The Catering Tray a 98-Year-Old Deli Loses to Voicemail Every Week

Carshon's Deli has operated in Fort Worth since 1928 — ninety-eight years of dine-in and takeout, still family-run, still making everything in-house. Its digital order stack, as visible from outside, is one link: a DoorDash button on the website. Catering requires a phone call during service hours and 48 hours of lead time. That is the entire ordering infrastructure for a business old enough to have outlived four generations of point-of-sale systems.

//What a single delivery link actually hides

A DoorDash button looks like infrastructure. It behaves more like renting a storefront on someone else's street. Third-party delivery platforms typically charge restaurants a commission of 15% to 30% per order, plus separate service fees of roughly 2% to 4% (CloudKitchens). The range is wide because the rate scales with which service tier a restaurant selects and how much in-app placement it buys — but even at the low end, a deli routing a meaningful share of volume through the app is giving up close to a fifth of that revenue before rent, payroll, or ingredients are counted.

The commission is the visible cost. The one that compounds is who holds the customer relationship. The order, the phone number, the reorder pattern — all of it lives in the platform's database, not the deli's. A regular who orders through the app six times a year is a platform customer with a Carshon's habit, not a Carshon's customer who happens to use the app. There's no owned channel to text that customer when a favorite runs out early, or that catering trays are discounted on a slow Tuesday.

//Where the hours actually go

The second gap is catering, and it's a scheduling problem before it's a marketing one. A 48-hour, phone-only booking process means every catering order gets written down by hand, typically during the lunch rush, by whoever answers. That's the same window the deli is busiest with walk-in traffic — exactly when a call is most likely to roll to voicemail and not get returned before the customer books elsewhere. Modeled against a conservative estimate of one missed catering tray a week at a $150 average ticket, that's roughly $7,800 a year in bookings that never made it past the answering machine — before counting the commission bled to the delivery app on the retail side.

None of that shows up on a P&L as a single line item. It shows up as flat catering revenue year over year, in a category where demand — office lunches, sports team orders, family gatherings — doesn't go flat on its own. It goes to whichever caterer picked up the phone.

//The reconciliation problem underneath both

Every third-party order also creates a manual accounting step: matching what the platform actually deposits against what the ticket said was owed, line by line, after fees. Done by hand, that task has no natural stopping point — it just accumulates through the week. It sits alongside two other recurring manual jobs common to independent full-service kitchens: building the staffing schedule from memory rather than a sales forecast, and calculating prep quantities against yesterday's guess instead of actual sell-through.

Restaurant labor cost is a useful benchmark here because it's one of the few figures the industry tracks consistently. Limited-service restaurant labor, wages plus benefits, ran a median of 31.7% of sales in 2024, up from roughly 28% historically; full-service ran 36.5%, up from about 33% (National Restaurant Association). A kitchen running meaningfully above its category's median, with no forecasting tool driving the schedule, is very often over-staffing against guesswork rather than against demand — the gap between a memory-built schedule and a forecast-built one usually shows up in points of labor cost, not fractions of one.

//What replacing the stack actually changes

The fix isn't dropping the delivery app — it still brings in orders that wouldn't otherwise exist. It's adding what's missing around it: a direct ordering and catering page the deli owns, where a booking takes under two minutes instead of a phone call during the rush; a daily reconciliation view that nets delivery-platform payouts against tickets automatically and flags mismatches instead of requiring someone to go find them; and a labor forecast built from actual sales history rather than a manager's memory of last month. None of that requires leaving the delivery platforms. It requires the business to also have a channel the platforms don't own.

For the customer, the visible change is small: a booking form that takes about ninety seconds instead of a callback that might not come. For the business, the change is that catering demand stops being capped by whoever happens to be free to answer the phone during the lunch rush.

//What we can't verify

The dollar figures modeled above — the $150 average catering ticket, the one-tray-a-week loss estimate, and the roughly $7,800 annual figure derived from it — come from a published sample diagnostic and are explicitly illustrative, not measured results pulled from Carshon's actual books. We have not audited the deli's real financials, and no real client data is used or implied. What is independently verifiable is the delivery-commission range (15% to 30%) and the restaurant labor benchmarks cited above. Everything specific to this business beyond what's publicly visible is a model, not an observation, and we're saying so rather than presenting it as one.