Fresh supermarket monthly full-count SOP - freeze the paperwork, blind recounts, fixed rules for weighed goods

You are a physical-inventory process consultant for a fresh-food
supermarket. Turn our month-end full count into an archivable SOP
covering the day before, the day of, and the day after. Anchor it on a
mistake we have already made: last count produced a pile of variances,
and half turned out to be "paperwork still flying" - receipts not yet
posted, write-offs stuck in approval. Those variances were fiction.

Three sections:

[T-1, the day before]
- Paperwork freeze announcement from 14:00: all departments told that
  goods arriving after tomorrow's receiving cutoff book to next period
- The receiving cutoff is a parameter in the SOP (default 10:00 on count
  day; later arrivals go to a marked "do not count" staging area)
- The close-of-business paperwork checklist, each line with an owning
  role: all receipts posted, all write-off tickets approved, all inter-
  store transfers moved in and out, all returns logged
- Count sheets issued by numbered zone: sales floor by rack group,
  backroom by storage zone, one sheet per zone, with the rule printed
  on it: "Count only your assigned zone - do not help the neighbors"

[Count day]
- Weighed-goods rules, the loudest argument in fresh counting, fixed as
  a table:
  - Sealed cases: case weight x case count; do not open
  - Open bulk: weigh for real, one decimal place
  - Prepackaged: count packs by label; do not open and reweigh -
    reweighing introduces more error than the labels carry
  - Bundle packs (buy-one-get-one): count as one set, never as two items
- First count: each zone counts itself, recording actual quantities
  only; system on-hand numbers are not handed out (prevents copying)
- Recount: a different person, blind - blank sheet, no access to first
  counts. Only matching numbers count
- Third count: where first and recount disagree beyond the item
  threshold, a third person counts; two-of-three wins
- Preliminary variance table out by 17:00

[T+1, the day after]
- Within 48 hours, every variance reviewed: recount plus document-trail
  check (can receiving, write-off, transfer or return paperwork explain
  this number?)
- Review verdicts, exactly four allowed: miscount (correct to the
  two-of-three figure) / paperwork unposted (book it) / real variance
  (route to loss prevention) / unexplained (suspend and escalate).
  A fifth verdict, "probably fine," does not exist
- Real variances above the reporting threshold go to the store manager
  and field supervisor the same day

End the SOP with a compact role-by-time responsibility table: every
action, who leads it, who supports.

Do not:
- Write goal-language like "ensure data accuracy" - only actions,
  deadlines, owners
- Fold "root-causing the variance" into this SOP - that is loss
  prevention's process, after the review. Mix them in and recounters
  stop reporting real variances
- Recommend "getting a count system / count machines" - write this for
  paper sheets and a dedicated entry clerk

Why it is written this way

"Paperwork still flying" is the largest fake-signal source in a full count: at the instant the book figure is extracted, unposted receipts and half-approved write-offs bleed pure process noise into the book-to- physical gap - and chasing that noise wastes analyst-days. So the SOP's center of gravity sits in freezing, not counting; the T-1 zeroing checklist is what makes every downstream number mean anything.

Blind recounts exist because recounters copy. The moment a recounter can see the first count, they drift toward it and the check quietly dies. "Blank sheet, matching numbers count, third count breaks ties" buys independence with process instead of character - any review that depends on discipline alone becomes a transcription exercise within three months.

The weighed-goods table exists because that is where fresh counts scream at each other: sealed cases by standard weight, open bulk by actual weight, prepack by label count. Leave the rules open and every zone counts the way it finds convenient, cross-zone variances become meaningless, and the review drowns in "well, how did your side count it?" Each line closes one species of argument.

Four verdicts and no fifth: suspensions and escalations look bad, but hand the reviewer a soft option like "probably fine" and they will take it every time - and vague verdicts swallow every real variance. Better ten suspended lines than one fog.

Variables