— Eight agents · one operator OS

The agents that name the leak.

Each one reads a slice of your operation — sales, labor, voids, 3P fees, tips, catering, shift sentiment — and tells you the one thing to fix. Per store. Per name. Every figure source-tagged.

01 · Voids

Void Hunter

For the COO · Owner

One name above the peer band.

— What it catches

  • Voids vs each store's own peer median (not industry benchmark)
  • Per-store + per-name void rate, flagged if > 1.5× peer median
  • Annualized excess-vs-peer dollar amount
  • Pattern detection — not verdict

— Data it needs

Employee performance CSV (Toast / Square / Clover / PDQ) with Location, Employee, Net Sales, Void Amount columns. Or a live POS connection.

— What you'll see

KPI strip · per-store table sorted by void rate · top-15 names sorted by void $.

02 · 7 signals

Leak Detector

For the COO · CFO · Owner

Seven theft signals, ticket-level.

— What it catches

  • Void after payment — paid then voided. The classic skim.
  • Cash-only voiders — ≥80% voids on cash tender, ≥5 voids total
  • Comp abuse — above peer band or > 10% of own revenue comped
  • Promo stacking — two or more discounts on a single ticket
  • Discount after close — discount applied after ticket closed
  • Day-of-week patterns — ≥40% of voids cluster on one weekday
  • Micro-comp pattern — 10+ comps averaging under $5 each ("no charge add bacon")

— Data it needs

Ticket-level CSV (Toast Sales Detail / Square Transactions / Clover Reports) with Location, Employee, Ticket Total, Tender, Void/Comp/Discount columns. Optional Closed-At timestamp enables the day-of-week pattern detector.

— What you'll see

Seven color-coded signal cards · top-20 employee table sorted by composite risk score (red ≥ 50, orange ≥ 20).

03 · Labor

Labor Leak

For the COO · Manager

OT drift before payroll closes.

— What it catches

  • Overtime drift per employee · ranked by minutes
  • Ghost shifts · clocked ≥ 60 min with zero sales attached
  • Early clock-ins · >5 min before scheduled start
  • Late clock-outs · >15 min after scheduled end
  • OT $ estimate · per-shift wage × 1.5, summed network-wide
  • Schedule-vs-actual gap concentration by name

— Data it needs

Timesheet CSV with Location, Employee, Scheduled Start, Scheduled End, Clock In, Clock Out timestamps. Optional Net Sales (for ghost-shift detection), Wage Rate (for accurate $ drift; defaults to $15/hr).

— What you'll see

KPI strip · per-employee table sorted by OT minutes · ghost-shift list with start times · total drift $ at OT rate.

04 · Tips

Tip Variance

For the CFO · Manager · Crew

Service slipping shows up here first.

— What it catches

  • Per-name tip rate (tips / net sales) computed week-over-week
  • Flag every server with WoW delta < −2pp on >$50 in tips
  • Network WoW total delta · spot fleet-wide service drops
  • Auto-buckets daily exports into ISO weeks

— Data it needs

Tip summary CSV with Location, Employee, Week-or-Date, Net Sales, Net Tips columns. Toast Payouts / Square Team tips / Lightspeed Shifts all work.

— What you'll see

Network WoW % · per-employee table sorted by most-negative delta · prev-rate vs curr-rate vs delta (pp) vs prev-$ vs curr-$.

05 · Catering

Catering Leak

For the Chef · CFO · Owner

Where the order ran but the receipt didn't.

— What it catches

  • Total reconciliation gap (invoice − POS) per store
  • Unmatched orders · invoice > 0 with no POS match
  • Flagged orders · gap > $50 AND > 10% of invoice
  • Customer concentration · who owns the biggest piece of the leak
  • Network gap ratio · gap / total catering invoice

— Data it needs

Catering CSV with Location, Customer, Invoice Amount, POS Amount columns. Optional Order ID + Event Date for the unmatched / flagged order list.

— What you'll see

KPI strip · per-store table sorted by total gap · unmatched orders list · flagged orders list · top customer concentration.

06 · BCS

Beverage Cost Score

For the Owner · Chef · Bar Manager

Pours vs inventory. One score 0-100.

— What it catches

  • Per-store BCS 0-100 with severity coloring
  • Per-category shrink · liquor · beer · wine · NA
  • Revenue lost (shrink units × unit price)
  • Network rollup score · sanity check across the fleet
  • The free outreach hook · one CSV → one page handed back

— Data it needs

Beverage CSV with Location, Category, Consumed (inventory delta), Poured (from POS). Optional Unit Price for revenue-lost calculation.

— What you'll see

KPI strip · per-store table with BCS score · per-category breakdown with shrink units + %.

07 · Drift

Vendor Drift Detector

For the Chef · CFO · Owner

Silent price creep, caught.

— What it catches

  • Per-SKU price drift % vs prior period
  • Flag every SKU with > 5% upward drift
  • Per-vendor rollup · who is creeping the most
  • Auto-buckets daily / weekly exports into months
  • Dollar amount of network-wide upward drift

— Data it needs

Vendor invoice CSV with Vendor, SKU, Period (or Invoice Date), and Unit Price. Optional Category for category-level rollups.

— What you'll see

Per-SKU table sorted by drift % · vendor rollup · total upward drift $.

— The rule that runs the agents

Every figure ships source-tagged.

Verified

We can re-pull this from a primary source and defend it to the penny.

Estimated

Modeled from a benchmark or assumption. We name the assumption next to the number.

Unverified

Source not wired yet. Number is illustrative — operator-only.

No competitor in this category publicly source-tags figures. None publicly disclose model error. We checked. Read the case where we caught our own $8.3M number, walked it down to $1.81M in writing, and made source-tagging the operational rule that came out of it.

— Try one. Right now.

60 minutes. Your real numbers.

Drop a Toast / Square / Clover / PDQ export at /trial. Void Hunter and the Leak Detector run on your real data in 30 seconds. No card. No human in the loop.