University of Hawaiʻi at Mānoa · 4 stations · ~1,600 meals a day
The screen the kitchen
would see every morning.
Today every diner's card is scanned at the front desk, so we know exactly how many people eat. We do not know what they eat, what we over-cook, or what we run out of. This prototype shows what changes once four station sheets and one scale start feeding a dashboard.
01 · User journey
One student, one lunch, five data points.
Nothing about the student's experience changes. What changes is that each step now leaves a record.
-
11:52
Card scan at the front desk
Kai taps a Mānoa One Card. Transact records a swipe with a timestamp.
Captured today headcount, arrival time, plan type -
11:54
Picks a station
Island Kitchen has Kalua Pork today. The station lead's production sheet says 190 portions were forecast and 214 cooked.
New forecast vs. prepared, per dish, per station -
11:56
Plates up
A standard 10 oz spoodle. The portion size on the recipe card is the tool on the line, so portions cooked equals portions served.
New portion standard = recipe standard -
12:20
Eats. Some is left on the plate.
Kai leaves a little rice. Twice a week the dish room weighs the scrapings bin and divides by the door count.
New plate waste per diner (sampled) -
14:05
Service closes
Island Kitchen counts 21 portions left. 14 go to tomorrow's fried rice, 7 hit the scale next to the bin with the reason code "overproduced".
New leftover, disposition, weight, reason
Meanwhile, the kitchen's day
02 · Check-ins
We already know when people arrive.
Card scans give the arrival curve for free. The curve tells the line when to fire the last batch and tells the schedule when the rush actually is.
Average diners by weekday and meal
Steady-state weeks, holidays excluded. Tuesday lunch is twice Saturday lunch. Cooking the same amount on both is where today's waste and today's rush both come from.
Lunch headcount across the semester
Weekly rhythm, holiday dips, Thanksgiving recess, finals. Dashed line: the calendar-aware forecast once the system is on.
03 · Dashboard
The Monday-morning page.
Six panels. The chef reads it in two minutes and picks one thing to fix this week.
Kitchen waste cost per meal, by week
Shaded: baseline weeks. Dotted line: system switched on.
Dishes that ran out, by week
Fewer stockouts and less waste at the same time is the whole point.
Forecast vs. actual diners, last 14 lunches
How wrong were we, and in which direction.
Top waste dishes by dollars, baseline weeks
Ranked by money, not pounds. Ten dishes carry most of the bill.
Stockout map: station × meal × weekday
Share of dishes that ran out before the meal ended. Dark cells are the rush.
04 · Food waste
Where the food actually goes.
A scale next to the bin, a reason code, and a cost per portion turn pounds into a number leadership can act on.
Portions prepared per week: eaten, held, repurposed, discarded
Kitchen waste dollars by reason
Overproduction is the lever. Trim and spoilage are smaller and slower to move.
Kitchen waste dollars by station, per 4 weeks
Which line to stand at on Monday.
Waste log, last entries
What the scale station records| Time | Station | Item | Stage | Reason | Net lb | $ | By |
|---|
05 · Popular days
Every dish has a best day.
Take-rate = portions served ÷ diners at that meal. A dish at 35% on Monday and 20% on Friday gets two different production numbers, not one.
Take-rate by weekday
Cells show the share of diners who take the dish. Darker = more popular. Best day per dish is outlined.
06 · Tomorrow's production sheet
Tuesday lunch, generated last night.
Forecast diners × each dish's Tuesday take-rate, then a margin sized by what it costs to be wrong. Cheap and repurposable dishes get a generous margin. Fresh fish does not.
| Station | Dish | Class | Forecast | Percentile | Make (Q*) | Cook now | Hold | Day-two use |
|---|
Percentile = the newsvendor critical fractile, cost-of-shortage ÷ (cost-of-shortage + cost-of-leftover), floored at the 90th percentile for service level. "Hold" is par-cooked and fired only if the card scans say the diners are coming.