Deli stall owner-operator - which products earn their shelf slot: rank by raw-weight margin, slot revenue, and spoilage rate

You are an operations analyst for a cooked-food stall in a wet market.
My counter has a handful of slots; the products currently in them are
listed in my data. I want to know which slots are earning their keep
and which one to consider replacing. Rank strictly along the three
axes below - do not substitute other metrics.

Input I will provide (per day, for the last [M] days):
- Per product: raw material purchased (weight and cost), weight and
  revenue actually sold, revenue sold at discount, spoilage weight
  at close
- Slots occupied per product (shared slots split by facing share,
  which I mark)
- Closing leftovers that were NOT discarded and carried to the next
  day, as a separate column - these don't count as spoilage, but a
  product carried over most days gets flagged

The three axes (fixed formulas, no substitutions):
1. Margin per raw weight = (full-price revenue + discounted revenue
   - raw material cost - raw cost allocated to spoilage) / raw
   weight purchased
   Spoilage MUST be folded in: cold salad carries a high ticket
   price and a nightly discard; leave it out and it ranks first
   forever.
2. Slot revenue = average daily revenue of the product / slots
   occupied
   The counter is this trade's most expensive asset; a slot that
   doesn't sell is paying the market's rent for nothing.
3. Spoilage rate = spoiled weight / (full-price + discounted +
   spoiled weight)
   This one decides whether the slot needs discount-babysitting
   each evening, and how deep.

Output format:
- Master table: product / margin per raw weight / slot revenue /
  spoilage rate, each marked "above median" or "below median"
- Slot-output ranking, lowest first; the bottom product gets a
  dedicated note: swapping it for a median-output product is worth
  how much per day, per month
- Carryover flag: products carried over on at least half the sample
  days listed separately - their low spoilage rate just postpones
  the discard, it doesn't remove it
- Replacement suggestions are only allowed as a half-slot trial:
  the incoming product takes half a slot for [K] days, with the
  three numbers to record daily, the threshold that counts as
  validated, and how to back out if it misses
- Days missing from the sample are simply missing; mark the actual
  sample size in the table

Constraints:
- No "boost promotion" or "increase sales" advice - slot allocation
  is deployment; whether it sells is a different battle
- Do not cite any industry-average margin or "normal spoilage
  rate" - only the ranking my own weeks of data produce
- Flag outlier days (holidays, storms) on their own line, but the
  decision to exclude them is mine - do not drop them on your own
- No pleasantries; start directly with the master table

Why it is written this way

A stall owner's instinct is "stock what sells," and that instinct has two blind spots. First, a revenue lens favors high-ticket, thin-margin products - impressive at the register, ordinary in the pocket. Second, it is blind to the fact that the counter itself is the scarce resource: the stall's biggest costs are rent and commission computed on those few feet of frontage, so a slot's output is its real report card. That is why axis 2 uses revenue per slot rather than per product - "is this slot worth it" becomes a question with a number attached.

Folding spoilage into cost (axis 1) and surfacing spoilage rate (axis 3) are two faces of the same trap: same-day items like cold salads and white-cut chicken often post the highest shelf margin of the whole stall, while the nightly discard never enters the owner's mental math - "looks profitable, nothing left at month-end" is exactly how that feels. The carryover flag closes a third loophole: a low spoilage rate can just mean the discard got postponed to tomorrow, and the product quietly eats reheating loss and reputation while the data looks innocent.

Replacements are restricted to half-slot trials because a few weeks of data cannot support a clean cut - season, the neighbor's stall, weather all make the ranking wobble. The trial's three numbers (what to record, what counts as validated, how to back out) turn a swap from a bet into an experiment that is allowed to fail. Outlier days are flagged but the owner keeps the delete key, for the same reason: it is his data, and the analyzer's job is to show the cost of deleting, not to delete.

Variables