The setup: a 60-seat trattoria in Brooklyn. Italian-American, BYOB-adjacent, lunch and dinner, six days a week. Owner-operator who is also the chef. Pre-ORBIS: prime cost trending 67 to 69 percent, owner working 70 hours a week, no time to audit invoices, three license renewals missed in the previous year (caught last minute, paid late fees). The restaurant joined the 90-day pilot in March 2026. What follows is what we saw, in the order it happened, with the numbers that matter.
One thing to say up front. This is a case study, not a customer testimonial. The owner asked to stay anonymous and we agreed, because the Brooklyn dining market is small enough that a postal code plus a cuisine plus a seat count is already most of an identifier. So there are no quotes here, no name, no photo. There are real numbers, real dates relative to onboarding, and a real arc. If that costs the post some of the warm-and-fuzzy that customer testimonials normally bring, fine. The numbers carry the weight.
The starting line.
Snapshot of the restaurant on Day 0, taken from a 12-week trailing window:
- Annual revenue: about $2.0M
- Prime cost: 68.4 percent
- Food cost: 36.1 percent (target 30 percent)
- Labor cost: 32.3 percent (target 28 percent)
- Owner-operator hours: 70 per week
- Software stack: Toast POS, 7shifts for scheduling, Square for books, no invoice tool
At intake the owner listed three pain points without hesitation: supplier overcharges he suspected but never proved; scheduling chaos every time staff called out; never knowing his actual margin until six weeks after month-end. The room was full most nights. The math was leaking anyway.
What ORBIS did, week by week.
Weeks 1-2: Onboarding plus baseline.
Eight questions in WhatsApp, Gmail and Toast connected, and ORBIS spent two weeks reading invoices to learn the baseline. No alerts yet, by design. First labor schedule drafted in week 2; the owner reviewed it, accepted with two voice edits, and moved on.
Weeks 3-6: First overcharge catches.
ORBIS started flagging discrepancies. Week 4: a Sysco invoice with romaine up 14 percent against the 8-week baseline; claim filed, $94 credit landed week 6. Week 5: duplicate billing on a Baldor delivery, $212 credit. Week 6: a spec swap on imported tomato (16 oz to 14 oz at the same line price); vendor agreed to credit forward. Total caught in four weeks: $618. The owner's exact reaction, paraphrased so we are not putting a quote in his mouth, was that he had suspected the romaine line for about a year and never had a way to prove it. Now he had a receipt.
Weeks 7-9: Margin tightens.
Food cost dropped from 36.1 percent to 33.8 percent as the supplier-side leakage closed and the owner started taking purchasing recommendations. The biggest one: sub the imported San Marzano for a closer alternative on the pasta dishes. Blind tasting passed in the kitchen. Savings: about $0.42 per portion, which on a dish that runs ~180 covers a week is real money inside of a month. Same period: ORBIS started catching prep-side overproduction on the lasagna line, and the kitchen rebalanced the batch size from 18 portions to 14. Smaller delta, same direction.
Weeks 10-13: Schedule plus compliance autopilot.
By week 10 ORBIS was drafting Saturday and Sunday schedules autonomously. The owner spent about 12 minutes a week reviewing, mostly to swap one shift around an FOH manager's class schedule. Compliance side: the liquor license renewal fired 90/60/30 alerts; paperwork started in week 11; license renewed without a fine for the first time in two years. Health permit and FDNY assembly cert both got logged with future-dated alerts so the same near-miss does not happen next cycle.
90-day results.
| Metric | Day 0 (12-wk trailing) | Day 90 (12-wk trailing) | Change |
|---|---|---|---|
| Prime cost | 68.4% | 63.1% | -5.3 pp |
| Food cost % | 36.1% | 32.4% | -3.7 pp |
| Labor cost % | 32.3% | 30.7% | -1.6 pp |
| Supplier disputes filed | 0 | 7 | +$2,840 credits |
| Owner hours / week | 70 | 56 | -14 hrs |
| Late license fines | $640 (prior year) | $0 | -$640 |
Convert the changes to dollar terms and the picture is concrete. A 5.3 percentage-point move on $2.0M in revenue is roughly $106,000 a year in gross margin. Add $2,840 in dispute credits inside one quarter (annualizes to about $11,000). Then the 14 hours a week the owner got back, roughly 700 hours a year, went into the wine list and the early scoping of a private-dining program. The pilot did not pay for itself once. It paid for itself in three different ways at the same time.
What did not change.
Honest section, because a case study without one is marketing. Revenue stayed roughly flat at $2.0M; the pilot was not a growth tool, it was a cost tool, and that is what it did. Staff turnover stayed at 15 percent for the quarter; ORBIS does not fix the labor market. The owner still works weekends, because he wants to; that is his call. And one specific line item (cheese) stayed high: the supplier is the only one in his network delivering the quality he wants, so ORBIS flagged the variance and the owner chose to absorb it. Not every flag becomes a fix. That is what credibility looks like.
What is next.
The pilot ended in week 13. The restaurant moved to the Base plan plus the Reviews add-on and ORBIS kept going. The next milestone the owner is watching is the cheese line: when a credible alternative supplier shows up in the city, ORBIS will flag it, and that is the next half-point of food cost waiting to be claimed. See how the pilot works →
Pilot anonymized at the restaurant's request. Numbers are real but date ranges and identifying details have been adjusted to protect the restaurant's identity in a competitive Brooklyn dining market.