BBQ four-state inventory count SOP - frozen blocks, thawed trim, marinated, skewered: separate books per state, and every state change that leaves no record is a black box

You are an inventory management consultant for a BBQ restaurant. I am the count specialist. Write "four-state inventory counting" as an archivable SOP. Conversion coefficients are this shop's measured numbers at [ ] - not yours to fill.

Background: the same meat exists in our shop in four states -
1. Frozen blocks (original case / primal cut, in the freezer)
2. Thawed trim (thawed and cut, not yet marinated)
3. Marinated, waiting to be skewered
4. Skewered, ready to grill (held chilled)

The SOP writes down to who counts, when, how, and where each number gets recorded, covering:

1. Common baseline: all four states fold into "net kg" on the count sheet - frozen x house yield [ ], thawed trim at actual weight, marinated at post-marination weight, skewers at per-skewer spec [ ] x count. Where hand-skewered specs drift (a veteran and a new hire produce different skewers), state which spec is currently in force and how much drift triggers a re-measure
2. Timing and witnesses: count after closing, two people present (count specialist + kitchen handler), never during service - draws from stock don't stop at peak, and a mid-service count is wrong by construction
3. Method per state:
   - Frozen: whole blocks counted + loose trim weighed, two separate lines in the book
   - Trim / marinated: weighed with container tare; whether marinade liquid counts is written to one fixed convention, settled once and never flipped
   - Finished skewers: counted by skewer; visibly off-spec skewers (oversized, undersized) get their own line - never folded into the normal count at the standard spec
4. State-conversion logging (the core of the SOP): every thaw, marination, and skewering writes one line - date, quantity, hands. The conversion log is the only explanation source for count variance: without it, everything between "thawed two cases Monday" and "eight hundred skewers out Wednesday" is a black box
5. Variance records: conversion loss (process-driven, re-measurable) and count variance (over/short, awaiting explanation) in separate accounts, no netting - offsetting count variance with conversion loss is how missing stock hides inside craft
6. Counting cycle: bottled goods weekly spot counts, meat in all four states monthly full counts, dry goods monthly, conversion log reconciled weekly
7. Escalation: the same item posting same-direction variance two months running beyond [ ] moves from "record" to "dedicated investigation," with the starting ledger and time window for the backward walk written down

Three red lines (discovery voids the count - redo it):
- Counting alone - numbers exist to feed ledgers, and one person's count has zero evidentiary value
- Revising a count after the fact - what was counted is what gets written; a revised sheet is worse than none because it arrives wearing the costume of credibility
- Filling in "roughly" - an estimate in the ledger becomes one more unexplainable source in next month's variance

Constraints:
- Yield and skewer-spec coefficients must carry "measured by this shop on [date]" - "typical BBQ shop" numbers are forbidden
- No "tighten inventory discipline" filler - every rule has to land on a form or an action
- Two pages maximum; the "method per state" section must survive being torn out and taped to the freezer door

Why it is written this way

Four states are the structural difficulty of BBQ inventory, and the reason this SOP exists. A convenience-store count adds two spaces (shelf + backroom); a BBQ count converts between four states - the same animal changes shape four times between the door and the grill, and each state keeps books in its own unit (cases, kilos, skewers). Without per-state ledgers a count sheet is one grand total against another, its variance explainable by nothing; with them, every variance at least localizes to "between which two conversions."

Conversion logging is the heart of the document. The only thing connecting the four state-books is the conversion record - delete the two lines "thawed two cases Monday, skewered out eight hundred Wednesday" and everything in between is a black box. Every attribution in monthly variance analysis assumes those lines exist. That is also why the conversion log reconciles weekly while the full count runs monthly: the log is written daily, and discovering it broke at month-end means a month-sized hole.

Separate accounts for conversion loss and count variance is the easiest rule to get wrong and the most lethal. Netting them produces the bookkeeping effect of "yield ran low this month" over what may be twenty kilos of meat gone - process loss is an explainable, re-measurable physical quantity, count variance is an anomaly awaiting explanation, and mixing them destroys both: management learns to absorb every shortfall as "process drift," and actual loss waits forever for a blow-up that already happened months ago. The escalation rule - same item, same direction, two months, past threshold - puts a boiling point on that slow simmer.

The three red lines all guard provability, not accuracy. Counting alone, revising after the fact, filling in estimates - none of these break the numbers themselves; they break what the numbers can prove. When a month's variance needs explaining to the manager or the owner, a two-signature, unaltered count sheet with dated measured coefficients is the count specialist's only shield. The penalty is void-and-recount rather than a talking-to, because provability cannot be patched after the fact - it can only be redone.

Variables