Bruce feels the word ‘churn’ should only be used for customer relationships based on contracts, e.g. in banking, mobile phones, media subscriptions. It’s not always simple, even then – many people have bank accounts with which they have no activity, so are these really ‘customers’ for any practical purpose?

Overall customer-loss fractions vary considerably as a relationship develops – typically a high fraction churn in the first year, a smaller fraction in the next, and so on. So it’s only possible to know what to do on customer retention if this cohort-by-cohort pattern is understood.

Very many major firms in contractual industries fail to report churn accurately [for this and other reasons] – Netflix even got sued by investors who claimed they had been misled by its churn reporting.

An important result of misreporting churn is that analysts and the firms themselves frequently estimate ‘customer lifetime value’ quite inaccurately – and therefore the ‘value’ of a customer base.

Differing churn rates amongst distinct customer segments can cause overall churn rates apparently to change, when in reality so such change is occurring [start with 50,000 customers churning at 20% and 50,000 churning at 10% and the overall rate will appear to fall] – which can lead to incorrect marketing, pricing or other responses.

… all clever stuff, and if you need to know more, investigate Bruce’s work some more.  

For non-contract situations, purchasing repeat rate [fraction of customers in one period who also purchase in the next] or repurchase rate [fraction who buy the same brand on the next purchase occasion] are more relevant. [Bruce recommends Farris et al, Marketing Metrics, 2006].

Whilst I see that transactional data can only give true churn measures in contractual cases, we do need to understand it in some way in other cases. E.g. I have in recent times come across a law firm, a consulting firm, and a marketing agency, none of whom were conscious of client losses. In each case, when we asked how many clients they had, they reported the number with accounts, though very old ones were sometimes informally disregarded. Key to helping them was to identify how many clients of which type and size were active, or lost, or [which was key] merely dormant. This can’t be known from the transactional data, but as in many other cases had to be found out by asking. This is how my friend Lars Finskud at Vanguard Strategy works his magic in consumer goods, pharma, and other sectors.

Often, the customer is really lost much earlier, when their activity falls off. So it’s good to see McKinsey suggesting this as a useful way of segmenting customers in this article looking at activity segments amongst mobile phone subscribers.

Intriguingly, consultancies and other large professional firms themselves often don’t know when a client is lost. Some clients will only ever seek one-off projects, of course, but larger ones might become regular customers by buying project after project. So when is one of these ‘annuity’ clients lost? .. when they have not bought for 3 months .. or 6 months .. or 12?

Over the last 3 years, I’ve come across this problem with a marketing agency, a law firm, and an HR consultancy. None of them knew how many real clients they had [i.e. active ones]. In each case, it turned out there was a small gold mine to tap amongst clients who had gone dormant – after being contacted, one even said “Good to hear from you. That was great work you did for us last year, and we were surprised you did not follow up.” !!