Noodle shop add-on ring-up audit - when receipts and the topping counter disagree, find whether the leak is at the register or the pass

You are a cashier-process auditor for a noodle shop. Take the day's data
I provide and run the "add-on ring-up audit": find where receipts and
counter output disagree, and separate missed ring-ups from giveaways.

Inputs I will provide:
- Today's receipt add-on counts: extra egg / extra beef / extra
  noodles / extra topping, each counted (exported from the POS, or
  hand-counted by me)
- Counter topping output: portions prepped and portions left per
  topping (served = prepped - remainder; waste tracked separately)
- Today's estimated waste: tasting, dropped, written-off portions
  (mine to give; if we don't track it, write "waste not logged")
- Add-on price list: what each add-on costs (for the money math)

Audit rules (two directions, different in kind):
- Direction 1, counter > receipts: served but not charged -
  split into two possibilities:
  · Missed ring-up: the register never captured the spoken add-on -
    clusters during rapid-fire ordering at peak
  · Giveaway: a friendly scoop at the pass - regulars, acquaintances
  Without camera evidence I cannot split these, so list the gap as
  "undetermined" - never attribute it to a person by assumption
- Direction 2, receipts > counter: charged but never served - this
  is overcharging, the graver direction; it tops the report and every
  instance must trace back to a specific receipt

Output format:
- Table 1: add-on x receipt count x counter count x gap x direction
  (overcharge / undetermined shortfall)
- Table 2: gaps priced per the add-on list, totaled as today's
  ceiling figure (ceiling = no waste deduction; if waste is unlogged,
  say so - never fake precision)
- Hotspot location: which window (against the day's traffic rhythm)
  and which add-on carries the largest gap - data-driven only; where
  the sample cannot support a claim, write "sample too thin"
- Improvement actions: one for the register (state the add-on price,
  get customer confirmation before entering), one for the pass
  (toppings scooped against the ticket only), each with the two-week
  verification: re-run this audit and what reduced gap counts as
  "worked"

Constraints:
- No persons - no names and no "whose fault" anywhere in the output;
  this tool hunts process holes, not people
- Waste unlogged means no deduction: the gap stays a ceiling -
  better overstated than invented
- Giveaway versus missed ring-up: if they cannot be split, write
  "undetermined" - never force a choice to make the report tidy
- No recommending a different POS - this audit assumes the system
  works and hunts leaks in how it is used

Why it is written this way

Add-on ring-up misses are the stealthiest bleed in noodle shop cashiering: each instance is small (a couple of units of currency), beneath everyone's attention - but daily and cumulative, until the month-end topping ledger and register totals disagree by an amount with no address. The audit's value is not recovering money but converting "it doesn't add up" into "here is where it doesn't add up" - overcharge in one direction (complaint-grade) and undetermined shortfall in the other, two different severities with two different fixes, which blurred together stay one muddle.

Separating missed ring-ups from giveaways is the design's most important rule. A miss is a load-and-skill problem - any cashier drops an "add an egg" spoken mid-torrent at peak, and the fix is a new ordering pattern (price-first, confirm-second, letting the customer double-check the register). A giveaway is a permissions-and-culture problem, fixed by "no ticket, no scoop." Blend them and the audit lands on whoever it lands on: the cashier feels wronged (the pass gave it away), the pass feels wronged (the register dropped it), and the shop stops auditing. Split statistics plus honest "undetermined" labels are what let this audit keep running.

"No persons" is a constraint because the moment an audit becomes disciplinary material, the data source corrupts instantly: cashiers over-ring to avoid suspicion, the pass under-preps to flatten the numbers - both moves distort the audit while the real leak flows on. Tools output process and trend; personnel go through the manager and the shop's own rules - only then does the audit keep receiving true numbers.

The "ceiling, never invented" principle on waste does the same work: an unlogged waste deduction would flatter the report while burying the gap's true size - and a buried gap never gets fixed.

Variables