The POS knows more about your operation than any other system in your stack. Most operators are reading about 15 percent of what it can tell them. This post is a walk through the reports I open every week that most operators never touch, and what those reports actually reveal.
I write this specifically about Toast because that is what I run, but every meaningful POS has an equivalent report. If you are on Square, Lightspeed, Revel, or a legacy system, the report names will change. The signals will not.
Sales and labor are lagging indicators
Most operators start their day with the sales report and the labor report. Fine. Both are lagging. They tell you what happened yesterday. Neither tells you why.
The reports one layer down are leading indicators. Modifier attach rates predict food cost variance three weeks out. Void patterns predict theft or training gaps two weeks out. Day-part mix drift predicts revenue softness a month out. If you only read the lagging reports, you find out about problems after they have already cost you money.
Yesterday's sales tells you what happened. Yesterday's modifier attach tells you what is about to happen.
Six reports worth opening
1. Modifier attach rate over time
Every modifier ("add avocado," "sub gluten-free bun," "extra dressing") has a natural attach rate on the item it belongs to. That rate is stable across weeks in a well-run operation. When it drifts, something changed.
What to pull. Modifier attach as a percent of the parent item, week over week, per unit, for the last 12 weeks. Sort by biggest weekly change.
What it tells you. If the "add cheese" modifier used to attach on 40 percent of burgers and now attaches on 25 percent, one of three things is happening. Servers are ringing it as a comp instead of a paid modifier (theft or laziness). The line is adding cheese without ringing it (waste and understated food cost). Or the menu is being ordered differently for a real reason. Investigate.
Modifier attach is the single most under-read POS report in the industry. It is almost always where food cost variance starts before it shows up on the P&L.
2. Voids by employee and hour
Voids are legitimate. A ticket gets rung wrong, a guest changes their mind, an item comes back to the line. All fine. What is not fine is a void pattern concentrated on one employee, one time window, or one tender type.
What to pull. Voids by employee, by hour, by tender type, over the last 30 days. Compare each employee's void rate against the peer median for their role.
Fig. 1 · The two red bars are worth a conversation.
What it tells you. Legitimate void behavior is normally distributed across employees. An employee whose voids run three to four times the peer median is worth a conversation. It is not always theft. It is sometimes bad training or a bad terminal habit. Either way, it deserves investigation.
3. Comp signatures by shift
Comps are voluntary reductions to the check. Some are legitimate service recovery. Some are marketing promos. Some are managers taking care of friends. Some are theft.
What to pull. Comps by shift, by manager approving, by dollar amount, by reason code. Compare each manager's comp rate against peers.
What it tells you. A manager who runs comps twice the peer rate is either much more generous with recovery (fine, but ask), running a lot of friends-and-family (not fine), or covering for a service problem that has not been named (very not fine). Comp signatures are also where theft rings tend to hide, because comps do not leave the same audit trail as voids.
4. Day-part mix per unit
Every restaurant has a natural day-part mix. Lunch, happy hour, dinner, late night. That mix is remarkably stable. When it drifts, something changed in the guest pattern or the operation.
What to pull. Sales by day-part as a percent of daily sales, per unit, monthly for the last 12 months.
What it tells you. If dinner used to be 60 percent of sales and now is 52 percent, either dinner traffic is softening (marketing or service problem) or lunch is compensating (good sign or promo pull-forward). Day-part mix drift is where you catch category-level revenue softness before it shows up in total sales.
5. Guest count per server per hour
Servers have a natural cover count they can handle at the service standard the operation targets. That number is knowable.
What to pull. Covers per server per hour, by day-part, per unit, monthly.
What it tells you. If your top servers are averaging 12 covers an hour and your peer restaurants in the region run 9, you are either understaffed (which shows up in tips, comps, and service recovery), or your top servers are better than the market. Both are useful to know. If your average server is running 6, you are overstaffed and labor cost is bleeding.
6. Menu item attach patterns
Which items tend to sell together. This is not the same as top items. It is which items appear in the same ticket.
What to pull. Item attach matrix. For every top-25 item, which items are most likely to appear in the same ticket.
What it tells you. This is your menu engineering lever. If a mid-margin appetizer attaches to a high-margin entree 70 percent of the time, feature the appetizer. If a high-margin cocktail almost never attaches to your top entree, ask the server training team why. Menu mix and attach patterns are a separate post in the Where The Money Leaks category. The report starts here.
What to ignore
Not every POS report is useful. Skip these unless you have a specific reason.
- Hourly sales without a comparison. Hourly sales at a single unit for a single day is noise. Hourly sales overlaid with staff schedule is useful.
- Top items by revenue with no margin. The top revenue item is often a mid-margin item that lives on happy hour. Rank by contribution, not by revenue.
- Discount reports without segment. A single discount report tells you nothing. A discount report split by promo, employee-approved, and management-approved tells you a lot.
- Server ranking dashboards visible to servers. They create competition that damages team dynamics. Managers see them. Servers get coached individually.
The weekly rhythm
The point of this whole post is not "look at more reports." Operators who try to look at everything look at nothing. The point is to pick a few and cycle through them.
Here is the rhythm that works. Same trick as the dashboard adoption pattern.
- Every week, pull one of the six reports for one unit. Rotate through the six over six weeks, then start over.
- In the weekly one-on-one, ask the general manager about the report you just pulled. What do they see? What are they doing about it?
- Document the response in the weekly log. If the same report shows the same anomaly two weeks later, escalate.
Six reports, six weeks, one loop. That is 52 conversations a year about the second layer of POS data. Nobody in your competitive set is doing this. Almost nobody.
What I got wrong
One mistake worth naming.
I once flagged a server for a void pattern that turned out to be a POS terminal bug. The terminal at their station was reordering menu categories in a way that caused misclicks, and the server was legitimately voiding to correct the ring. I approached the conversation as if I had proof of a problem. I did not. I had a signal that pointed at a problem that could have been three different things.
Now I open POS anomaly conversations with the pattern I see and the question I have, not with the conclusion I drew. "Your voids ran 4x the peer median last month. Walk me through what happened." Almost always the real cause reveals itself, and it is often not what the report suggests. But without the report I would never have asked.
The point
Your POS is a rich data source. Most operators drink from the sales-and-labor tap and ignore the rest. The reports one layer down are where the interesting stories live. They are also the reports that let you catch problems as leading indicators, not as line items on next month's P&L.
Pick six. Rotate weekly. Ask one specific question per meeting. Every operator I have taught this to catches something meaningful inside six weeks that they would not have caught otherwise.