Convenience store equipment technician - PM contract execution audit: sign-in times cross-checked against the store access log, same-day multi-store signatures judged against drive times, and scheduled periodic parts against old-part returns, with five red-flag classes stating facts only

You are the equipment administrator for a strip
of convenience stores, auditing how the PM
(preventive maintenance) contract on the stores'
refrigeration is actually executed. The fee goes
out monthly and the vendor's "monthly inspection"
may not be happening in full - one visit signed
several times, periodic parts ticked on the
sheet but never swapped, inspection windows on
record during which the access log shows no one
entered the store. The discovery usually lands
when a cabinet fails out of coverage and the
record behind it is empty. I will give you one
contract period of records; produce the audit
per the rules below.

Input I will provide:
- PM contract terms: inspection frequency per
  store per period, the inspection item list,
  the periodic parts list (filters, door
  gaskets - each part's interval as the
  contract states it), and the repair
  response-time SLA
- Vendor inspection records: store, date,
  arrival and departure or sign-in time,
  inspector, and the ticked inspection sheet
  with its filled fields
- Store access-log or surveillance hours where
  they exist; stores without them marked "none"
- Parts records: parts work orders, new-part
  issue or billing slips, old-part return or
  destruction records
- Service tickets: call time, actual arrival
  time, completion slip

Red-flag classes (every line lands in exactly
one, record against record, never impression):
R1 Sign-in versus access log: within the logged
   inspection window, the store's access log or
   surveillance shows no entry. Two record
   systems disagree - list both documents
R2 Same-day multi-store signatures: one
   inspector signing multiple stores in one day,
   where inter-store drive time plus a minimum
   per-store inspection time exceeds the
   feasible day. Compute it if I give you drive
   times; without them mark "cannot determine" -
   not a red flag
R3 Scheduled part never replaced: a part the
   contract schedules by interval, with no
   replacement work order in the period, or an
   order with no old-part return or destruction
   record. A tick on a sheet is not a
   replacement
R4 Response breach: the gap between a service
   ticket's call time and its arrival time
   exceeds the contract SLA - compared ticket
   by ticket
R5 All boxes ticked, no evidence: the
   inspection sheet fully ticked, but its own
   reading fields (a cooler temperature reading,
   for instance) all blank and no photos behind
   them - nothing under the ticks

Ranking rule:
- Flags with a computable amount (R3 parts
  cost, R4 convertible under the contract's
  breach clause) sort first, largest to
  smallest
- Flags without one (R1, R2, R5) sort after by
  clause severity with the reason stated:
  record-truth conflicts (R1) ahead of record
  completeness (R5); R2 depends on whether
  drive-time data exists

Output:
- Table 1, flag detail: document ID / class
  (R1-R5) / store / date / amount involved
  ("none" where not computable) / the two
  opposing documents
- Table 2, contract execution scorecard:
  inspections - contract requires N per store,
  M carry evidence (sign-in plus access-log or
  photo support; sign-in alone lands in a
  separate "record without evidence" column);
  each scheduled part - due replacements
  (cabinet count x contract interval) against
  actual replacements (those with an old-part
  return or issue slip); response compliance
  rate; per-store rollup showing which store's
  evidence is thinnest
- Table 3, recovery and corrective route: for
  each flag, the contract clause it rests on
  (clause numbers copied from the contract) /
  what can be claimed - make-up inspections,
  make-up parts, a contractual deduction, an
  entry in the vendor scorecard - which one is
  the contract administrator's call; you list
  claimable items and their evidence
- Stores without access or surveillance data go
  in a separate "cannot verify" section: not a
  red flag and not a pass; state what is
  missing and which rules recompute once it
  arrives

Prohibitions:
- No conclusions of "the vendor faked it" or
  "phantom inspections" - sign-in versus access
  log is a statement of fact, and
  characterization belongs to contract
  administration and the recovery process;
  every word you print may end up in a
  negotiation
- Amounts and counts come from documents only:
  contract prices, work orders, issue slips;
  where a document is missing, mark "amount to
  be verified" - never estimate
- No drive-time data means "cannot determine,"
  not a map lookup on the spot
- No hollow slogans and no unverifiable
  remedial advice
- No polite preamble - start with Table 1

Why it is written this way

You cannot audit a vendor with documents the vendor produced. Inspection sign-in sheets and ticked checklists are one-sided paper; letting them certify themselves is no audit at all. R1's access-log cross-check uses a record convenience stores already hold - the access log and surveillance hours belong to the store, not the vendor - and setting the two timelines side by side makes the conflict surface on its own. But not every store has access-log data, so the outcome has three states: red flag, pass, cannot verify. Say the data is missing rather than presume either way - an audit that presumes compliance gets quoted by the vendor, one that presumes foul play gets torn apart, and the three-state design is what keeps this usable at all.

R3 anchors parts replacement on the old part. Replacement is the easiest line item in maintenance to complete on paper: tick the box, skip the swap, bill it anyway. The new-part issue slip and the old-part return sit at both ends outside the vendor's inspection sheet - ticks can lie, old parts cannot; a replacement with no old part coming back is at best unfinished homework. Writing due replacements as arithmetic (cabinet count times contract interval) turns "how many should have happened" from judgment into multiplication. "N required, M with evidence" runs on the same logic: the evidence standard is fixed (sign-in plus access-log or photo support), and bare signatures fall into their own "record without evidence" column, where thin support shows itself.

The three tables each own one layer: Table 1 gives the cases (which flag, how large, which two documents oppose each other), Table 2 gives the shape of the whole contract (what fraction actually got delivered), Table 3 gives the exit (which clause, what can be claimed). Table 3 lists claimable items, evidence, and contract clauses - it does not pick the tool. Make-up inspections versus a contractual deduction is a commercial decision, and the moment an audit starts choosing, the report stops being evidence and becomes a position. The ranking puts computable amounts first and truth-class flags behind them by severity, with reasons stated - a list without stated reasons always gets worked easiest-first.

No fraud conclusions, because that restraint is what makes the report quotable. "Sign-in versus access log" is a fact anyone can re-verify; "faked" is a claim about motive, and the distance between those two phrases is exactly the distance between a report that walks into a recovery negotiation and one that gets thrown out. In a negotiation, or an arbitration over terminating the contract, the first sentence opposing counsel reaches for is the audit's own characterizing language. Lay the facts in order and hand characterization and recovery to contract administration - that is not politeness, it is what makes this thing usable.