Seafood stall inventory specialist - the four-category counting-basis SOP: how to estimate live stock, drain shellfish, and book mortality every single day

Write a "Seafood Counting-Basis SOP." The core difficulties: live
stock swims in tanks and cannot be weighed piece by piece,
shellfish sits in water carrying water, and mortality happens
every day - not just on count day. The SOP locks the basis for
each of the four categories so any two months' counts are
comparable.

1. The four categories (each: measurement / valuation basis /
honesty handling)

Live (in tanks):
- Count heads per tank per species (large stock counted; small
  stock like shrimp eyeballed by density, recorded as a range
  such as 80-100, never a point) x average unit weight of the
  size grade = estimated weight
- Average unit weight source: the latest arrival's size record;
  for species with volatile incoming sizes, the mean of the last
  [N] batches
- Honesty handling: every tank logs its method (counted |
  eyeballed). Eyeballed tanks are flagged separately on the count
  sheet and analyzed separately - an eyeballed tank's variance
  precision is not the same animal as a counted tank's, and
  blending them means finding causes in neither
- Crabs valued net of rope weight, at the rope ratio measured
  at receiving

Iced (counter):
- Count only after close, when the counter is cleared (or holds
  only marked keep-overnight items) - a midday count counts
  nothing
- Keep-overnight items valued at their marked-down price, not
  original: an asset is worth only what it can still sell for

Shellfish (troughs):
- Lifted and drained [N] seconds, then weighed. Water-weighted
  figures are not assets; they are flattery
- Dead-but-unculled counted at zero - cull on discovery, never
  left inside the count to pad volume

Frozen:
- Pieces x spot-measured net weight per piece (glaze included
  in the basis)
- Thaw-refrozen product (clumping, uneven glaze) flagged
  separately and valued at its markdown worth

2. Daily mortality logging (the other half of the basis,
outside count day)
- Log mortality daily: species / quantity / destination
  (markdown | write-off), recorded the moment dead fish leave
  the tank - not reconstructed at month-end
- Month-end identity check: opening stock + purchases - sales -
  logged mortality ~= counted closing stock
- When the identity fails, the audit order is fixed: first
  ticket-item reclassification (live sales booked as iced is
  the classic), then the purchase receiving records for size
  grade, and only last, people

3. Count sheet format
- Four category blocks; each row: species / method / quantity /
  valuation basis (purchase price | markdown) / value
- A variance rate per category, never one blended store-wide
  rate - 3% on live and 3% on frozen are problems of entirely
  different natures

Format: two pages max, tables first; the [N] values are
calibrated once by this store, then written into the SOP as
constants - and never nudged month to month.

Why it is written this way

This SOP defends against one thing above all: basis drift. The bulk of a seafood stall's count variance is usually not missing goods - it is definitions. This month the shrimp get eyeballed at 90; next month the same tank gets eyeballed at 100. This month the shellfish drain ten seconds; next month, rushed, three. Every basis wobble turns variance into noise, so each category must answer three questions: how it is measured, where the measurement's basis comes from, and how reliable the method itself is (the honesty handling).

"Flag eyeballed tanks separately" is honesty acknowledging reality: counting shrimp one by one is impossible, so estimate - but once estimates and counts blend into one variance rate, variance analysis is dead; flagged separately, at least you know which numbers deserve how much trust and which tanks to recount first. Recording ranges (80-100) instead of points serves the same end: uncertainty written honestly on the number beats a pretend-precise "90."

The identity check and its three-step audit order exist to prevent "variance found, people blamed." A seafood stall's count gaps live mostly in the records: a live sale ticketed as iced, a size grade mis-transcribed by one decimal. Putting "people" last is not diplomacy - it is audit efficiency: the first two steps run on tickets and receiving forms, cheap and high-hit; open with an accusation instead, and the team reads the count as a witch hunt, and next month's mortality log starts quietly shrinking.

"Written in, never nudged" is the last link in basis discipline: a basis that may be adjusted monthly is not a basis - every adjustment "eliminates" a batch of variance that then nobody ever knows occurred.

Variables