Your POS already knows which menu items are losing you money and which service windows run short on staff. Most operators never look. The data problem in most restaurants is not too little data. It is that the data exists and nobody is reading it.
The five health metrics from The Five Restaurant Metrics That Tell You Whether the Business Is Actually Healthy are the headline. The POS data is where the detail lives. A food cost percentage above target is the headline; the item-level sales data tells you which dish is producing it.
Most operators look at their POS for end-of-night totals. The same system holds three months of item-level detail, server performance data, and timing patterns. That data would change several decisions currently being made by intuition. The POS Decision Method is how to get that data into use.
This article is for the operator who has never explored beyond end-of-night totals. The reports are already there.
Why POS Data Sits Unused in Most Restaurants
The obvious reason is that nobody asked for it. The POS interface is designed around taking orders, not producing analytics. Most operators set it up to run orders and never explore the reporting menu.
The less visible reason is that the reports, when found, feel overwhelming. A standard POS analytics interface produces more data than is useful. The question is not “what does the POS know?” but “what three questions would change a decision I am making right now?”
The deepest reason is that POS data answers operational questions, not strategic ones. The POS is not useful for deciding what kind of restaurant to become. It is very useful for deciding which items to cut, when to add a server, and which service window has a waste problem.
The POS Decision Method
The menu performance read surfaces which items to keep, cut, or reprice
Item-level sales data shows how many times each dish was ordered and what it contributed to total revenue. Sorting by volume and revenue reveals three categories: items that sell and contribute, items that sell but do not contribute, and items that neither sell nor contribute.
Items in the third category are candidates for immediate removal. They occupy menu space, require prep, and produce neither volume nor margin. The data makes the decision objective rather than based on how much the chef likes making the dish.
Items in the second category (high volume, low margin) require a decision. Raise the price, adjust the recipe, or accept that the item drives traffic that orders other things. Item-level sales data makes that analysis possible.
The timing pattern read reveals where volume and staffing are misaligned
Most POS systems track covers by hour across a week. The resulting pattern shows which service windows are consistently high volume, which are consistently low, and where the volume is unpredictable.
A Tuesday lunch window that runs at forty percent of Friday dinner volume does not need a Tuesday dinner schedule. A Friday dinner window that consistently spikes in the last ninety minutes needs staffing in that window, not spread evenly across the full service.
Scheduling from this pattern rather than from a standing template produces better coverage at lower labor cost. The data is already in the POS; it only needs to be read.
The server performance read identifies patterns worth investigating
Server-level data shows average check per cover, table turn time, and sometimes void and comp rates. These are patterns, not performance evaluations. The data identifies where a conversation is warranted, not what the conclusion should be.
A server with a consistently higher average check may be upselling effectively. A server with a consistently lower turn time in a high-volume window may be creating a bottleneck. Both patterns are worth understanding before attributing them to individual performance or style.
The data rarely provides the full explanation. It provides the starting point for a conversation that identifies whether a pattern is worth addressing.
How the Conductor Reads Your POS Patterns
The owner opens the Conductor and asks what the POS data shows about menu performance and where staffing and volume are not aligned. The Conductor is Kiluma’s context-aware AI. The question starts with this restaurant’s own export data, not a generic analytics dashboard.
It draws from the Living Library, the active layer where the restaurant’s POS export data has been saved. It returns specific findings: which items are low-volume and low-margin, which service windows show consistent volume spikes, which server patterns are worth investigating. These are findings from this restaurant’s own three months of data.
Kiluma is the knowledge layer, not the POS. Toast or Square runs the register and generates the data. Kiluma reasons over what the restaurant has exported and saved so the owner can ask specific questions rather than navigating a dense reporting interface.
Pull One POS Report Before the Next Scheduling Decision
Before building next week’s schedule, pull the covers-by-hour report from the POS for the last four weeks. Find the two service windows with the highest and lowest average volume.
Schedule from that pattern rather than from the template used last month.
The Restaurant That Was Running on Intuition Now Has Data
Your POS already knows which menu items are losing you money. The operator who read the first line of this article and thought “I should actually look at that” has the data. The POS Decision Method is what makes it actionable. Try Kiluma free for 14 days at kiluma.ai.