What to do with six weeks of POS exports

Somebody eventually tells you the answers are in your data. Then you export six weeks of sales out of the POS, open the first file, and it is four thousand rows of every item anyone rang in, and you close it again.

That is not a discipline problem. Nobody has an afternoon.

The mistake is the open-ended question

The instinct is to hand the whole pile over and ask what it says. That produces a paragraph of things you already knew, dressed up.

Sales are highest on Friday. Your top seller is the burger. Lunch is busier than you thought on Wednesdays. Fine. Nobody is going to change a shift over that.

A pile of exports is only useful against a question you could already have argued about. If you cannot say what you would do differently depending on the answer, do not ask it yet.

Three that are worth it

What did this exact week look like the last three times it happened?

Not the average. The last three. Averages hide the thing you need — that two of the three were ordinary and the third was a disaster because of an event nobody wrote down. Three actual weeks side by side is a thing a GM can reason about; a mean is not.

Which items move together?

This one earns its keep in a kitchen. If the fish special and the house salad rise and fall together, that is one prep decision, not two. Most kitchens are counting a dozen things separately that behave like four.

Where does the same item ring at different prices?

Across stores, this is where the money hides. A modifier that is free at one location and eighty cents at another. A comp reason one manager uses forty times a month and the rest use twice. Nobody is stealing. The buttons are just different, and nobody has compared them since the last menu change.

The one that is not worth your afternoon

"Predict next month's sales."

You will get a number. It will be roughly last month with a trend on it, and you will have no way to tell whether it is any good until the month is over — at which point you will not remember what it predicted.

Forecasting is a real thing and it is not a first project. Ask it of six weeks of exports and you get confidence without accuracy, which is worse than no answer at all, because somebody will schedule against it.

The shape that works

Export by day, by item, by store, for a period long enough to include the week you care about at least three times. Six weeks is not that. A year is.

That surprises people. Six weeks is what the POS defaults to and it is enough to see a pattern inside a month, but if you want to know what the second week of November looks like, six weeks of summer cannot tell you.

Keep the export boring. Date, store, item, quantity, gross, discount, comp reason. Every extra column is another thing to explain before anyone can use it.

Then check one answer by hand

This is the step people skip and it is the one that decides whether anybody trusts the rest.

Take one number the analysis produced — one item, one day, one store — and go and count it in the POS yourself. If it matches, you have a tool. If it does not, you have found out now, on one number, instead of in a district meeting.

I have never regretted the twenty minutes. I have regretted skipping it.

What you keep

Not the analysis. The questions.

Six weeks from now the exports are stale and the answers are gone. What survives is the list of three questions worth asking of a sales file, and the fact that somebody in your group now knows how to get an export that answers them.

That is a smaller-sounding outcome than a dashboard and it is worth considerably more, because it does not stop working when you change POS.

---

Not sure reporting is where your week is going? Take the free check — eight questions, no email, and it names the two things worth attacking first.