Inventory reconciliation for restaurants

Know why your inventory doesn't match.

StockSight connects what your POS sold, what your recipes should have used, and what's actually on the shelf — then shows you where the difference comes from.

Get early accessSee how it works

The gap

Three numbers that rarely agree.

Restaurants already track sales. What's missing is the layer that turns sales into expected ingredient usage and checks it against a real count. StockSight is that reconciliation layer.

What the POS says

400

burgers sold

Your POS knows the menu item. It doesn't know the bun.

What recipes expect

100

buns left

500 at the start of the week, minus 400 used by sales.

What's on the shelf

65

buns counted

The physical count at close on Sunday.

Variance

−35

buns · −35%

Waste? A miscount? A recipe that's off? StockSight helps you find out.

How it works

From sales to variance in four steps.

No rip-and-replace. StockSight sits beside the POS you already use and does the math your spreadsheet can't keep up with.

01

Map menu items to recipes

Tell StockSight what goes into each dish, once. Sizes, modifiers and per-location recipes are supported in the model.

Cheeseburger

menu item
  • Burger bun1 unit
  • Beef patty1 unit
  • Cheese slice1 unit
  • Lettuce20 g
  • Tomato15 g
  • Sauce10 mL

02

Import POS sales

Upload an export or connect your POS. Raw data is stored as received, and re-importing the same file never double-counts.

CSV

shift4_sales_oct06-12.csv

100 sold

03

Count on a phone

Staff walk the store and count item by item. Each submission is timestamped and attributed, and history is never overwritten.

Walk-in→Freezer→Dry storage→Bar→Prep

04

Reconcile

Expected usage is calculated from every sale and compared with what was counted, per item, per location.

Burger bun

100

Beef patty

100

Cheese slice

100

Lettuce

2 kg

Tomato

1.5 kg

Sauce

1 L

The reconciliation, in one line

expected = starting count + deliveries − (sales × recipes) − recorded waste ± adjustments

variance = actual − expected

Neutral by design

Flags the gap. Never points fingers.

Variance has many causes, and most are operational. StockSight separates what's documented from what isn't and gives you the context to investigate. The cause is your call.

Every variance, three ways

Quantity

−35

buns

Percent

−35.0%

of expected

Value

−$22.75

at unit cost

$ unknown% n/aMissing cost or zero expected is shown as unknown, never as $0 or 0%.

  • Waste

    Expired, spoiled, damaged, prep errors, overproduction, customer returns. Logged with a reason, so it never looks like missing stock.

  • Adjustments

    A corrected count, a supplier discrepancy, a transfer between stores. Made by managers, kept in the audit trail.

  • Over-ordering

    Deliveries that consistently outrun real usage, like ordering 100 kg when you use about 60.

  • Unexplained variance

    The gap that's left once everything documented is accounted for. Repeated patterns get surfaced, never assumed.

  • Potential shrinkage

    The umbrella view of stock that can't be accounted for: miscounts, recipe drift, unrecorded meals and more.

Mobile counting

Built for a phone, a clipboard hand, and a walk-in freezer.

Big tap targets, a numeric keypad, and one item per screen. Counts follow the same physical route every time, so nothing gets skipped between the walk-in and the bar.

  • Partial units

    0.5 case or 1.35 kg, stored exactly as counted.

  • Any schedule

    Daily, weekly or biweekly. The dashboard always shows how old the last count is.

  • Every change is kept

    Counts are append-only. Corrections are new entries, not overwrites.

  • AI as a shortcut

    Photo or voice input is a suggestion you confirm. Manual entry always works.

POS

Keep your POS. We'll read from it.

Every connector converts sales into one standard model, so refunds, voids, comps and employee meals are handled the same way whichever system you run.

Shift4

CSV / Excel upload

Clover

Sheets export

Square

Planned

Toast

Planned

  1. Your POS→
  2. Raw data, kept as received→
  3. Normalized→
  4. Standard sales model→
  5. Reconciliation

Imports are idempotent. Rows that fail validation are reported back to you, never silently dropped.

Franchises & groups · Coming later

Every location, one view.

For operators running 2 to 20 stores. Compare variance across locations, spot the store that needs attention, and drill into any single store's counts. Owners see everything; managers and staff see their own location.

Unexplained variance, % of expected usageLast count
  • Downtown4.8%6 h ago
  • Westside2.1%1 day ago
  • Airport7.4%3 days ago
  • Northgate1.3%6 h ago

Where we're going

From counting stock to understanding it.

  1. 01Now

    Track

    What inventory do I have?

  2. 02Now

    Reconcile

    What inventory should I have?

  3. 03Next

    Explain

    Why is it different?

  4. 04Later

    Predict

    What should I order?

FAQ

Questions operators ask us.

Do I have to replace my POS?+

No. StockSight reads sales from the POS you already use, through a file upload or a connector, and never writes back to it.

How often do we need to count?+

Whatever fits your operation: daily, a few times a week, weekly or biweekly. More frequent counts give finer detail, and the dashboard always tells you how old the latest count is.

What about shifts that run past midnight?+

Each location has its own timezone and business-day cutoff (4 AM by default), so a 1 AM sale lands on the right business day.

Does it depend on AI?+

No. Photo and voice input are optional shortcuts. Anything AI suggests is confirmed by a person before it touches your inventory, and manual entry always works.

Who can change inventory numbers?+

Staff count and record waste. Adjustments need a manager or owner, and every change is kept in an audit trail rather than overwritten.

Find the variance you can't see today.

We're onboarding a small group of independent restaurants and franchise operators. Tell us a little about your setup and we'll be in touch.