit. I went through what those commissions do to unit economics in the third-party delivery post; this is the same money disappearing from your books instead of your margin.

One honest caveat: principal status is a conclusion from your actual contract, not a default. Most restaurants are the principal. Not all are. Have someone read the agreement rather than assuming.

The identical error happens with card processing fees. US merchants paid $198.25 billion in processing fees in 2025, and the average combined Visa/Mastercard credit interchange rate reached 2.36%, up from 2.02% in 2010. Recording only the net batch deposit removes somewhere around 1.5–3% of card revenue from both sides of your P&L and hides a cost line that more than nine in ten operators name as a top challenge.

What the nightly entry has to contain

Here is the shape of a correct daily sales journal entry posted from the POS close:

Four of those lines are where operators go wrong, and each has the same signature: the entry still balances, so nothing errors.

Tips are a liability, not revenue. Charged tips arrive inside the deposit, so an operator reconciling to the bank books the whole deposit as sales. Revenue is overstated by the tip total, prime cost percentage looks artificially good, and the money you owe your staff appears nowhere on the balance sheet.

Gift cards are a liability, not revenue. A card sold is a contract liability; revenue is recognized at redemption, where the card acts as a tender type rather than a sale. Recognizing it at purchase pulls revenue forward by up to a year and omits a real obligation.

Comps and discounts belong in contra-revenue, visible, not netted into sales. Netting them destroys the comp-volume signal — which, as the fraud data shows, is one of the few internal-control indicators a POS gives you for free.

Cash over/short is a control account, not a plug. Which brings up the most common way it gets destroyed.

How the integration actually breaks

Every failure below produces a balanced entry. That is why they survive for months.

Mapping drift. You add a menu category in the POS. It has no GL mapping. The integration either drops it — sales quietly vanish from the P&L — or dumps it into a default account, moving your food cost percentage for no operational reason. The tell is a period where sales fall but food cost percentage holds steady.

Unmapped tender types. House accounts, third-party gift cards, employee comp tenders, delivery-prepaid. Each unmapped tender becomes a daily out-of-balance that someone plugs to cash over/short. Within a month, the account that would have detected drawer theft is noise.

Reopened-day double posting. A manager reopens a business day to enter a late check and re-closes it. The integration posts the day twice. Sales double, sales tax payable doubles, tips payable doubles. It balances perfectly. You find out at the sales tax filing.

Multi-entity mapping. Two locations in one legal entity, or one location split across two entities for liquor licensing. The integration maps by location, the return is filed by entity, and inventory transfers between them are never eliminated. Consolidated cost of goods is wrong in both directions.

Three-way delivery reconciliation. Platform statements will never match your POS. Platform-funded promotions, customer refunds issued by the platform, and adjustment fees appear on the settlement report and never in the POS. The reconciliation has to run platform statement to POS to GL, not POS to bank.

The calendar question, and its real cost

Restaurants are weekend businesses. A calendar month has 28 to 31 days and either four or five of each weekday. Comparing a month with five Fridays to one with four mixes a calendar artifact with actual performance, and no amount of analysis separates them afterward.

The fixes are a 4-4-5 calendar — the retail standard, where fiscal 2026 runs 1 February 2026 through 30 January 2027 — or a 13-period calendar of 28 days each, where every period contains exactly four of each weekday.

Now the part the advocacy usually omits: 13-period accounting de-synchronizes you from everything else. Rent, insurance, loan payments, sales tax filing periods, and most vendor statements are calendar-monthly. Somebody reconciles that boundary every single period. For a single location with one bookkeeper, that burden can exceed the analytical benefit.

The honest recommendation is conditional: adopt it when you have multiple units, a weekly ordering and scheduling rhythm, and someone whose actual job is the weekly number. Otherwise close weekly on a calendar month and accept the comparability noise.

Either way, remember the 53rd week. Since 52 times 7 is 364, a four-week-based calendar adds a 53rd week roughly every five or six years — most recently fiscal 2023. Year-over-year comparisons and any covenant tied to annual sales break in that year unless someone normalizes them.

About that prime cost benchmark

You will read everywhere that prime cost should be 55–65% of sales. I went looking for the study behind it and there isn't one I can reach. The number traces to a paid membership site whose survey methodology isn't public, and from there to consultant and vendor blogs citing each other. It may well be roughly right. It is not a measured benchmark, and nobody quoting it says so.

What is worth stating precisely is the definition, because this is where self-calculated numbers go wrong: prime cost is total cost of goods sold — food and beverage — plus total labor including payroll taxes and benefits, as a percentage of total sales. Leave out the taxes and benefits and you understate it by several points, which is exactly the error that makes an operator think they are inside a benchmark they aren't.

Also worth knowing: the Uniform System of Accounts for Restaurants, the industry-standard chart of accounts framework, was last revised in 2012. It predates third-party delivery, the current revenue recognition standard, and the present tip-reporting regime. It is still a good skeleton. It is not current guidance.

Tips: three separate tax consequences most operators miss

A tip is not a service charge. The four-factor test: the payment is made free from compulsion, the customer has unrestricted discretion over the amount, the amount isn't dictated by employer policy, and the customer generally determines who gets it. Fail any of them and it is a service charge, which is wages — payroll-taxed as wages, and ineligible for the FICA tip credit. I covered the operational side of this in the tip pooling post; the accounting consequence is that an auto-gratuity moved onto the menu changes three things at once and the menu change doesn't announce any of them.

The FICA tip credit is computed against $5.15 an hour. That is the federal minimum wage as it stood on 1 January 2007, frozen into the statute ever since — nineteen years. The credit is 7.65% of employer FICA on tips above the amount needed to bring wages to that figure, claimed on Form 8846. For many independents it is the single largest tax benefit available, and it requires that tips were properly reported. If your tip reporting is sloppy, the credit is unclaimable.

The 2026 W-2 changes. The tip deduction enacted in 2025 is capped at $25,000, phases out above $150,000 of modified AGI ($300,000 joint), runs through 2028, and reduces income tax only — Social Security and Medicare still apply. Final regulations issued in April 2026 list the qualifying occupations and confirm that service charges and mandatory auto-gratuities do not qualify unless the customer can disregard or modify them. Mechanically: beginning with amounts earned in 2026, qualified tips go in W-2 Box 12 with code "TP", and the employee's Treasury Tipped Occupation Code goes in new Box 14b. That was not required for 2025, and it is a data field most restaurants have never populated.

Two smaller things worth getting right

You are almost certainly not required to use accrual accounting. The gross receipts threshold for the cash method is $32,000,000 for tax years beginning in 2026. Essentially no independent restaurant or small group is compelled into accrual by federal tax law. Adopt it because it matches cost of goods to the sales that generated them — a real management benefit — not because someone told you it was mandatory.

Gift card breakage is not all yours. Unredeemed balances get recognized in proportion to the pattern of redemption, but amounts you are required to remit to a state as unclaimed property stay a liability permanently and can never become revenue. Roughly half the states escheat unredeemed balances after a dormancy period of three to five years. I wrote about gift cards as a revenue channel; this is the part that isn't revenue at all.

What it costs to do properly

I'm not going to quote you a monthly outsourced bookkeeping price. Every range I found came from firms selling the service, quoting wildly different scopes — some transaction coding only, some including sales tax filing, tip credit computation, and weekly prime cost reporting.

The defensible anchor is a wage: the median for bookkeeping, accounting and auditing clerks was $49,210 in May 2024, with the tenth percentile under $34,600 and the ninetieth above $72,660. That is the build-versus-buy comparison, and it is a measured number rather than a sales quote.

Context for whether it's worth it: the National Restaurant Association's 2026 industry report found 42% of operators were not profitable in 2025, against $1.55 trillion in projected industry sales. In that environment, a food cost percentage that is wrong because delivery posts at net is not a bookkeeping detail. It is the difference between knowing and guessing.

Disclosure: I work at Katalyst, and POS-to-accounting integration is something we build. So take the vendor-shaped advice with appropriate suspicion — and note that the highest-value item in this post is a check anyone can run tonight on any system: open last month's P&L, find your delivery sales, and see whether the number matches your menu prices or your deposits. If it matches the deposits, everything below it on that statement is wrong.

Related Katalyst products

• See Katalyst analytics
• See Katalyst Flex POS