Each topic below has its own section, an original illustration, and sample metrics for Corner Grill. Live connections are marked as roadmap.
Sample Live heatmaps on the roadmap
1. Employee movement — workflow, not surveillance
EVA is designed to watch stations and paths so a manager can see whether the line is covered during a rush. It is not built as a tool to score individual workers by face. Facial recognition is off by default and is not shown in this demo.
During a rush, the useful question is usually “is the station staffed?” not “where is this person?” EVA’s movement view is planned around work areas:
- Station coverage heatmaps. Warm color builds where tickets, plates, and foot traffic concentrate — typically the pass, the grill, or a drinks well. Cooler color shows quieter prep or dish. The map is about the floor plan, not a named face.
- Time away from station. If the grill is empty while tickets are on the board, the demo can show minutes the station was uncovered in an hour. That is a coverage clock, not a break-room camera (break rooms and restrooms are not EVA camera locations).
- Bottleneck spots. When plates stack at the pass or runners cross the same aisle, the map marks a bottleneck so a manager can add an expo, move a trash can, or change a firing pattern.
- Steps per order. Path traces can estimate how far a ticket travels from prep to the pass. The sample figure of 48 steps is a planning estimate from those traces, not a pedometer on a person and not a productivity score.
- Idle vs busy. A station can look idle because the book is light, or busy because one person is covering two wells. Side-by-side busy percentages help a manager rebalance the next rush, not dock a paycheck.
- Coverage gaps during rush. A short gap — eight minutes with no expo between 8:10 and 8:18 PM in the sample — is the kind of hole that shows up later as slow-service reviews.
What is in the demo today is a sample heatmap illustration and the metrics below. Continuous, live movement maps from the cameras are on the roadmap. EVA does not use facial recognition by default; if a future option is ever offered, it would be a disclosed, off-by-default setting, not the standard workflow view.
Sample data Figures are labeled samples for the fictional Corner Grill location. They are not measured results from a live restaurant and not a promise of performance.
Sample KDS connection on the roadmap
2. Ticket times — fire to pass, then pass to table
A plate can be late because it cooked slowly, or because it sat at the pass. EVA is designed to time both legs of the trip and to sit next to the fire stamp from a kitchen display system.
Each dish and each station can carry its own clock:
- Fire-to-pass. When a ticket is fired on the line, the clock starts. When the finished plate is set on the pass (the pass camera sees it), the clock stops. In the sample, the average is 11 minutes 20 seconds against a 12-minute target. Ticket 148 in the live demo is the same idea in a single alert: fired at 10:33 PM PT, still waiting at 10:42, a 9:04 plate wait.
- Pass-to-table. A second clock starts when the plate is ready and stops when a runner takes it, or when a table-status change says the food has landed. The sample average is 3 minutes 40 seconds. This is how a manager separates a slow grill from a slow run.
- Late-ticket alerts. Tickets that cross a threshold — 15 minutes in the sample — rise as a watch or an alert on the EVA tile, so the expo does not have to stare at every chit. The demo already shows one late plate at the pass.
- Rush breakdowns. Times can be sliced by hour, station, and dish. The sample 7–9 PM rush shows grill fire-to-pass at 13:10, sauté at 10:40, and expo pass-to-hand at 4:05. That points at the grill, not “the kitchen” as a blob.
Kitchen display data, including Toast KDS as an example
Fire times are most accurate when they come from the kitchen display the line already uses, not from a camera guessing when a sauté pan moved. EVA is planned to read ticket-fire timestamps from a kitchen display system. If a restaurant already runs Toast KDS — or another KDS that can share fire times — those stamps can be matched to pass-camera timing and to the ticket number on the rail.
That connection is on the roadmap and is shown here with sample data only. This demo is not a live hookup to Toast, it is not an official Toast product, and it does not imply a partnership with Toast. “Toast KDS” is named as an example of a kitchen display many independent restaurants already have. Other kitchen displays with a usable fire-time feed are in the same planned design.
Sample data Figures are labeled samples for the fictional Corner Grill location. They are not measured results from a live restaurant and not a promise of performance.
Sample Live POS timestamps on the roadmap
3. Table turnover — seated, ordered, food, check, bussed
A table earns its keep only when it is ready for the next party. EVA is designed to stamp each stage of a turn and to flag a table that is empty but still dirty while a waitlist is standing.
- The five stamps. Seated when guests land, ordered when the first ticket fires, food when plates leave the pass or land on the table, check when the bill is opened or settled, and bussed when the table is reset. Together they show whether a long turn was in the kitchen, at the table, or in the aisle with a dirty two-top.
- Average turn time per section and server. The sample has Section A at 48 minutes and Section B at 61 minutes. Server averages in the sample — Maria at 46 minutes, Jake at 58 minutes — come from the section and server assignment on the ticket, not from facial recognition on the floor.
- Empty, dirty-table alerts. Table 12 in the sample has been empty for 9 minutes and is still not bussed. That alert is meant for the floor, the same way a late plate alert is meant for the pass.
- Waitlist impact. If four parties are waiting and two tables are dirty, the hold-up is not “we need more seats” — it is a reset problem. EVA is designed to put those two numbers on the same card.
- Revenue per seat-hour. A simple planning ratio: net sales in the window, divided by seats, divided by hours. The sample is $14.20. That figure is an example calculation for Corner Grill, not a measured result and not a promise of higher revenue.
Joining camera table-status with live point-of-sale seated, ordered, check, and closed stamps is on the roadmap. The demo shows the sample metrics and the dining-room illustration so you can see the shape of the report today.
Sample data Figures are labeled samples for the fictional Corner Grill location. They are not measured results from a live restaurant and not a promise of performance.
Sample Pour vision and auto-matching on the roadmap
4. Inventory and theft signals — variance, doors, and an evidence report
Missing product is usually a puzzle with several pieces: a count, a sale, a door, a void, a delivery, or a drawer. EVA is designed to put a short clip and a timestamp next to that puzzle so a manager can review it. A clip is a signal, not an accusation, and nothing in the demo auto-disciplines anyone.
- Shelf and walk-in counts vs sales (variance). A snapshot count is compared with what was sold. In the sample, chicken is two cases short of what sales and the last count would predict. Oil and chicken in the 9:05 PM PT walk-in snapshot are already under par on the Inventory tile — the same short items, two views.
- Back-door and after-hours activity. A rear exit held open, or a door after close, raises an alert. The live demo already includes a propped back door (6 minutes at 10:36 PM PT in the evening sample) and this page uses an after-hours still at 11:14 PM PT so you can see the night case.
- Bar pours vs rings. A camera over the well can be trained to count pours and compare them with items rung on the ticket. The sample shows 8 pours and 6 rings on well vodka between 8 and 9 PM. Pour-level computer vision is on the roadmap; the figure here is a sample of the report, not a live pour counter.
- Voids and comps matched to clips. When a manager comps or voids a ticket, EVA is designed to attach the pass or register clip from that minute. Ticket 144 in the sample has a void with a clip on file. Matching is planned; the demo shows the shape of the row, not a live POS void feed.
- Deliveries vs invoice shortages. If two cases of oil are on the invoice and the walk-in count after delivery is short, the receiving door clip from that window is the first place to look — a short shipment, a miscount, or product that left the dock. The sample shortage is two cases of oil.
- Cash-drawer events. A no-sale drawer open, or a drawer after close, can be lined up with the register camera. The sample notes a no-sale open at 10:51 PM PT. This is a review cue for the manager, not an automatic theft finding.
Sample data Figures are labeled samples for the fictional Corner Grill location. They are not measured results from a live restaurant and not a promise of performance.
Sample manager evidence report
Sample report Built so a manager can review a clip with a timestamp. This is not from a real restaurant and it is not a finding of theft.
- Location
- Corner Grill (fictional demo site)
- Event
- Rear exit door open after hours
- Camera
- Back door / rear hallway
- Timestamp
- Wednesday, 11:14 PM Pacific Time (America/Los_Angeles)
- Clip
- Sample still, 18 seconds · labeled “not a live feed”
- Related counts
- Walk-in chicken variance −2 cases (sample) · no matching late invoice in the window
- Related POS
- No ticket, void, or paid-out in the same minute (sample)
- Access
- Owners and designated managers only
- Manager review
- Open — EVA does not close this as theft. A person still decides what, if anything, happened.
Hardware and privacy, in short
EVA uses compact cameras on PoE or Wi-Fi, with edge processing and night vision on dim doors and walk-ins. Work areas only — not restrooms or break rooms. Facial recognition is off by default. Notices, access limits, retention, and audio consent (including in all-party states such as California) are covered in more detail on the EVA cameras page. That discussion is general information, not legal advice.