My views on the market, tech, and everything else

(I Can’t Get No) Satisfaction and Negative Churn

My grandfather said to me once: “Two negatives can create a positive in English, but what is the point of that?” He continued by saying: “It is clever dialogue when Bert says to Mary Poppins ‘If you don’t want to go nowhere…,’ but that phrasing was intended as humor, rather than to improve clarity.” He was referring to the lyric “(I Can’t Get No) Satisfaction” in a song by the Rolling Stones I was listening to at the time. That story reminds me of a joke:

A linguistics professor was lecturing to his class one day. “In English,” he said, “a double negative creates a positive. In some languages though, such as Russian, a double negative is still a negative. However,” he pointed out, “there is no language wherein a double positive can form a negative.” A voice from the back of the room then spoke up, “Yeah, right.”

I have similar feelings to those expressed by my grandfather when I encounter the business term “negative churn” since it is also a double negative. The origin story of the term “negative churn” I have heard is that it was invented by Bessemer Venture Partners. My guess is that Bessemer had a portfolio company that was losing customers due to customer churn, but the business was gaining enough revenue from existing customers that the net change in revenue was net positive. In short, the tern “negative churn” was probably created as a form of Kryptonite to downplay the existence of customer churn. It seems clear that some people believe that negative churn is useful in talking up the value an increase in net revenue to counter the existence of customer churn. This Kryptonite-style rhetorical approach reminds of the way doctors use Latin terms to impress patients or some people use terms like “adjusted EBITDA before items.”

Derek Pilling identifies another problem wit the terms negative churn here:

“What if the revenue gained from existing customers who purchase more over time is less than revenue lost from existing customers who purchase less over time. Such a scenario would indicate that Negative Churn is NEGATIVE! Can Negative Churn be negative? Should we call this scenario Negative, Negative Churn? Of course not.”

The simplest way to describe what some people call “negative churn” would be to rename it “expansion revenue” or “account expansion.” There are a few helpful of blog posts and articles on the internet about this type of revenue written by several people including Lincoln Murphy. After including this tweet in his essay:


Murphy writes that businesses which know how to grow expansion revenue are:

“really good at promoting additional use that requires more resources or additional functionality which expands revenue, as well as using up-sells and cross-sells to grow per-customer revenue. In my experience, since SaaS providers with high churn rates simply don’t do the things necessary to reduce churn in the first place, they rarely do things necessary to expand revenue at all, let alone enough to overcome what is lost through churn. From what I’ve seen, providers with an unacceptable SaaS churn rate don’t know how to – or for whatever reason simply choose not to – leverage up-sell, cross-sell, or even down-sell opportunities, so their churn rate is rarely offset by expansion revenue.”

What are some examples of expansion revenue? One example of expansion revenue is upselling and another is cross-selling.


I will discuss upselling first, which happens when a business sells a higher-end version of the same product. David Skok explains upselling quite simply using an illustration:


Joel York writes:

“Upselling increases the average recurring revenue per customer over time. It is an increase in price, not volume. SaaS churn is fundamentally a volume problem, negative virality. Upselling is better viewed as a transition from one average price scenario to another. It can soften the impact of churn, but it cannot reverse it.”

A so-called freemium business model is just one example of an “upsell” or cross-sell. The freemium model is classically illustrated here in a South Park episode:


And here:


One unsavory and sometimes illegal form of upsell is the bait and switch, which is a sales tactic that lures customers into a selling environment with low prices on unavailable items with the aim of upselling them on a more expensive product. If the seller has the “bait” in stock and is willing to sell it but talks the buyer into buying something else, that is not “bait and switch.” If the seller does not intend to or cannot sell the “bait,” it is often illegal.

In addition to using upselling to increase revenue per customer, the same objective can be created via “cross-selling,” which is selling different products to the same customers. In his post on negative churn Tomasz Tunguz explain that a cross-sell includes selling: “adjacent products that increases the value of the core product (ex. AWS S3, EMR, etc).” Salesforce.com both does cross-selling and enables it in its products:

Some examples of cross-selling include an electronics retailer offering a deal on a computer case, mouse, and screen cleaning wipes to a customer who purchases a new laptop, or an insurance provider offering renters’ insurance to its car policyholders.

One of the more notable examples of a failure related to cross-selling done improperly is the Wells Fargo example. This bank created incentive programs for cross selling that went disastrously off the rails. I could write a long blog post on what went wrong in the case of Well Fargo. Have they abandoned cross-selling? No. That cross-selling remains core to any financial service supermarket’s business model. American Banker magazine notes:

“Wells has stopped reporting its fabled cross-selling ratio, and executives at the bank have spoken about moving from a sales-oriented culture to one that is more service-driven. But one of the clearest messages from the embattled company’s investor day was that selling additional products to existing customers remains a key part of the bank’s strategy. Top Wells executives crowed Thursday about steps they are taking to improve customer referrals from the bank’s sprawling branch network to its wealth management unit. And in what seemed like déjà vu, the bank’s head of consumer lending said that one of his key strategies for building the mortgage business is to attract more of Wells Fargo’s existing customers.

Other businesses in other industries rely on cross-selling in a big way. For example, a software business like Twilio or a hardware and software business like Cisco are successfully cross-selling other services (like video and security features) to customers.  Open View Labs has a slide which illustrate cross-selling:


Upselling and cross-selling are very desirable even if conflating them by using the term negative churn is not. David Skok has created a slide explaining negative churn but underneath it all is really “expansion revenue:


How do you do upsell? Skok again has a simple but clear explanatory slide below in which the key addition to the usual sales funnel (in blue) is an extender for expansion revenue (in green pointed to by the red arrow):


The cost to acquire a new customer by means of an upsell versus a cross-sell are quite different. David Skok points out: The median CAC ratio per $1 of upsells is $0.27, or 24% of the CAC to acquire each new customer dollar. The CAC ratio number for expansions is $0.20, or 18% of the CAC to acquire each new customer dollar. For renewals, it is $0.13, or 12%. These results are substantially similar to previous years.” This survey that Skok is basing his statements on illustrates the difference:


One important aspect of all this is to understand what Bill Gurley said here:

“The LTV formula, when used correctly, can be a good tactical tool for monitoring and comparing like-minded variable market programs, especially across channels. But like any model, its proper use is entirely dependent on the assumptions used in that model. Also, people who have a hidden agenda or who confuse a model with reality can misuse it.”

It is possible to use a churn model that contains equations like this:


And many variables like this:


That something is possible, does not mean it is always useful. It depends. I am sometimes pleased with the result when someone does this sort of very complex modelling work on a variable like churn and even more pleased that it is not me doing that work. The best goal is to be approximately right, rather than precisely wrong. Complex is not necessarily better. A complex model is only as good as its assumptions and the real world is filled with risk, uncertainty and ignorance. Some probability distributions of possible future states of the world are not known and some future states of the world are not known. Uncertainty and ignorance are always present. The real world is rarely merely risky like roulette. Formulas can’t make uncertainty and ignorance disappear.


Customer churn is never good and its negative financial impact can often be nonlinear. Replacing a customer who churns means more customer acquisition cost whereas keeping that customer is usually far less expensive (renewal CAC). Pilling adds to what he said above: “Churn should be analyzed independently from the revenue lift from upsell (or extension) that has the potential to drive organic revenue growth in an installed base of customers. Conflating the two is dangerous.” It is important to start with first principles in looking a these issues related to negative churn as is illustrated in this analogous case from another problem set:


One aspect of modern business and financial terms that isn’t talked about nearly enough is how far behind Generally Accepted Accounting Principles (GAAP) are behind actual business and financial practices. GAAP often still assumes that a business is just like the pin factory that Adam Smith wrote about in his book Wealth of Nations. As just one example, customer “churn” does not have a standard definition. Churn is a measure of negative growth. But the definitions of churn people and businesses use are all over the map. One article on the topic reveals that the authors found: “43 different ways public SaaS companies were accounting for the metric.” If mobile operators and other types of businesses where included the figure would be far higher.

Here are some example definitions:

Retention rate: The probability that a customer keeps his relationship with a business over a given period. The retention rate can be estimated from an analysis of historical data.

Churn Rate: The probability that a customer ends the relationship with a business during a given period. The churn rate is (1- retention rate). In other words, churn is the reciprocal of the retention rate.

To make matters even more complex, there is “user churn” and there is “revenue churn” which are not the same. There is net churn and gross churn. And there is logo churn. Will the accounting profession ever respond to the fact that we have evolved beyond an economy compose of Pin Factories? Maybe. But perhaps I should not complain since understanding the difference between a pin factory and a modern business is a huge source of investing alpha. Endowing a Pin Factory Chair of Accounting and SEC reporting might be a good idea to preserve potential investing alpha for future generations.

This is Charlie Munger on the limitations of accounting generally:

“…accounting [is] the language of practical business life. It was a very useful thing to deliver to civilization. I’ve heard it came to civilization through Venice which of course was once the great commercial power in the Mediterranean. However, double-entry bookkeeping was a hell of an invention. And it’s not that hard to understand. But you have to know enough about it to understand its limitations because although accounting is the starting place, it’s only a crude approximation. And it’s not very hard to understand its limitations. For example, everyone can see that you have to more or less just guess at the useful life of a jet airplane or anything like that. Just because you express the depreciation rate in neat numbers doesn’t make it anything you really know.”

Despite this complexity, some definition of churn must be selected since this variable must be used in a formula that is a helpful guide to running a business when used correctly. Similarly, a definition of revenue per customer must be determined and calculated. Be careful out there.

If you want to learn more about churn you might want to try my classic essay: “Everyone Poops and has Customer Churn” https://25iq.com/2017/01/27/everyone-poops-and-has-customer-churn-and-a-dozen-notes/

End Notes:

















Categories: Uncategorized