Why average delivery time misleads every Monday stand-up
Average delivery time is a comforting statistic. It fits on a slide. It moves a little each week. It is also three clocks wearing one coat.
Ticket-to-bag is the kitchen: fire, plate, pack. Bag-to-pickup is handover: the pin, the queue, the rider who is “nearby” but two sois away. Pickup-to-door is travel: lights, rain, a hotel porte-cochère. When you average them, you can yell at chefs for a street problem, or throw rider bonuses at a bump-bar that has no timestamp discipline.
The Sukhumvit case we teach in Kitchen Cycle Diagnostics looked like a slow kitchen on the marketplace SLA. Prep was under nine minutes on most dinner tickets. The lag sat in handover behind a hotel drop-off. The average never said that. The split did.
If your POS cannot export item-level fire times, you will still have a blurry kitchen clock. That is a real limitation, not a moral failure. Publish the blur. Do not invent precision. In the flagship Ticket Anatomy module we fail a quiz if you still report a single ETA after week two. It sounds harsh. Monday stand-ups are harsher when the wrong station gets the blame.
A practical rule: never put average delivery time on a founder slide without the three component medians beside it. If you only have two clocks, say which one is missing. Logicpacketgrid will not pretend SQL can mint a timestamp the kitchen never wrote.