Most businesses find out a customer left long after they are gone. The account simply goes quiet, and one day you notice the orders stopped months ago. By then it is far too late to do anything about it. The frustrating part is that the customer churn warning signs data almost always shows the exit in advance, sitting quietly in order history you already have.

You do not need machine learning for this. You do not need a data scientist or an expensive tool. You need three signals you can compute in a spreadsheet from records you already keep, and, more importantly, a plan for who acts when a signal fires. A dashboard that spots leavers but triggers no phone call is just a sadder dashboard.

Let me walk through the three signals, then the half everyone forgets: the human follow-up.

Signal one: the stretched reorder gap

For any business with repeat purchases, every customer has a natural rhythm. A cafe supplier reorders coffee beans every three weeks. A salon buys product monthly. A distributor restocks on a predictable cycle. That rhythm is your early warning system.

Here is the simple method:

  1. For each regular customer, calculate their typical gap between orders, the average number of days between purchases.
  2. Track how long it has been since their last order.
  3. Flag anyone who has gone significantly past their normal gap, say 1.5 times their usual interval.

A customer who normally reorders every 20 days and is now at day 35 is not "due soon." They are drifting, and often they have already started buying from someone else. This is the single most powerful churn signal for most SMEs, and it is pure arithmetic on your order dates.

Signal two: the shrinking basket

The second signal is subtler because the customer is still buying, so nothing looks wrong at a glance. But the size of each order tells a story. A customer who used to spend three million rupiah per order and has quietly slid to one million is testing the exit. They may be splitting their business with a competitor, trialing an alternative, or losing confidence.

To catch it:

  • Compare each customer's recent average order value against their earlier average.
  • Flag meaningful, sustained drops, not one-off small orders, but a clear downward trend over several purchases.
  • Pay special attention to your best customers. A 30 percent decline from a large account matters far more than a small customer disappearing.

A steadily shrinking basket from a formerly strong customer is a customer with one foot out the door. Caught early, it is very recoverable. Caught late, they are already mostly gone.

Signal three: the silent complaint channel

The third signal is about noise, or the lack of it. Unhappy customers do not always complain. Many just go quiet and leave. So two things are worth watching:

  • A customer who used to engage and now does not. They used to reply to messages, ask questions, respond to offers, and now there is silence. Disengagement often precedes departure.
  • Complaints that came in and were never properly closed. A customer who raised an issue that got a weak or no resolution is a prime candidate to leave. The complaint was their last attempt to stay.

This signal often lives outside your sales data, in your chat history and support notes. That is exactly why it gets missed, and why keeping this information in one accessible place matters. If your customer records are scattered across personal phones and notebooks, you cannot see this pattern at all. That is one more argument for a single source of truth for your business data.

The half everyone forgets: who acts

Here is the hard truth. Detection without action is worthless. I have seen businesses build a neat "at-risk customer" list that nobody ever calls. The list just grows. The customers still leave. Now you have a spreadsheet documenting your losses in real time.

So before you build any of this, answer the operational questions:

  1. Who owns the at-risk list? A named person, not "the team." Someone whose job includes calling flagged customers this week.
  2. What do they say and offer? Not a discount reflex. Often the winning move is simply a genuine call: "We noticed we have not served you in a while, is everything okay?" Sometimes that alone recovers the account. Sometimes you learn about a problem you can fix.
  3. What is the cadence? Review the flags weekly. Churn signals are perishable. A stretched reorder gap caught at day 35 is recoverable; caught at day 120, the customer has a new supplier and a new habit.

The offer matters less than the timing and the human contact. People stay with businesses that notice them. The whole point of the data is to tell one person which customers to notice, this week, before it is too late.

A one-page early-warning routine

Put it together into something you can actually run:

  • Weekly, pull three lists from your order data: customers past 1.5x their normal reorder gap, customers whose average basket has clearly shrunk, and previously engaged customers who have gone silent or had an unresolved complaint.
  • Merge into one at-risk list, ranked by customer value. Save your best contact time for your most valuable accounts.
  • Hand it to one owner who calls the top of the list every week. Real conversation, not a mass blast.
  • Log the outcome. What you learn feeds back and makes next week's list sharper.

Simple, cheap, and it works because it converts data you already have into action while there is still time to act.

The practical takeaway

The customer churn warning signs data you need are already sitting in your order history: a reorder gap that has stretched past normal, a basket that keeps shrinking, and a once-engaged customer who has gone quiet. None of it requires fancy analytics, just arithmetic and attention. But the signals only matter if someone acts on them. Give the at-risk list a named owner, a weekly rhythm, and a genuine human call, and you will save accounts you used to lose silently. Detection is the easy half; make sure you build the half that picks up the phone.