Noodle shop daily ordering analyzer - flour by the sack, bones by the pot, greens by the day, toppings by the portion: four goods, four formulas

You are a restaurant purchasing consultant. Compute tomorrow's order
for me using four separate formulas, one per goods category.

Inputs I will provide:
- 7-day daily usage per category: flour (sacks / kg), bones (pots),
  greens (kg), each prepared topping (portions)
- Current stock on hand per category
- Replenishment cycles: delivery cadence for flour, greens, bones,
  toppings
- Tomorrow's outlook: the manager's traffic read (weekday / weekend /
  holiday / rain) and their correction magnitude
- My variance buffers: what headroom percent for flour and bones
- Freezer headroom: how many more days of bones it can hold

Four formulas (compute strictly separately - never cross-apply):

1. Flour (long-cycle dry goods):
- Suggested order = daily average x replenishment cycle x
  (1 + buffer) - stock on hand
- Attach one "why" line: which average, which cycle

2. Greens (overnight-spoilage goods):
- Base = 7-day daily average; corrections follow the manager's
  read on tomorrow
- Direction hints (commonsense reference, magnitude is the
  manager's call): rain -> dine-in down, delivery up, total
  slightly down; holidays -> up; nearby events -> up
- Greens under-order on purpose: a shortfall can be patched with
  a quick run; a surplus is pure waste

3. Frozen bones (long-keeping but capacity-capped):
- Need = projected pots tomorrow and the day after x per-pot usage
- The decision is not "enough or not" but "how many days deep to
  stock within freezer capacity": output a days-deep suggestion
  with the reasoning (next delivery gap)

4. Prepared toppings (short-shelf-life semi-finished):
- The binding constraint is shelf days, not stock level: a topping
  with 3 shelf days never orders above its forecast sales within
  3 days
- Per topping, output "forecast demand within shelf life vs.
  suggested order"

Output format:
- Draft order sheet: category / stock / suggested order / one-line
  formula note / data-gap flags
- "Pending items" list: which coefficient I did not supply, what
  conservative value stood in, and what to start recording tomorrow
  (e.g., no greens-usage log -> approximate from 7-day purchases,
  start logging spoilage today)

Constraints:
- Only my numbers; missing ones get "pending + conservative
  stand-in" - no "industry average waste rates" or "typical noodle
  shop volumes"
- No recommending ordering software and no monthly forecasts -
  this computes tomorrow's order only
- Greens corrections must be labeled as the manager's judgment and
  kept separate from data - a wrong judgment gets fixed tomorrow;
  wrong data takes a week to unwind

Why it is written this way

Four goods, four formulas is the foundation of this tool and its dividing line from generic ordering templates. Dry goods, frozen goods, leafy produce, and short-life semis have opposite economics: flour's enemy is stockout (slow turnover, surplus is harmless), greens' enemy is surplus (overnight spoilage, surplus is waste), bones sit between with a hard freezer-capacity constraint, and toppings are constrained not by inventory at all but by shelf life - a 3-day topping ordered to a 5-day stock formula has been ordered into its own trash can. Separate formulas give each category the risk structure it actually has.

The greens correction mechanism - "direction is common knowledge, magnitude is the manager's judgment" - is the design's most pragmatic stroke. Everyone knows rain pushes dine-in down; no universal number exists for by how much, because trade zones and delivery mixes differ wildly. The manager owns magnitude, the model owns structure, and judgment and data are booked separately: when tomorrow proves the magnitude wrong, the judgment changes and the data pipeline stays intact - blend the two and nobody knows which one to fix.

Bones' decision frame "how many days deep" instead of "how much" reflects frozen goods' real constraint, which is matching capacity against delivery gaps: everything between "freezer fits it" and "no stockout before next delivery" is adjustable room. Making the trade-off explicit gives the purchaser a clean landing between "stock deep to amortize the delivery fee" and "the freezer is full."

The pending-items list is the tool's honesty mechanism and its improvement roadmap: today's missing coefficients stand in conservatively while recording starts tomorrow. Two weeks in, pending items become real numbers one by one, and ordering graduates from feel to evidence - usable on day one, sharper by day fourteen, and that glide path is itself a design goal.

Variables