Community supermarket receiving - depot truck spot-count checklist; hundreds of SKUs per truck cannot be piece-counted, so sample in layers by value and risk, with what to count, how many, and how to annotate discrepancies all fixed

You are a receiving advisor for a community
supermarket. Build me a spot-count checklist
for full depot deliveries, used by the
receiver at the truck. Constraints stated
up front: [SKU count] SKUs per truck, the
driver can wait [waiting time] and no more -
a full count is impossible. The goal is not
coverage; it is pressing the few minutes
onto where errors are most expensive.

Four layers, each fixing what to sample,
how many, and what fails:

Layer 1. Count in full (every unit)
- Cigarettes and high-value resellables
  (supplements, premium liquor, razor
  heads): unit by unit
- Any SKU with a discrepancy on record
  within the [lookback window]: restored
  to full count, dropping back to sampling
  after [clean streak] consecutive clean
  full counts
- Fail: any single unit short means the
  whole line gets signed at actual count

Layer 2. By weight (heavy goods)
- Rice, flour, oil, beverages, home care -
  carton-standard heavy goods: sample
  [weighed cartons] cartons on the scale
  or count by piece
- Fail: actual net weight below carton
  standard (after tare) by more than
  [weight tolerance] triggers a full
  pallet count
- Note: for shrink-wrapped pallets count
  pallets first, then sample cartons;
  intact wrap allows counting outer
  cartons only, broken wrap forces
  carton-by-carton

Layer 3. By date (lot codes)
- Sample [date cartons] cartons per pallet
  for production dates: a pallet mixing
  more than one lot forces a date check
  on every carton on it
- Fail: any batch with remaining shelf
  life under [shelf-life floor] is
  rejected on the spot or signed as
  conditional acceptance - never waved
  through unhandled

Layer 4. By appearance (damage, contamination)
- Crushed, wet, dented, bloated cartons:
  stack separately and open to count
  contents - shortages inside damaged
  cartons are the easiest to bury under
  "it was damaged anyway"
- Fail: contents short or contaminated -
  photograph, annotate separately, never
  shelved as normal goods

Annotation rules (how discrepancies are
written, all fixed):
- On the delivery note: item / documented
  qty / received qty / reason (short-
  shipped / damaged / wrong item) - all
  four fields or the annotation is void
- "Received qty" is mandatory; "3 short"
  alone is not a valid annotation
- Rejected lines are crossed out and
  marked "rejected," driver countersigns;
  a driver refusing to sign gets a photo
  of the note plus the plate number, and
  the annotation is written anyway
- Annotations in [copies]: driver, store
  copy, depot copy each get one

Format:
- One page A4; each layer gets tick boxes
  and a "sampled how many / how many
  exceptions" counter
- One timestamp line at the top: arrival
  time / finish time - written for
  yourself. Overrun cuts layer 3's volume,
  never layer 1

Forbidden:
- No "check every item carefully" - an
  instruction nobody can execute
- No "sample as circumstances allow" -
  every layer's volume is a fixed number
- No "work it out with the driver" -
  discrepancies go through annotations,
  not goodwill

Why it is written this way

This checklist accepts a premise every receiving manual refuses to: piece counting does not exist in depot receiving. Hundreds of SKUs, a driver watching the clock for the next drop - whoever follows the textbook gets buried by the next truck. So the design's core is not "how to count everything" but "where the minutes go": cigarettes in full, because one carton is a month of count variance; heavy goods sampled on the scale, because carton-standard cheating hurts more than unit shorts; dates sampled per pallet, because mixed lots are the main feed of post-promo near-expiry. The four layers map exactly to four failure modes, each with a wildly different price per error.

"Discrepancy-history SKUs return to full count" is the only rule with memory in the sheet: the biggest waste in receiving checks is re-verifying stable suppliers while sampling trouble SKUs at the same rate as everything else. With the lookback window and clean streak as parameters, sampling fire automatically concentrates on problem items - something intuition cannot sustain; only a rule can. Likewise, "damaged cartons always opened" plugs the softest seam: shortages inside damaged cartons slide by on "it was damaged anyway," and that moment is precisely the only chance to assign responsibility on the spot.

The annotation rules are written finer than the counting itself, because a perfect count with a sloppy annotation counts for nothing: "3 short" is worthless in a claim - without the documented-versus-received comparison, even the discrepancy cannot be proven. The four-field entry, the driver's countersignature (or the photo-and-plate fallback when refused) - that is the entire evidence chain for reconciliation. One extra minute at the truck buys back a week of arguing at month-end.

"Timestamp + overrun cuts layer 3, never layer 1" is the escape hatch for reality: when the morning truly catches fire, the checklist must say what to sacrifice first. Date sampling can shrink (the cost is rising mixed-lot risk, bounded); cigarettes and high-value goods can never (the cost is money, immediately). A checklist without priorities gets abandoned whole on its first busy morning.

Variables