threefoldWORKSLet’s talk
Menu

WORKFLOW

Food-cost gap widening at one branch

Illustrative example Restaurants

The count is monthly, the food cost percentage arrives with the month-end books, and the branch that has been losing a little every week has been losing it for thirty days before anyone can see it.

The workflow

One branch counts its stock, what the recipes say it should have used and what it actually used drift apart week after week, and nobody is told which of portioning, spoilage or pilferage is doing it while the trail is still warm.

Illustrative example Theoretical vs actual food cost · closed

Weekly count · Branch D

Week 37 · ten highest-cost items
  1. Count sheet readDone
    Count photo from the branch · Monday 08:20
  2. Theoretical vs actualDone
    Gap 4.2 points · third week wideningEvidence attached · week 37 count, week 34 to 37 trend
  3. Branch answer and decisionDone
    Portioning on one dish · branch confirmed4.2 ptsClosedRetraining the line · re-count next week
    Decided by the owner · Tuesday 10:40

The owner made the only decision in it.Every step on record

What happens today

The count is written on a printed sheet, photographed, and sent to the office on Viber, where somebody types it into the stock spreadsheet. The food cost percentage comes out with the month-end books. Counting monthly leaves a thirty-day blind spot, as the accounting firm that works with Philippine restaurants describes it, and a blind spot is exactly how long a small weekly loss gets to keep running.

The same firm describes what changes when a chain grows: controls that worked informally at one location stop working the moment there is a branch the owner cannot personally oversee, and that is usually when food cost percentages become unreliable. By the time the number is unreliable at the chain level, it is nobody’s number in particular — the branch that is actually drifting is averaged in with the branches that are fine.

What ordinary rules handle

Most of this is arithmetic on a weekly clock, and we build it as ordinary rules, because ordinary rules are cheaper to run, easy to check, and behave the same way next month.

  • Turning the week’s sales into a theoretical usage from the recipes, and setting it against the counted usage per branch.
  • Tracking the gap per branch and per week, so a branch is compared to its own last four weeks rather than to the chain average.
  • Flagging a gap past the threshold the chain sets, and routing it to the branch manager who owns that count.
  • Reminding on a schedule while the explanation is outstanding, and holding the exception open until it is answered.

Where the assistant follows through

This is the part we call follow-through: the assistant checks the records, prepares the action, routes it, chases the reply, and raises what it cannot settle.

What rules cannot do is read a photographed count sheet with two items crossed out and rewritten, or decide that the branch’s reply — “yung chicken, natapon yung isang batch nung Sabado” — is a spoilage answer and not a portioning one. AI reads the sheet and the reply and makes those calls, and it is a smaller part of the job than it sounds.

  1. The count sheet becomes a draft count. The photographed sheet is read off against the item list. Draft, because nothing has been posted yet.
  2. Theoretical is set against actual, for that branch alone. The week’s recipes against the week’s count, and the gap put next to the same branch’s previous weeks.
  3. A widening gap is raised with its evidence attached. The count, the sales, and the four-week trend travel with the exception, so the next person to look at it is not re-doing the check. A gap of three to five points between theoretical and actual food cost is the point that same accounting firm treats as the signal to look; whatever number a chain picks, the assistant uses the chain’s own.
  4. The branch is asked which of the three it is. Portioning, spoilage, or pilferage — asked as a question with the numbers already attached, in the thread the branch already uses, and chased on schedule if it goes unanswered.
  5. What the answer cannot settle is escalated. A gap that widens for a third week, or an answer that does not fit the count, goes to the owner as a pattern rather than being logged as three separate weeks.

Where a person decides

Retraining the line, re-pricing the dish, or investigating the branch is the owner’s call, and the assistant cannot take it. It measures the gap, collects the branch’s account of it, and stops there — not because a check failed, but because that is where the line was drawn when the workflow was built.

The three answers cost very different things, and one of them is about a person’s job. So the decision stays on the record with the name of whoever made it, and the week it was made about. A quarter later, the question “what did we decide the last time this branch drifted, and did it work” has an answer.

Where to start

Start with one branch and the ten items that carry most of your food cost — the ones you would notice on the invoice. One weekly count, one gap, one question routed to one branch manager is enough to see whether the drift is real and what it is worth in a month.

Nobody at the branch has to learn anything new. They count the way they count now and send the photo to the same group, and the manager answers a message.