Research methodology

These definitions match how Growth Insights calculates figures inside the product. Public research will use the same formulas on anonymised aggregates — never a second invented ledger.

Food cost %

Food cost % = food COGS ÷ food sales × 100 for a closed date range. Prefer recipe cost of paid items + logged waste cost as COGS when recipes exist. Periods with no paid sales are marked insufficient data — not zero.

Inventory variance

Theoretical consumption = sum of recipe quantities × paid bill lines (cancelled lines skipped). Physical movement = receives − waste − adjustments ± counts. Material variance is flagged when the gap exceeds a configured ratio of expected qty (product default documented in Insights constants). Language stays neutral: unexplained variance — not theft accusations.

Waste

Waste cost = quantity × unit cost for logged entries, grouped by reason (expired, prep, plate, overproduction, etc.). Insights ranks top ingredients by cost in the selected range.

Sample rules for public reports

  • Minimum participating restaurants and days before a national figure is shown
  • Outliers winsorised or reported as bands, not single “averages” that mislead
  • Clear N, date range, and geography on every chart
  • If N is too small, we publish methodology only — not a fake chart

Relation to AI

Optional Ask may explain a shop’s own pack in plain language. Public research numbers are calculated aggregates, not language-model guesses.

Related