Convenience store loss prevention officer - surveillance blind-spot self-audit checklist: checkpoints derived backwards from "what footage would I need if this happened," each camera judged on whether it resolves what that incident requires, failures ranked by incident likelihood and fix cost - recover usable angles before buying hardware

You are a convenience store loss prevention
consultant. I am the store's LP officer running the
quarterly surveillance self-audit for blind spots
and dead cameras. Organize what I give you into a
results-and-repair-ranking table using the
framework below. The checklist is not "inspect the
cameras" - it is "what footage would I need if this
incident happened." Every camera is judged on
whether it resolves what that incident requires.

Input I will provide:
- Camera inventory: ID per channel, mount position,
  approximate angle
- Retention: how many days of recording the drive
  holds, whether it auto-overwrites
- My spot-check playback results: which channels I
  reviewed yesterday and how they looked (faces
  recognizable / human shapes only / color blobs)
- Store layout, sketch or described: relative
  positions of the entrance, tobacco wall, register
  counter, fresh-food prep counter, backroom door,
  and instant-retail picking area

Checkpoint framework (for each incident, fix what
the footage must resolve first, then check whether
it does):
C1 Robbery / snatching: entrance resolves front
   faces; register-front shows actions at the
   counter. This footage exists to identify a
   person - "a person is visible" is not the bar
C2 Carton theft from the tobacco wall: one channel
   on the face, one on the side. The face counts
   how many cartons were taken; the side shows
   which way the person left - front-only coverage
   goes blind two steps after the grab
C3 Counter disputes / counterfeit bills: a close-up
   on the counter that shows the denomination
   handed over and the examine-and-return motions
C4 Fresh-food shrink / sampling out of control:
   overhead of the hot-food and prep counter
   showing hand movements and discard actions
C5 Backroom loss / unauthorized entry: the full
   rear-door corridor, including the outside view
   in the moment the door opens
C6 Instant-retail mispicks: the staging and
   handoff shelf, resolving which order was picked
   by whom
C7 Baseline (checked once across all channels):
   retention days >= [retention requirement] -
   set from this store's slowest discovery path;
   a monthly-count store needs the tobacco wall
   covered for at least a month; host clock synced
   to real time - thirty minutes of clock skew and
   the footage never lines up with the POS journal;
   night-mode illumination actually resolves detail
   (judge at night, not on daylight playback)

Pass criteria (fixed):
- C1-C6 pass only when each channel reaches
  "resolvable" for its own requirement - a face
  identifiable, carton count possible, actions
  distinguishable. "A human shape is visible"
  fails
- Any C7 item failing marks every channel as
  non-compliant for evidence purposes

Output:
- Table 1, results: checkpoint / camera ID / what
  it must resolve / does it (pass / fail / area
  not present in this store) / specific failure
  cause (angle / resolution / retention / clock /
  night)
- Table 2, repair ranking (failures only): higher
  incident likelihood and lower fix cost rank
  first - re-aiming, clock sync, and better
  lighting always precede installing new cameras;
  each row states the fix and its acceptance test
  (which playback to re-check afterward, against
  which standard)
- Table 3, uncovered list: checkpoints with no
  camera at all - this is the basis for a hardware
  request, listed separately from repairs

Prohibitions:
- No brand, model, or price recommendations - the
  ranking deals in angles and likelihood only;
  what to buy is procurement's business
- Do not cite a "required retention days" figure
  from any regulation - retention is set from this
  store's own count and reconciliation cycle, with
  the reasoning stated, not from an invented
  external mandate
- Never judge resolution on channels I have not
  played back - mark them "needs playback" and
  give me the method (pull yesterday at a set
  hour; the ruler is: pin a newspaper two meters
  out and see whether the headline is readable)
- No "comprehensive sweep" or "regular
  maintenance" phrases without a frequency; no
  polite preamble - start with Table 1

Why it is written this way

The checklist works backwards from incidents rather than forward from cameras because the classic failure of a surveillance audit is checking "is it on, is the light green" instead of "will it be enough when it matters." Under camera-inventory thinking, twelve channels all pass; under incident thinking, the tobacco wall has front-only coverage and the side - exactly the direction a thief actually leaves through - is empty. Stating "what this footage must resolve" at each checkpoint turns "usable" from an adjective into an acceptance test: face identifiable, cartons countable. Anything short fails; there is no middle state.

C7 pulls retention days and clock sync out as baseline items because they are the cheapest failures and the most common. Retention is derived from the discovery lag, not a generic number: in a store that counts monthly, the missing cartons can surface three weeks after the theft, and a drive that overwrites at twelve days has already destroyed every frame of evidence. Clock skew is the sneakier failure - the footage exists but no longer lines up with the POS journal, and in a refund dispute a half-hour offset makes the evidence nearly worthless. Both fixes cost almost nothing, which is why the ranking rule hard-codes that zero-cost actions always precede hardware.

The model is forbidden from judging resolution it has not seen because checklists invite fill-in-the-blank compliance. It does not know what each lens actually shows; ticking "pass" on my behalf would be fabricating data. So the input accepts conclusions only from channels the LP officer has played back, and everything else is marked "needs playback" with a method attached - the newspaper test is a resolution ruler that requires no expertise to run.

The uncovered list stays separate from repairs because they are different jobs with different audiences: repairs are the LP officer's own work; gaps are the basis for a budget request. Merge them and they dilute each other - the re-aimable cameras make the situation look less severe, while the genuinely uncovered areas look already handled.

Variables