Customers rarely leave slamming the door

A customer who cancels, you hear about. That is unpleasant, but it is clear and you can respond to it.

The customers who really cost you money do not do that. They order less often. Then in smaller quantities. Then once they do not order at all. And at a certain point there are six months between orders and the conversation you could have had can no longer be had.

Nobody said anything. Nothing happened. And you have still lost a customer.

Why you miss this

Because a number that drops slowly sets off no alarm.

In your revenue report you see a total. If one customer goes from ten orders a year to six, that is barely visible in your total, certainly if a new customer came in meanwhile. Your total stays flat and you think things are going well.

In the meantime there is a movement under that flat number you cannot see: existing customers eroding, compensated by new customers you acquire at a high cost. That is an expensive way to stay in the same place.

What you have to measure

Frequency, not volume. That is the whole point.

A customer who spent the same this year as last year but across four orders instead of ten is on the way out. The amount hides that, the number of times does not.

So build a picture per customer of how often they normally order, and measure how much time has passed since the last one against that normal rhythm. A customer who orders every three weeks and has now been silent for seven is a signal. A customer who orders twice a year and has been silent for four months is not.

That is the difference between a list of customers you have not seen for a while and a list of customers where something is changing. The first is a report, the second is usable.

How you define active

This is where almost everyone goes wrong. The common definition is: a customer who has ordered in the past twelve months.

That sounds watertight, and it is too coarse. Someone who buys every month and someone who buys once a year end up in the same bucket, while those are two completely different customers with a different risk.

If you take frequency into your definition, that single category splits into groups you can actually do something with. And only then can you build a meaningful signal, because the signal depends on what is normal for that group.

What you do with the signal

An alert nobody follows up is one alert too many, so be strict about this part.

Decide in advance who receives it and what that person is supposed to do. Call, visit, send an email. And set the number of signals so that it stays manageable. Five a week that get followed up are worth infinitely more than fifty that get ignored.

Put a feedback loop on it as well: what came out of it? If after three months you can see that half the customers who were called have ordered again, you know what this is worth and it stays alive. If not, you know your signal is no good and you can adjust it.

Where to be careful

Not every decline is a problem you can solve. Sometimes a customer has simply become smaller, or their own market has changed. A phone call does not fix that and calling too often works against you.

And season is the classic trap. A customer who orders less every July is not a risk in July. So your normal rhythm has to hold per customer and per period, otherwise you get a wave of false alarms every summer and everyone stops looking.

What you need for it

Not much. Your sales history per customer per date, and enough history to establish a normal rhythm. Two years is comfortable, one year is workable.

That sits in your ERP or your invoicing. It is not a new system, it is asking a different question of data you already have.

The claim

Your revenue figure does not tell you that your customers are walking away, because it adds new and departing customers together. Frequency per customer does tell you, and it sits in the same data.