Why off-premise catering margins are shrinking (and three things operators can actually do)

Operators have been saying this since mid-2023. The margin on off-premise catering, particularly corporate drop-off, has been getting tighter. Our read from working with off-premise operators is that gross margin on a standard drop-off that was 38% in 2019 is closer to 28 to 31% in 2026, and on some accounts worse.

This is not a single-cause problem. Here is the anatomy.

What actually changed

Food cost inflation stuck at elevated levels

The USDA Food Price Outlook shows food-at-home and food-away-from-home prices rising roughly 25 to 28% cumulative from 2019 through early 2026. Catering food cost input has tracked similar, with protein and dairy running hotter than the overall index.

The part of this that hurts off-premise specifically: the menu items that sold well in 2019 at $16 per head now cost the operator $8.20 to $9.10 to produce, versus $6.50 in 2019. Unless the menu priced moved to $19 per head in proportion, the per-plate margin compressed by 200 to 250 basis points. Most operators held pricing in 2020 to 2022 to keep volume, and are still catching up.

Labor that stays expensive even as hiring normalizes

The labor shortage of 2021 to 2023 drove fully-loaded hourly costs for catering kitchen and FOH labor from roughly $24 an hour in 2019 to $31 to $34 an hour in 2024. That has not come back down meaningfully even as staff shortages eased. Health insurance costs rose in parallel, and most operators who added benefits in 2022 to retain staff cannot remove them without losing the people.

On a standard 40-person drop-off that ran with one assembler and one driver in 2019, the labor cost moved from about $95 to about $140. That 45-dollar delta per order compounds hard at volume.

Fuel, at a different elevated normal

Fuel is cheaper than it was at the 2022 peak but has settled meaningfully above 2019 levels. The EIA gasoline history shows US retail regular at roughly $3.40 in early 2026 versus $2.60 in 2019. For an operation running six delivery routes a day with typical mileage, that is roughly $1,100 a month of extra fuel spend baked into the P&L.

Driver shortage as a structural tax

The CDL driver shortage got the headlines, but the non-CDL delivery driver pool is also tight and expensive. Operators running off-premise routinely pay $24 to $28 an hour for drivers who in 2019 were $18 to $20. Some of this is minimum-wage floor increases in specific markets. Some of it is simply the delivery labor pool got more competitive with gig work.

Corporate dining patterns shifted

This is the structural change nobody warned operators about. Corporate catering volume in 2019 was concentrated in large all-hands lunches and quarterly events. Post-2022, hybrid work broke the pattern. Corporate catering volume is up roughly 18% on a unit basis but the average order size is smaller and the days are more distributed. You are doing more Wednesday-specific 12-person lunches and fewer Tuesday 80-person all-hands.

The kitchen math on a 12-person order is dramatically worse than on an 80-person order. Same assembler labor time on the small one as a percentage of revenue, same route time for the driver. A 2026 off-premise operator is doing roughly the same top-line revenue as 2019 with 30% more orders to produce it.

Three things operators can actually do

Not "raise prices." Everyone is raising prices. The structural question is what to do on the cost side, because price increases have a ceiling that is being tested right now.

Play 1: Per-plate margin enforcement, not per-plate pricing

Most operators price by target margin at proposal time and then lose 200 to 400 basis points on execution. The difference between what was sold and what was run is the real problem.

Specifically: a sales team quotes a menu at a contribution-margin floor of 42%. The proposal gets approved. Then:

• The menu substitution 48 hours out uses a more expensive protein than the quoted one, and the substitution is made without a margin recheck. 150 basis points gone.
• Labor on-site runs 20% over what was quoted because the layout required more servers. 200 basis points gone.
• The client requests an add-on late, priced at cost-plus-10% because the coordinator wanted to be accommodating. 100 basis points of blended margin gone on the full order.

The fix is not tighter sales discipline. The fix is a system that flags the margin delta at the moment each of those decisions is being made, and requires a senior-level approval to override. This is what Caterforia's margin-floor enforcement does. It is unglamorous. It recovers 150 to 300 basis points of margin on an operation that has been running without it.

Play 2: Route density via routing software, not driver intuition

The 2019 off-premise route was built by the dispatcher looking at a map and guessing. Six stops, rough order, 35 minutes of commute time, call it done. At 2026 fuel prices and driver wages, that guess is leaving 15 to 20% of the route efficiency on the table.

Real route optimization (actual multi-stop optimization with traffic data, not the dispatcher's heuristic) recovers that 15 to 20% in fuel and driver time. For an operation running six routes a day, five days a week, that is roughly $1,800 a month of recovered labor and $600 a month of recovered fuel.

The operators who resist this argue their dispatcher is better than the software. In small rural markets with a dispatcher who has been driving for 20 years, that can be true. In any market with more than 500,000 population, it has not been true since about 2020. The math on traffic variation alone makes the software more accurate than the human.

Play 3: Menu engineering toward labor-efficient items

This is the play most operators have not made and should. Menu engineering in catering has traditionally been about food-cost optimization: which items have the best contribution margin per plate. In a labor-expensive 2026, the better frame is contribution margin per kitchen-labor-minute.

An item that contributes $4.20 per plate but requires 3.5 minutes of assembly is worse than an item that contributes $3.80 per plate and requires 1.5 minutes of assembly, at scale. The second item produces $2.53 of margin per labor-minute; the first produces $1.20.

The Caterforia reporting layer runs this analysis automatically once you have a quarter of event data in the system. You get a ranked list of your menu items by margin per labor-minute and you can make informed decisions about what to promote, what to reprice, and what to quietly retire.

Most operators who run this report for the first time find that three to five of their most-promoted items are actually in the bottom third on labor-efficiency terms. They were pushed by the sales team because the food-cost percentage was good. The labor cost was invisible.

What not to do

A few plays that get pitched but do not hold up.

Do not try to cut out the coordinator layer. There are vendors pitching "AI sales coordinators" that are supposed to replace your inside sales team. The realistic version of AI in catering sales is a drafting assistant. Replacing the coordinator is a capacity-reduction play that loses you more revenue than it saves.

Do not try to exit corporate drop-off entirely. Several operators have pitched us on shifting out of drop-off into full-service only. This works if you have the sales team and the brand to pull it off. For most operators, corporate drop-off is still the volume backbone and the cash-flow smoothing. Exiting it costs more than fixing the margin.

Do not commoditize with the delivery aggregators. Listing on ezCater, Hungry, or Sharebite at their standard take rate is a quiet way to push another 12 to 18% of margin to a third party. The aggregators are a sensible acquisition channel for specific use cases. They are not a sensible backbone.

The economics of fixing it

Operators who run all three plays well see gross margin on off-premise return from the 28 to 31% floor back toward the 35 to 38% range, sometimes higher. That is the difference between a struggling catering business and a healthy one.

The tooling matters here. You cannot enforce margin floor without real-time cost cascade. You cannot do route density without routing software. You cannot run margin-per-labor-minute analysis without structured event data.

Caterforia is built around exactly these three plays. Start at $1 a month and try it against your current stack for a quarter. The math either works or it does not.