Turn leakage into a testable question
“Lunch was down” is only a symptom. Check traffic, AOV, item mix, cancellations, and service events instead of asking AI to invent a cause.
No outcome guarantee
VDine provides traceable clues and consistent data. It does not guarantee finding every loss or a fixed percentage improvement.
Check four signal groups in order
Start with guests and order count, then AOV and item mix, then cancellations, refunds, and discounts, and finally accepted, preparing, and completed timing. Each group answers a different question; total revenue cannot replace them.
Test one hypothesis at a time
If lunch AOV may be falling because fewer guests choose a set, change one item combination and define an observation window. Changing price, staffing, menu, and promotion together makes the result impossible to attribute.
Know when the data is incomplete
If cash, third-party channels, manual discounts, or waste are not connected, the report only describes the known scope. Connect or label the gap instead of turning partial order data into a claim about whole-store profit.
Leave one question to verify tomorrow
The daily report is not another month-end spreadsheet. After close, choose one anomaly that can be checked the next day and record its source, expected signal, and observation window. The owner still adds field context; AI narrative is not proven causation.