He cooked the line at a hotel restaurant on Meeting Street for eleven years, and in 2013 he started catering on weekends out of a rented church kitchen in West Ashley with a borrowed van. Last year he did a hundred and forty-three events. About sixty percent are weddings, which means his year has two spines — April through June and September through November — and inside those the Saturdays stack. One Saturday last October he had three weddings running at once: one downtown, one on Johns Island, one out toward Wadmalaw.

Six hundred and ten guests between them, forty-one staff, three trucks, and no kitchen at any of the three addresses. Off-premise catering is a logistics business wearing a chef's coat. Everything that will be needed has to be on a truck, because the alternative is a forty-five-minute round trip during service, or the thing simply does not happen. The paperwork never kept up with that.

Every wedding is bespoke — nine stations and courses is normal, each with its own recipe, yield, and equipment. Contracts call for a final guest count ten days out; the real number arrives four days out, and it moves again. Dietary restrictions come from a different planner every time in a different spreadsheet, sometimes as a photograph of a handwritten list. He was rebuilding the event order by hand each time a count changed, then rebuilding the shopping list, then the prep schedule, then the pack list, usually at eleven at night.

Three plates went out wrong last year against a stated allergy. Twenty-two times somebody had to drive back for something. The money was blurrier still. He bid at a per-person price using ingredient costs he half remembered, and learned the real food cost weeks later when the invoices cleared, if he ever went back and looked.

Across the year it averaged about thirty-three percent against a target of twenty-eight, and he could not have told you which events were the problem. What we built starts at the signed contract. It reads the contract and the planner's guest file, produces the banquet event order, and versions it, so when the count moves from two hundred and twenty to two hundred and forty-one, everything downstream moves with it instead of waiting on him. It explodes each menu item into its recipe, scales to the current count using the yields he actually gets rather than the ones printed in a book, and emits four documents: purchase orders grouped by vendor and released against each vendor's own order cutoff, a prep schedule broken out by day and station, a pack list per truck per venue with every item counted out and counted back, and a staffing plan with call times by role.

The dietary piece it handles as its own problem. It normalizes whatever the planner sends into allergen categories, cross-references every ingredient in every recipe on the menu, and reports which dishes are already safe, which need a substitution, and what the substitution should be. Then it prints the card the captain carries — table, seat, restriction, and the plate that goes there. It prices the event against ingredient costs from his actual recent invoices, so he sees the margin before he signs rather than after.

After the event it reconciles what was bought against what was bid and tells him where the gap was. On stacked Saturdays it de-conflicts shared equipment across all three events, which is how the second wedding stopped being the one that got the dented chafers. Nine months and a hundred and eighteen events in, food cost has come down from about thirty-three percent to twenty-eight and a half. The drive-backs went from twenty-two in a year to two.

There have been no allergen errors. Prep now starts at six in the morning from a schedule that printed itself the night before, and he has started taking corporate work in the shoulder months that he used to turn down because he could not tell in advance whether it would make money.