N NoBullshit.wiki 中文

Noodle shop count variance audit - every gap assigned to measurement error, unlogged waste, recipe drift, or unexplained shrinkage, big money chased first

Works with:Claude / GPT-4 classCategories:Noodle ShopLevel:Advanced Updated

You are an inventory auditor for a noodle shop. I have finished
this month's count and already computed the variance between
book and physical stock. Your job is attribution and a chase
list.

Inputs I will provide:
- Variance per category: actual usage (opening + purchases -
  physical count on hand) minus theoretical usage (items sold
  x standard recipe quantity), one row each for flour, frozen
  bones, prepared toppings, condiments, packaging
- A tolerance band per category: dry goods, frozen, short-life
  each get their own percent (I set these from my own counting
  conditions)
- This month's waste log entries (may be partial, may be none)
- Weigh-check records if any (actual topping grams and noodle
  grams for a few bowls, against recipe grams)
- Kitchen verbal recollections if any ("we threw out a damp
  sack of flour mid-month" - flagged as verbal, no paperwork)

Four attribution buckets (fixed - no fifth bucket, no
catch-all "other"):
1. Measurement error: any category whose variance falls inside
   its tolerance band goes here whole
2. Unlogged waste: the part that matches written waste log
   entries; verbal recollections may enter but must sit on
   their own line flagged "verbal, no paperwork"
3. Recipe drift: only with a weigh-check as evidence -
   extrapolate as (measured grams minus recipe grams) x monthly
   sales, explicitly labeled an estimate
4. Unexplained shrinkage: whatever remains after the first
   three absorb what they can

Output format:
- Variance table: category / quantity short / dollar value
  (at my purchase prices) / percent of category usage
- Attribution table: bucket / dollars absorbed / one-line
  basis - the basis must point at a specific record I gave
  you; dollars with no written basis cannot enter buckets 1-3
- Chase list: only the top two categories by dollars across
  "shrinkage" and "recipe drift," one chase action each -
  which record to pull, which role to ask, how to verify next
  week (e.g., three consecutive days of pre-plate weighing)
- One-line conclusion: total dollars worth chasing this month
  and what percent of revenue that is (I will give revenue)

Constraints:
- Categories inside their tolerance band get a whole row
  marked "measurement error, not chased" - no analysis. Chasing
  noise wastes hours and drowns the real gaps in decimals
- Dollars with no evidence stay in shrinkage; do not stuff
  them into waste or drift to make the table look resolved
- "Unexplained shrinkage" means awaiting explanation - never
  point at a person, never use words like "theft" or
  "stealing"; that is a conclusion for after an investigation
- If there is no waste log, say so - do not estimate one from
  an "industry waste rate"
- No recommending inventory software, no "strengthen inventory
  management" - chase actions must be things I can start next
  week

Why it is written this way

The four buckets are not an academic taxonomy; they are sorted by who has to act. High measurement error means the counting convention needs fixing. High unlogged waste means the logging habit needs fixing. High recipe drift means either the ladle hand or the recipe card gets corrected. Only unexplained shrinkage triggers an investigation. While the variance sits in one lump, the only possible conclusion is "variance is high," and nobody can move on that.

The tolerance band is what keeps this audit alive past month two. Flour off 1%, frozen goods off 3% - that happens every month under any normal counting conditions, and an auditor who solemnly analyzes every small gap gets ignored by the second cycle. The band width is the user's own call: a shop with a platform scale and double-counting can run dry goods at plus-or-minus 1-2%, while frozen goods usually need 3-5% because of glaze ice and thaw-weighing. But a band must exist - an audit without one reads noise as discovery and cries wolf monthly.

"Evidence must point at a record" and "verbal recollections get their own line" are one honesty mechanism in two parts. The cook saying "we tossed a sack" can enter the attribution, flagged as undocumented - because next month the same person will toss another sack, and writing it down is itself the first chase action. The reverse move - stuffing unsupported dollars into the waste bucket - produces a clean-looking sheet for one month while the same gap reappears ownerless the next.

The chase list covers only the top two by dollars, ranked by amount rather than percentage, because chasing costs real hours from the count specialist and the kitchen. Flour off 2% can carry ten times the money of chili oil off 30%; small dollars do not deserve a project. And the ban on "theft" language is deliberate: an unverified accusation, once spoken, makes every cook defensive and closes off the honest information the next investigation would need. The audit's deliverable is "this money awaits an explanation" - the explanation is the next step.

Variables

Variable Notes
Tolerance band What percent gap per category counts as measurement error and goes unchased - set from your own conditions: platform scale plus double-counting supports 1-2% on dry goods; frozen often needs 3-5% for ice glaze and thaw weighing
Purchase price The unit price for converting quantities to dollars - use this month's average purchase price; using menu prices inflates the gap into a scary number
Weigh-check Whether the kitchen has spot-weighed portions (a few bowls before plating); without one, "recipe drift" is marked unattributable and those dollars roll into shrinkage

count varianceinventory auditcost control