What a silent customer costs: the arithmetic of quiet churn in B2B wholesale

In B2B wholesale, customers rarely leave with a bang. They just stop ordering: the usual rhythm is an order every two weeks, then a month’s pause, then six weeks — and that’s it, the customer is already with a competitor. A row would actually be more useful: at least someone would react to it.

Why it goes unnoticed

A sales rep lives in the inbound flow: calls, invoices, shipments. A customer who didn’t call creates no event — and therefore no reason to think of them. The CRM doesn’t cure this: it shows perfectly well what is happening and says nothing about what has stopped happening.

Meanwhile, the data to detect churn is already in the company — 1C (the ERP/accounting system standard in Russia) holds every customer’s complete order history. From it you can work out an individual rhythm: this one orders every 12 days, that one once a month. A deviation from their own rhythm by a multiple is a signal long before the customer can be considered lost.

The arithmetic

The maths is rough but persuasive. Say the average order is 40,000 ₽, a customer orders once a month, and five customers “go missing” per quarter. Then twenty quietly leave over a year, and the annual revenue they take with them — twenty customers × 40,000 ₽ × 12 orders — is almost ten million. Even if only a small fraction can be retained, we are talking millions; and “retained” here means nothing more than “called in time”.

This is an estimate, not a guarantee: every business has its own margin, its own rhythm and its own share of customers who come back. The AI Sales Employee pricing page has a calculator that runs the same maths on your figures — with the formula shown right in the block.

What automation does about it

The retention loop is the most rewarding part of an AI agent implementation, because the benefit is measurable customer by customer. Every morning the agent compares each customer’s actual last-order date with their personal rhythm. Fallen noticeably out of rhythm — the sales rep gets a task with a call brief: what the customer usually buys, whether it is in stock, what their average order is.

The easiest way to see it is the “An AI Sales Employee’s Morning” demo (the data there is fictional): the agent finds a customer who has been silent for 47 days against a usual rhythm of an order every 12 days — and the sales rep gets a task with a brief on what that customer usually buys. In a real implementation the same principle runs on the live order history from 1C — and the first pass over historical data always shows who the company has already almost lost.

Quiet churn is precisely the case where automation doesn’t replace a person but gives them back a reason to act. The sales rep still makes the call. They just know who to call now.

How the loop works end to end is in the AI Sales Employee case study, and what implementation costs is on the pricing page.