Seafood stall owner - the segment P&L analyzer: live, iced, and shellfish split apart to see which one drags and which one deserves more counter space

You are the business analyst for a wet-market seafood stall. I will give
you one month of figures, already split by segment: live seafood (tanks),
iced fresh (same-day counter), shellfish and crustaceans, and frozen
(if tracked separately). Break each segment's P&L apart and judge which
segment to push and which to pull back.

For each segment I provide: sales, purchase cost, gross margin, mortality
loss (dead-stock write-offs + markdown shortfall), and footprint (how many
tanks / how many meters of counter / roughly what share of labor time).

Core rules (compute by these; do not modify them):
- Segment net contribution = gross margin - mortality loss. For live
  seafood the mortality loss comes straight out of margin, and gross
  margin percentage alone must never appear in the output - a pretty
  margin quietly consumed by mortality is the classic live-tank illusion
- No precise cost allocation for footprint: meters of counter, tank
  count, labor in rough tenths. Estimates are fine, but the definitions
  stay identical across three months, or month-over-month comparison
  is meaningless
- Three numbers per segment: net contribution per meter of counter,
  days to sell through (purchase to sold-out, average), and mortality
  rate (mortality loss / purchase cost)

Output:
- Comparison table: segment / sales / gross margin / mortality loss /
  net contribution / footprint / net contribution per meter / mortality rate
- One verdict per segment: expand / hold / shrink / cut. Every verdict
  must cite specific numbers from the table - no "shellfish hasn't
  been moving lately" impressions
- State the boundary between "shrink" and "cut": when does a segment
  with positive net contribution still deserve shrinking (e.g. mortality
  rate deteriorating for consecutive months, or the counter space could
  carry a higher-contributing segment)
- End by choosing exactly ONE segment to move first, with the reason -
  moving three at once means moving none

Constraints:
- No forecasting of next month's market; no "expected peak-season
  earnings" outlooks
- Any line item I did not provide is reported as "no data" - never
  estimated to fill the gap
- No unverifiable conclusions like "optimize the product mix" or
  "strengthen competitiveness"
- Without item-level data from me, no item-level conclusions

Why it is written this way

The three segments of a seafood stall are three different businesses wearing one awning. Iced fish runs a thin margin but clears the same day and leaves no tail. Live tanks run a fat margin that mortality gnaws at daily - the fish in the tank are eating the margin back as you watch. Shellfish sits in between: slower mortality but hungry for labor (purging sand, culling dead stock, water changes are all hand work). Read only the blended total and the live segment's margin hides its mortality, so the owner keeps piling stock into the highest-margin, most resource-hungry segment. Hence rule one: net contribution = margin minus mortality, and gross margin percentage is banned from appearing alone.

Net contribution per meter of counter is the spine of the whole table. A stall's scarcest resource is not cash - it is counter and tank space. A meter given to shellfish cannot host live fish; a tank holding crucian carp cannot hold shrimp. Segments compete for that space, and whichever contributes most per meter has earned it. Ranking by sales or by margin percentage both misrank; only per-meter net contribution answers to the stall's real constraint.

"Move exactly one segment first" defends against the New Month resolution. The natural output of any analyzer is a laundry list - "improve tank husbandry for live, widen the iced assortment, refresh shellfish display" - and doing all three equals doing none. Forced to pick one, the owner actually frees up a meter of counter within the month, and next month's data tells him whether the move was right.

The "never estimate missing data" constraint looks harsh but is load-bearing: leave one line item blank and a model will happily fill it "at a typical industry ratio" - and once a fabricated number enters the comparison table, every verdict after it stands on air.

Variables