Your POS isn't short on data. It's short on someone asking it the right question. Following on from last week's piece on menu engineering, here are six reports worth pulling this month, and a prompt for each one. Export the report, paste it into whatever AI tool you're using along with the prompt, and adjust the wording to fit your own numbers.

1. Product sales

Pull your full product sales report for the last quarter, not just this week's snapshot.

Prompt: "Here is my product sales report for the last quarter. Based only on the numbers, which items should I drop, which should I push harder, which are underpriced for what they cost me to make, and which two or three items would work well bundled together. Explain your reasoning for each."

This is the one most owners resist, because it usually means admitting a favourite item has been quietly losing money for months. The data doesn't care how proud you are of it.

2. Hourly sales

Pull hourly sales across a full month.

Prompt: "Here is my hourly sales data for the past month. Identify my genuine busiest and quietest trading periods, hour by hour, not shift by shift. Recommend staffing levels for each period and suggest one operational change or promotion that would make sense for the quietest two hours of my trading day."

Most rosters are built on what felt busy last Tuesday. The report usually tells a slightly different story, and the gap between the two is where labour cost quietly builds.

3. Category sales

Break sales down by category rather than individual item.

Prompt: "Here is my sales broken down by category. Which categories are underperforming relative to the rest of the menu. Suggest three practical ways to lift average customer spend without adding new items to the menu."

Sometimes the answer isn't selling more coffee. It's making sure everything sold alongside the coffee is actually pulling its weight.

4. Staff sales

Pull sales by staff member if your POS tracks it.

Prompt: "Here is a sales report broken down by staff member. Who consistently upsells or drives a higher average ticket, who might benefit from coaching, and what patterns stand out across the team. Keep this factual and specific to the data, not general performance advice."

Most cafes can name their fastest barista without hesitation. Fewer can name who actually moves the average ticket up, and that's usually the more valuable thing to know.

5. Discounts and voids

Pull your discount and void report.

Prompt: "Here is my discount and void report for the past month. Flag any unusual patterns, timing, or staff members that stand out, and note anything that looks like it could be a training gap rather than a one off. Don't assume wrongdoing, just flag what's worth a closer look."

Not every discount is a friendly gesture and not every void is an honest mistake. Most are. The report is how you tell the difference without accusing anyone of anything.

6. Customer purchase history

Pull customer purchase history if you're set up to capture it.

Prompt: "Here is customer purchase history data. Based on real buying patterns in this data, design three offers that would increase visit frequency or average spend for specific groups of customers, not a single blanket discount for everyone. Explain who each offer targets and why."

A blanket discount sent to your whole list is easy. It's also the least effective thing you can do with this data. The pattern in front of you is usually more specific than that, and worth acting on as such.

What to actually do

Pick one report this week, not all six. Export it, run the prompt, and look for the one change you can make before your next roster goes up or your next menu review happens. Small, acted on, beats comprehensive and shelved.

Which of these six have you actually pulled in the last month, and which one are you avoiding because you already suspect what it'll show you?