There is a version of this subject that produces nothing: the owner who builds a beautiful dashboard, checks it every morning, feels informed, and runs the restaurant exactly as before. Data that does not change a decision is a hobby with extra steps. This last topic is about the crossing point — how to turn the groups and the lists into a small number of decisions each month, how to combine numbers with what customers actually said, and how to catch yourself using a figure as decoration for a choice you had already made.
🔗 Numbers find the question, words answer it
The two courses in this pair are halves of one instrument, and using either alone gets you into trouble in a predictable way. Numbers alone give you a precise fact with no explanation, which tends to produce confident wrong action: Thursday is down eighteen percent, so let us discount Thursday — when the real cause was a service problem that a discount will now subsidise. Words alone give you a vivid story with no sense of scale, which produces the opposite error: one furious review about the noise, and a restaurant rebuilt around a complaint that four hundred other people never made.
Used together they cover each other's blind spot, and the working method is unglamorous.
- Start from the number. A drop, a rising quiet list, a group shrinking. That is the question, and it comes with a date attached.
- Go to the words for that same period. Reviews, survey answers, the notebook by the pass. The reason is usually already written down.
- Check it against the room. Ask the staff who worked those shifts. They will often confirm or kill the theory in one sentence.
- Then decide, and write down what you expect to happen. The prediction is what makes it a decision rather than a reaction.
That last step is the one that separates people who improve from people who merely stay busy. Writing "if this is the wait, then the quiet list should stop growing by next month" costs you nothing and gives you a real answer later, instead of the usual fog where every change is declared a success because nobody said what success would look like.
📋 Two decisions a month
The discipline is the same one the feedback course arrived at, for the same reason: a long list of improvements is a list of none. Once a month, sit down with the groups, the quiet list and the themes from your comments, and pick two things — each with a person's name on it and a date to check. Not ten. Two.
What those two decisions look like in practice is usually smaller than people expect.
| What the data shows | The kind of decision it points at |
| A large share of sales from a small group of regulars | Move effort from acquiring strangers to protecting those people |
| A quiet list that is growing month over month | Stop and find the cause before spending anything on winning them back |
| Occasionals who come three times a year | One well-aimed reason to make it four; this is where an offer pays |
| A dish ordered constantly by your best customers | Protect it: do not cut it for being fiddly, and never quietly change it |
| A service or a day that only ever produces one-timers | Either fix what that service is doing, or stop promoting it |
Notice that none of these are about buying software or launching a campaign. The decisions customer data actually produces in a restaurant are about where you point your attention, which dish stays, which day gets looked at, and who gets a phone call — and that is exactly why this works without a budget.
🪞 When the data is only decoration
The failure mode to watch for in yourself is using numbers to justify rather than to decide, and it is easy to spot with one question: would you have changed your mind if the number had come out the other way? If the honest answer is no, you were not deciding with data. You were shopping for support.
It shows up in a few recognisable shapes. Picking the timeframe that flatters — comparing against the worst month rather than the same month last year. Reporting a percentage without the count behind it, so that "bookings from that group doubled" turns out to mean four instead of two. Watching numbers that only ever go up, like total customers ever collected, which cannot deliver bad news and therefore cannot inform anything. And the most common of all: measuring the things that are easy to measure rather than the things that matter, which is how restaurants end up with a folder of reports and no idea whether their best customers are still coming.
The antidote is to keep very few numbers and let them be capable of embarrassing you. For most restaurants that is about five: how many different customers this month, what share of sales came from repeat customers, how many are on the quiet list, how many of last month's quiet list came back, and the average spend per visit. Reviewed monthly, written down, compared only against your own past. If you have those five and you act on two things a month, you are doing more with data than most of this industry.
🧭 What you have built
Put the six topics together and you have a restaurant that knows its customers as people rather than as a total on a report. You found the data you were already collecting and stopped confusing transactions with people (topic 1). You merged four contradicting lists into one record per person, collecting less rather than more, and got clear about whose data it is (topic 2). You gave every customer three numbers — how often, how much, how long ago — and discovered the four or five groups hiding inside "our customers" (topic 3). You learned to use that knowledge the way a good host would, at the level of the group rather than the individual (topic 4). You built the monthly list of who is quietly leaving, which is the alarm no other system in your restaurant will ever sound (topic 5). And now it runs: five numbers, two decisions, every month.
This also closes the run of five marketing courses, and they were always one argument. The program gives people a reason to come back, email reaches the ones who already know you, social gets you found by the ones who do not, feedback tells you what they think, and data tells you who they are and which of them you are losing. The thread through all five is the same and it is worth saying plainly: it is far cheaper to keep the customers you have than to buy new ones, and almost everything in this industry is arranged to make you forget that. The five numbers in this course exist mostly to keep reminding you.