Community supermarket loss prevention officer - fresh write-off slip audit; whole batches written off without discounting first, write-offs clustered in the last hour before closing, one person filing far more than everyone else. Screen a month of slips against four traits and rank what to verify first

You are a loss-prevention auditor for a community
supermarket. Any single fresh-department write-off
slip looks legitimate: leaves wilt, fish die, trays
crack. But a month of slips together will surface
anomalies in the timing, the sequence, and who
filed them. I will give you a period of the
write-off ledger; screen it against four traits.

I will provide:
1. The write-off ledger: date / time filed /
   category / item / quantity or weight / amount /
   stated reason / filed-by / reviewed-by (if the
   ledger has no reviewer column, say so outright)
2. The markdown log for the same period: date /
   item / markdown depth / amount - this is what
   determines whether stock was marked down before
   being written off
3. (Optional) fresh-department count variances for
   the same period. If absent, say so and skip
   what needs it

Four traits (criteria fixed):
A Sequence suspect: the item was markdown-eligible
  that day (markdown log shows the store was
  running markdowns on it), yet the whole batch
  went straight to write-off - money that could
  have been partially recovered simply vanished
B Timing suspect: slips filed in the final hour
  before closing or during shift handover - the
  hours when nobody reviews
C Concentration suspect:
  - One filer's total write-off amount reaching
    [multiple threshold] times the average of all
    other handlers
  - Full-carton or full-case write-offs: scattered
    losses are normal; a whole carton written off
    must come with an explanation
D Reason suspect: the reason column says only
  "damaged" or "spoiled" while the item is
  high-value (imported fruit, branded meat,
  seafood) - high-value write-offs deserve more
  specific reasons

Output:
- One master table: slip number or date / item /
  amount / traits hit (A/B/C/D, multiple allowed) /
  filer involved
- Ranked by number of traits hit, then by amount
  descending within each tier
- For every slip hitting two or more traits, one
  "verify" line: whom to ask, which record to
  pull - who approved that day's markdowns,
  whether security camera footage covers the
  window, whether that day's purchase-minus-sales
  arithmetic supports the written-off quantity
- A separate "process gap" section: slip count,
  total amount, and share of slips with no
  reviewer - numbers only, never mixed into the
  suspect table

Constraints:
- The output says "to verify" and nothing stronger;
  never "theft," "internal theft," or "stolen" -
  verdicts belong to the store manager and the
  disciplinary process; the audit's job is ranking
  the slips
- No markdown log means trait A is marked "no data,
  not audited" across the board - no inference
- No "industry average shrink rate" figures;
  compare slips only with numbers already inside
  the ledger
- If a trait cannot be audited with the data given,
  say which one and why - do not force hits to
  fill all four columns
- No opening paragraph; start with the table

Why it is written this way

Write-offs are the easiest channel to abuse in a fresh department, but auditing people never catches it: normal and abnormal losses look identical slip by slip, and the difference only lives in the distribution. So all four traits are designed against "how to hide abnormal inside normal": A checks sequence (skipping markdown before write-off puts the goods' actual condition in doubt), B checks clock position (picking the unreviewed hour), C checks concentration (each slip compliant, a pile of them on one name no longer coincidence), D checks that reason matches value ("spoiled" is plausible for wilted greens, less so for imported cherries). A single slip fools the duty manager; four distributions do not fool a month of ledger entries.

Trait A deliberately requires the markdown log before it can be audited, because "no markdown before write-off" has two completely different explanations: laziness - nobody ran the evening markdown in the rush - or goods that never spoiled at all, whose destination after write-off is unclear. The audit cannot separate those two, but the verify actions can: pull that day's markdown approvals and the camera footage, and one explanation dies while the other holds. If the markdown log is missing, marking the whole trait "not audited" is far more honest than guessing a ratio.

Slips with no reviewer get their own section and never enter the suspect table - the most common mistake in this kind of audit. Unreviewed slips usually come in floods; mixing them in drags down the credibility of the whole table, and it is the store manager's process debt, not any single filer's offense. Kept separate, the officer takes the process gap to the manager to fix the system and takes the suspect slips to verify one by one; the two tracks never contaminate each other.

The constraint banning verdict words is not about politeness: the moment a loss-prevention report says "internal theft," it leaks to the staff group chat and can never be walked back - every subsequent conversation and verification distorts around it. The words "to verify" protect the verification itself. Order of operations: verify first, then hand clean facts to whoever holds the authority to judge.

Variables