The Question
A Malaysian steam bun shop runs the way hundreds do across the country. The owner walks in before dawn, looks at what ran low yesterday, and calls three suppliers. Eighteen products — buns, dim sum, pastries — every one perishable. A day's worth of stock on the shelf, sometimes less.
The rule for knowing when to order was a mix of gut feel and a reorder-point number set long ago. It fired when stock dropped to a set level — but that level had been set against an older, slower sales pace. Some products ran empty on Saturday afternoons before a Monday delivery could arrive. Others sat in over-ordered piles, tying up working capital the owner could have put to work on the fast movers.
The owner wanted a discipline, not a one-off order list. Something to run every morning, in minutes. Something that would still work when a product got hotter or a supplier slowed down.
They brought the question to Pau Analytics: how do I know when to order?
What the Data Showed
We took six months of the shop's own sales — 9,852 transactions across 182 days, every one of the 18 products, all three suppliers — and looked at every one. No forecasting model. No borrowed benchmark. Just the shop's real pace, counted honestly.
The weekly rhythm came back first. Monday through Friday sat in a tight 2% band — every weekday within 36 units of the others. Saturday jumped 30% above the weekday average. Sunday followed. Every single product peaked on Saturday. All eighteen. No exceptions. Prep planning had a rule it didn't have before: add 30% across the board on Saturdays.
Then the stability check. We compared each product's last 30 days against the previous five months. Zero products crossed the ±10% threshold. The strongest riser was Lotus Paste Bun at +9.8%; the weakest was Chicken Bun at −2.9%. For a perishable shop, this is the good kind of boring. It said the six-month averages could be trusted — the rule we were about to build would hold.
Then came the finding the owner needed to hear. We compared current stock against the shop's real 30-day sales pace — and every product was within three days of empty. Twelve of eighteen sat inside their supplier lead time. The old reorder flags showed Chicken Bun, Egg Tart, Siew Mai, Har Gow, and BBQ Pork Bun as healthy — all five were in fact within two days of an empty shelf. The flags were not wrong in intent. They were wrong in number. They had been set against an older, lower sales pace, and the shop had quietly outgrown them.
What Changed
The report gave the owner three things they didn't have before.
A formula, in three lines — trigger equals daily sales times lead time, plus a 30% Saturday cushion. One number per product. When stock drops to that number, call the supplier. The formula covers the delivery gap and the Saturday spike, both at once, and anchors itself to the shop's real sales rate, not a textbook.
A trigger number for every product — Siew Mai at 178, BBQ Pork Bun at 95, Tau Sa Piah at 30 — computed from the last 30 days. Not a permanent standard. A snapshot, on purpose, because the rule is designed to be refreshed, not re-proved.
A refresh discipline. Once a month, at the same time the owner does the books, the owner recounts the last 30 days and recalculates every trigger. Thirty minutes. Weekly is too noisy — one quiet Monday can swing a 7-day average by 15%. Quarterly is too slow for perishables. Monthly fits the rhythm the owner already runs for cash-flow checks — an extra 30 minutes on an evening already given to the books.
The Result
Today, the owner checks stock each morning against the trigger table, not the old reorder flags. On Fridays, the Saturday cushion is already in the number — no Friday-afternoon guess, no Saturday-morning scramble. The three slowest products — Tau Sa Piah, Coffee Bun, Pandan Bun — now carry 30% smaller order sizes; the working capital freed up lands in the top five, where the velocity warrants it. Saturday prep runs at +30% across every product — a rule, not a decision.
The Stock Manager AI agent runs the daily check automatically. The owner opens it in the morning, sees the products at or under their trigger, and places the order grouped by supplier in a single sheet. What used to take twenty minutes of counting and calling now takes three. The monthly refresh — the only task still done by hand — takes under half an hour.
"I used to wake up and worry about which shelf would run out today. Now I open one sheet and see what to order — by number, not by feel. The surprise is that the rule didn't need to be complicated. It just needed to be mine."