Fast food inventory count specialist - monthly full-count SOP: blind count sheets with a blank quantity column, re-counts done by a different pair, and a cut-off timestamp locked before anything in a live-service count

You are an inventory count procedure designer for a
fast food chain. Write the monthly full-count SOP
for stores. Readers: the inventory count specialist
and drafted-in staff.

Phase one: preparation (day before the count)
- Cut-off announcement: fix the accounting cut-off
  timestamp for this count [recommended 22:00
  after close - if a live-service count is
  unavoidable, see the addendum]; announce it to
  all staff; no requisitions or returns process
  from that moment until the count completes
- Bin reset: every item returned to its fixed
  location; open packages merged; semi-finished
  and opened items in their own zone; anything
  un-resettable goes in a marked "to count" pen
- Sheet printing: item names and units in bin
  order, quantity column blank - this is the
  blind sheet, and book quantities appear
  nowhere on the counting floor
- Unit table: easily confused units (bag/box,
  g/lb) carry their conversion at the sheet
  header, against "one bag recorded as one box"
- Pairing: teams of two - one counts aloud, one
  records and signs on the spot; nobody counts
  the storage area they personally manage day
  to day

Phase two: the count
- Bin by bin, item by item: the counter calls
  "name + quantity + unit," the recorder repeats
  it back before writing - both halves are
  mandatory
- Anything unresolvable (jumbled stacks, broken
  seals): flag it, never estimate - "about three
  cases" does not exist on a count sheet
- Opened and semi-finished items: converted to
  base units per the store's conversion standard
  (no standard = listed "unconverted"; inventing
  a conversion rate on the floor is forbidden)

Phase three: after the count (within [48] hours)
- Variance re-count: book vs. actual; any item
  over threshold (per-category thresholds,
  provided by me) is re-counted physically by
  a pair [outside the original team] - fresh
  eyes are a hard rule, because self-review
  re-reads the original error
- Still over threshold after the second count:
  into the variance report - item / book /
  actual / variance qty / variance value, both
  counters signing
- Adjustment: system stock adjusted per the
  report; the item joins "watch status" next
  month, and [2] consecutive same-direction
  variances escalate to variance analysis
- Archive: count sheets, variance report,
  adjustment records kept [12] months

Addendum: counting during service hours (only when
a post-close count is impossible)
- Cut-off to the exact minute: requisitions
  before the timestamp post to this cycle; after,
  to the next - and the timestamp is printed on
  the count sheet
- Work-in-progress (sold but not handed off,
  prepped but not sold) goes on a separate sheet,
  never mixed into stock
- A quick post-close check the same night:
  confirm every movement between cut-off and
  close was recorded into the next cycle

Prohibitions:
- Never hand out book figures before counting -
  blindness is the premise of the entire
  procedure, and destroying it leaves only
  theater
- Never fill any cell from "last month was about
  the same"
- Never net variances off each other (surplus
  quietly canceling deficit) - each item's
  variance reported separately, because netting
  cancels two problems of different natures into
  "no problem"

Constraints (for the AI):
- No invented accounting standards or statute
  clauses - approval authority for adjustments
  and escalation amounts carry "per this store's
  finance policy" [blanks] for me to fill
- Every threshold and duration is a [blank]
  variable; the SOP's structure depends on no
  specific number

Why it is written this way

Blind counting is the soul of the SOP and the only real dividing line between "a count" and "copying the ledger." Hand the counter the book figures to check against and human nature slides straight to "close enough, tick" - especially past midnight with a hundred SKUs to go. A sheet with a blank quantity column physically removes the copy path: you can only write down what you counted. The rule that nobody counts their own storeroom extends the same logic - self-counted stock never produces a variance worth having.

Mandatory fresh eyes on variance re-counts target the psychology of self-review: asked to re-check their own discrepancy, the original counter's first motive is proving "I wasn't wrong," and they re-count from the same spot through the same eyes - wrong the same way. A different pair is a second, independent measurement; the two signatures pin both measurements to named people, which is what lets the variance report survive the follow-up analysis.

The ban on netting deserves its own word: a little extra of item A quietly canceling a little short of item B is the most convenient closure in the world - and it destroys the most valuable signal there is, because surpluses and deficits have entirely different cause profiles (surpluses usually trace to receiving and posting errors; deficits point toward loss and leakage). Netting them is cutting both wires at once. The live-service cut-off rule is the same realism: without a timestamp locked to the minute, tonight's "accurate" number is falsified by the next requisition out the door - an unlocked cut-off invalidates the whole count.

Variables