Most restaurants do not have a customer database. They have four: a spreadsheet of email addresses, the booking system, the loyalty program, and the online ordering account list. The same person appears in three of them under two spellings and one old phone number, so nobody can answer a simple question like "how many regulars do we actually have?" This topic is about fixing that with the least possible work, and about the discipline of not collecting things you will never use.
🪪 One person, one record
The single idea that makes everything downstream possible is that each human being should appear once. Until that is true, every count you make is wrong in an unknown direction — you will double-count your best customers, because the people who engage most are exactly the ones who ended up in several systems.
To join records you need a key: one field that is the same person everywhere. In a restaurant that is almost always the email address or the phone number, so pick one and treat it as the identity. Then the merge is mechanical, and it is a morning's work, not a project.
- Export everything to one sheet. One row per record, with where it came from in a column, so you can always trace it back.
- Normalise before matching. Lowercase the emails, strip spaces and dashes from the phone numbers. Half your "different" customers are the same one typed differently.
- Merge on the key, keep the fullest version. Best spelling of the name, newest phone, and the earliest first-visit date — that date is what tells you how long you have had them.
- Keep the source column. When someone unsubscribes or asks to be deleted, you need to know everywhere they exist.
- Do it again on a schedule. Monthly or quarterly, forever. A merge done once decays back into four lists within the year.
You will not reach a hundred percent and you do not need to. The goal is that your best and most frequent customers — the ones the rest of this course is about — appear once each, because those are the people every decision here turns on.
🧹 Dirty data is worse than no data
A list nobody trusts stops being used, and a list that is wrong in a specific direction is actively dangerous, because it produces confident conclusions that are false. The most common problems are dull and easy to fix once you know to look.
| Problem | What it does to your numbers |
| Duplicates of the same person | Your best customers look like several mediocre ones, so frequency looks worse than it is |
| Staff typing a fake email to finish the sign-up | Silent poison: the list grows while the reachable list does not |
| Test and staff records left in | A "customer" with sixty visits distorts every average you calculate |
| Group bookings under one name | One record that behaves like twenty people; obvious once you see it |
| No first-visit date | You cannot tell a new customer from an old one, which breaks half of topic 3 |
Two habits keep it clean at almost no cost. Fix the collection point rather than the sheet — if servers are inventing emails to get through the loyalty screen, the screen or the incentive is the problem and no amount of cleaning downstream will help. And look at your own list with your eyes, sorted by visits, once a quarter: the top twenty rows will show you every kind of error you have, faster than any rule.
🙋 Ask for less than you think
Every field you ask for costs you sign-ups, and every field you store is something you have to protect, keep current and eventually explain. So the rule is to collect what you will actually use in the next three months and nothing else. In practice that is a way to recognise them and a way to reach them: a name and one contact field.
Be especially wary of the fields that feel charming and almost never get used. Birthdays are the classic — most restaurants collect thousands and send nothing, and the ones who do send a birthday offer could have got the same result from a date they already had. Dietary preferences, anniversaries, favourite table: genuinely useful in a small fine-dining room where a manager reads them before service, and dead weight anywhere else. If you cannot name the specific thing you will do with a field, do not ask for it.
Where you ask matters as much as what you ask. The easiest moments are the ones where the customer is already giving you information for their own benefit — booking a table, placing an online order, joining the program because they want the reward. Trying to collect data at the door, or from someone who just wants to pay and leave, gets you refusals and fake answers, which is worse than nothing.
⚖️ It is their data, and there are rules
Personal data comes with legal obligations that differ by country and sometimes by state — Europe's GDPR, Mexico's LFPDPPP, California's CCPA and others all cover this ground differently, and they change. This course cannot tell you what applies to you, and a restaurant course should not pretend to. What it can tell you is that something applies to you, and that the basic expectations behind all of them are the same everywhere and are mostly common sense.
- Tell people what you are collecting and why, in one plain sentence at the point where you ask.
- Use it for what you said and nothing else. Data given for a booking is not a marketing list.
- Let people leave. Unsubscribing and being deleted should be easy and should actually happen, everywhere they exist.
- Never sell or share it. There is no version of this that ends well for a neighbourhood restaurant.
- Protect it like cash. Restricted access, no customer lists on personal phones, no spreadsheets emailed around.
- Keep card data out of it entirely. That is the payment processor's job and its own regime, as the payments course covered.
The practical version is simpler than the legal one: hold what a reasonable customer would expect you to hold, use it for what they would expect you to use it for, and be able to explain the whole thing in a sentence without wincing. If you find yourself hoping a customer never asks what you keep, that is the answer — and it is the same instinct that topic 4 turns into a working test.