Cuedis

What your POS database already knows about your business

Every shop, restaurant and distributor that uses a point-of-sale or ERP system is sitting on a detailed diary of the business. Each sale, each delivery, each refund, each price change and each staff log-in is written down, with a time and a name. Most owners only ever see a small slice of it: the end-of-day total, a monthly sales report, maybe a stock count.

The rest is still there. It is just never read.

This post is about what that diary already contains, what it can tell you if someone actually reads it, and why it so rarely happens.

What is in there

The names differ from one system to another, but almost every POS or ERP database holds the same kinds of records:

  • Sales and their lines: what was sold, when, at which counter, by which cashier, at what price and with what discount.
  • Purchases and deliveries: what you ordered from each supplier, what arrived, when it arrived and at what cost.
  • Products: cost price, selling price, category, supplier, and often the history of price changes.
  • Customers and members: who buys what and how often (if you run a loyalty scheme).
  • Voids, refunds and cancellations: who did them, when and for how much.
  • Stock movements: transfers between outlets, adjustments, write-offs.
  • Staff: who was logged in, at which till, for which shift.

None of this was collected for analysis. It was collected because the system needed it to print a receipt or post a delivery. But put together, it describes almost everything that happens in the business.

What it can tell you

Here are the kinds of things that are hiding in plain sight in most databases. None of them needs new data. They only need someone to compare today with what is normal.

A supplier quietly getting worse. Each delivery has an order date and a received date. If a supplier who used to deliver in four days now takes nine, the evidence is in the purchase records weeks before anyone complains about empty shelves.

A margin slipping a little every week. When a supplier raises a cost price by a few cents and the selling price stays the same, nothing looks wrong on any single day. Over two months, the margin on that product line can drop several points.

A product that suddenly starts selling. A sharp rise in a slow product is a chance to reorder before you run out, or a sign that something changed: a competitor closed, a price is wrong, or someone posted about it.

Unusual till activity. Voids and refunds are normal. A cluster of them on one counter, one shift or one cashier is not.

A customer ordering less. In wholesale, a regular customer whose orders slowly shrink is usually moving business elsewhere. Their order history shows it long before they stop calling.

A bad day with one cause. When thirty products all sell less on the same afternoon, it is rarely thirty separate problems. More often the card terminal was down, a till was offline, or the shop closed early. The pattern is in the data.

Why nobody reads it

If the information is already there, why do most owners only find out about these things when they become expensive?

Reports answer questions you already thought of. A monthly sales report tells you what you asked it to tell you. It cannot tell you that a supplier you never think about has become unreliable, because nobody built a report for that.

Dashboards need someone to look at them. A dashboard shows everything at once and leaves you to spot what changed. On a busy day, nobody does. Many dashboards are opened often in the first month and rarely after that.

Normal is different for everything. A 10% drop means nothing for a product that swings 30% every week, and a lot for a product that has sold the same amount for a year. Knowing what is unusual means knowing what is normal for each supplier, product, customer and staff member separately. That is too much for a person to hold in their head across thousands of items.

The problems are slow. The most expensive problems (margin leaks, a slipping supplier, a customer drifting away) never cause a single alarming day. They build up quietly.

What reading it well looks like

The useful approach is to compare each new record with what is normal for that particular thing, and only bring the unusual to someone's attention. In practice that means:

  1. Learn normal per item. Each supplier's usual delivery time, each product's usual weekly sales and margin, each cashier's usual void rate.
  2. Remember seasonality. Month end, payday, school holidays, Chinese New Year and Hari Raya all change what normal looks like. A good system learns these patterns instead of raising alarms every year.
  3. Watch trends, not only days. A margin that loses a little every week should be flagged even though no single day looks wrong.
  4. Group related alarms. Thirty products dropping on the same day should be reported as one incident, not thirty.
  5. Say it in plain words. "Kelvin Electronics has been late on 24 deliveries since 7 April; it used to take 4 to 6 days" is useful. A red cell in a chart is not.
  6. Keep the evidence. When something is flagged, the owner should be able to see the records behind it.

Where to start

You do not need a data team to begin. A few simple habits help:

  • Ask whoever looks after your system for a read-only login to the database, so that anything reading it cannot change it. (We have step-by-step guides for SQL Server and MySQL.)
  • Pick three things that would hurt most if they went wrong without you noticing: perhaps your top supplier's delivery times, the margin on your best-selling category, and refunds.
  • Check them weekly against the same weeks last year, not only against last week.

This is also why we built Cuedis. It sits beside the MySQL or SQL Server database you already have, reads new records every few minutes, learns what is normal for each supplier, product, customer and staff member, and sends a short briefing each morning about what changed. It only ever reads, and it runs on your own computer.

Whether you use a tool or a weekly habit, the main point stands: the data you need to run the business better is already being collected. It is worth reading.