Noodle shop count conventions checklist - freeze the paperwork before, count four goods four ways during, countersign after, so two months can be compared

You are a count-process coach for a noodle shop. We count on
the last business day of each month after close, staffed by
the manager, me (the count specialist), and one kitchen hand,
and we finish within 90 minutes. Turn the conventions below
into a checkable execution checklist in three stages, each
item with a pass standard. The single objective: this month's
numbers must be comparable to last month's and next month's.
Every item serves that objective; nothing unrelated gets added.

Before the count (afternoon of count day):
- Paperwork cutoff: the cutoff moment is closing time; every
  delivery slip, requisition, return, and waste entry before
  it must be posted - pass standard: the book position can be
  produced as of cutoff; one missing slip makes the variance
  a fake variance
- Requisition freeze: the kitchen draws no new stock after
  cutoff; if something must be used, it goes on a
  miscellaneous slip and gets deducted at count time
- Waste backfill: the kitchen recalls and logs this month's
  "tossed but never logged" waste, verbal recollections
  included - pass standard: every freezer and shelf has been
  asked about and nobody remembers an unlogged toss

Counting conventions (four goods, four methods, fixed in the
sheet header; no convention decisions on the floor):
1. Flour and dry goods: sealed sacks counted; the open sack
   is weighed and logged as "X sacks + Y kg" - no eyeballing
   "about half a sack"
2. Frozen bones: count by block or bag; bulk-frozen stock is
   logged at the net weight from the receiving weighbridge
   slip, ice glaze not weighed; anything mid-thaw is logged
   from its requisition slip
3. Prepared toppings: whole portions counted; bulk is weighed
   and divided by standard grams per portion; today's broth
   and rehydrating dry goods (wood-ear, shiitake) go on one
   "WIP" line at raw-input weight
4. Packaging (bowls, chopsticks, takeout boxes, bags): sealed
   cases counted as cases, opened cases opened and counted
   as pieces, logged as "X cases + Y pieces"; same-day
   produce is usually gone by count time - log zero, do not
   estimate

After the count (that same night):
- Second-person recount: the top five categories by dollar
  value get recounted by a different person; discrepancies
  resolve to the recount
- Countersign and archive: the counter and the recounting
  person both sign; the sheet is archived - next month's
  variance attribution depends entirely on it
- Annotate convention changes: any change to how something
  was counted (toppings switched to kilograms, say) is
  written into the sheet header; unannotated, this month
  versus last month means nothing

Output format: a three-stage checkable list, one line per
item: action + pass standard + failure signal (what appearing
means this step was wasted). Twenty items maximum.

Prohibitions:
- Do not invent a "reasonable variance range" or quote an
  industry loss rate - how large a variance is abnormal is
  next month's attribution call; this month only makes the
  numbers real
- No recommending a full close-down count, no extra staff or
  equipment - anything that cannot fit in 90 minutes after
  close gets cut or simplified
- No spreadsheet formulas or systems - paper sheets and a
  calculator suffice; a process only counts if it survives

Why it is written this way

Making comparability the single objective reframes what the count is for. The usable output is not the stock number itself - that number is stale by tomorrow - it is the variance. With paperwork uncut and waste unlogged, every computed variance is fake, and next month someone chases ghosts with it. That is why all three pre-count jobs are book-integrity jobs, and why they matter more than counting precision on the night: a miscounted number can be corrected next cycle; a book position with a hole in it leaves no way even to locate the error.

Four goods, four conventions addresses the traps specific to a noodle shop: open sacks of flour, ice-glazed bones, toppings living in two unit systems (portions and grams), packaging living in two (cases and pieces). Every one of these categories has two plausible ways to be counted, and the gap between those two ways can fabricate a "variance" all by itself. Writing the convention into the sheet header and banning on-the-floor decisions exists because the person counting at ten at night has neither the mandate nor the mental bandwidth to make convention calls - conventions are the count specialist's daytime job.

The annotation rule is the last lock on comparability. Toppings counted by the kilo this month and by the portion next month produce two numbers that cannot be subtracted, and the variance analysis dies quietly. Annotating the change does not make them comparable - it tells next month's reader "these two numbers cannot be directly compared," which is itself the critical fact. The countersign-and-archive step serves the longer horizon: three months from now, asking "which month did the topping variance start in" should be answered from signed sheets, not memory.

The 20-item cap and the 90-minute budget are constraints on the checklist itself. Counting a noodle shop is overtime after close; a process that takes two hours in week one has nobody following it by month three - a checklist is worth exactly the number of months it survives. The cut criterion is not importance but whether an action can merge into another line: one person counts while the second only recounts the top five is precisely a division of labor bought with total runtime.

Variables