Tobacco & liquor store inventory specialist - zoned two-person stocktake SOP - cigarettes by the carton, counter before stockroom; premium liquor blind-counted bottle by bottle with batch codes; everyday liquor sampled; the counter never sees the book figure, so filling in from memory becomes impossible

You are a stocktake process consultant for a
tobacco & liquor store. Write an archivable
"Zoned Two-Person Stocktake SOP." Context: value
density is high - a carton or a bottle of
variance is real money - and the biggest
distortion isn't a miscount, it's "counting to
the book": the counter knows the expected number
and the pen follows it. The SOP's core objective:
the count reflects only what physically exists.

Scheduling:
- Timing: outside business hours or the lowest-
  traffic window; register frozen for the
  duration; inbound, returns, and adjustments
  carry a cutoff time - anything after it posts
  to the next cycle, no queue-jumping
- Staffing: pairs, one counting and one
  recording; the premium-section counter may
  not be the person who keeps that section's
  daily log (blind counting can't survive
  counting your own books)

Zones:

1. Premium liquor (full count - blind)
- The counter holds no book figure: the sheet
  has three blank columns - product / quantity
  / last digits of the batch code - and no
  "expected" column. The sheet is printed
  specially; there is nothing to copy from
- Record bottle by bottle: product, quantity,
  batch code digits
- Only after the zone is done does the sheet
  reach the reconciler, who matches it against
  the ledger - variances surface after the
  fact, with no route back to the counter for
  a convenient fix
- Recount rule: any slot showing a variance
  gets re-counted by a different person; both
  sheets are filed

2. Cigarettes (full count - by location)
- By the carton; loose packs from opened
  cartons by the pack. Counter display first,
  then stockroom - two locations counted
  separately, recorded separately, never merged
  (a matching merged total conceals errors that
  offset between locations)
- Allocated fast-movers counted carton by
  carton; slow movers may be counted by the
  sealed carton, sampling one carton per line
  for inner-pack integrity
- Ledger reconciliation: counter cartons +
  stockroom cartons + sales during the period =
  last count's closing. The equation doesn't
  close, it's a variance - and the stocktake
  isn't done

3. Everyday liquor and beverages (sampling +
case count)
- Category sampling: the top three movers per
  category always counted; the rest sampled at
  the [sample rate], chosen by price-tag
  sequence (every Nth tag), never by feel
- Full cases counted by the box; one box per
  line opened to verify bottle count
- Any category showing a variance in sampling
  escalates to a full category count

4. Records and signatures
- Read-back confirmation: the counter calls
  "X brand, three cartons," the recorder
  repeats "three, noted," and only then writes
  - every number passes through two sets of
  ears
- Count sheet, variance sheet, and recount
  sheet signed separately and filed; bound
  monthly, cross-checkable against daily
  close-out sheets

Variance path (three steps, none skipped):
1. A different person recounts the slot; still
   off ->
2. Into the variance-attribution process
   (paperwork review: unposted receipts /
   unlogged returns / SKU mix-ups) - that
   process is another document's territory;
   this SOP only guarantees the true variance
   gets handed over ->
3. Premium-item variances reach the manager and
   owner the same day - never overnight

Red lines:
- Nobody "previews the book before counting" -
  including the inventory specialist
- No variances squared on the spot: two counts
  disagree, both versions are filed verbatim,
  step 3 decides
- Filling in from the book or from memory is a
  violation, full stop - in this trade's
  stocktakes, an invented number is far worse
  than an honest miscount

Why it is written this way

The blind-count sheet "physically has no expected column" is the hardest design in the SOP. The dominant source of count distortion was never poor eyesight, it's priors: the counter glimpses "should be 12" on the ledger, counts 11, and counts again - until it's 12. That isn't fraud, it's human nature, but the result matches fraud exactly: the variance dies inside the stocktake, and by month-end there's nothing left to trace. Removing the expected column from the paper, separating counter from reconciler, and never routing variances back to the counter - three structural moves that together produce actual blindness. Calling for "count what's really there" without structure has never once produced a blind count.

Cigarettes recorded by location - counter versus stockroom - plugs the "total cover" hole. Inter-location drift (counter short, stockroom long) is invisible in a merged total, and it's precisely the signal worth seeing: a counter shortage points at the sales floor, a stockroom surplus at unposted paperwork - both diagnostics die in the addition. The reconciliation equation (last closing = counter + stockroom + period sales) upgrades the stocktake from counting goods to closing the books, and "the numbers match but the paperwork doesn't" is treated as a variance because that's what it is - books and shelf have already parted ways.

The read-back convention guards against transmission error - "three" heard as "four" - which at this value density is hundreds per syllable. And "an invented number is worse than an honest miscount" stands as its own red line because it's the ethical floor of the whole document: a miscount is an honest capability problem that a recount fixes, and a fabricated number turns the stocktake from a control into a forged instrument. An SOP that doesn't say this out loud never built the ground it stands on.

Variables