Catchy title, isn't it?
After a very good Q3 earnings call by Sirius XM (NASDAQ:SIRI) on November 1, the share price took off through the day's trading, zipping from $2.80 right on up to $2.96 on over 100 million shares traded. Amazingly, at the end of the day's trading, a large block of 5 million shares was bought above the asking price, and then similar volume was subsequently sold into the market, driving the share price right back down within minutes to the day's opening price to close flat.
It has many investors going "hmmm" and looking for answers as to why. Of course, answers abound, and everyone has an opinion. I'd like to focus on one opinion as posted within the comments section of my article here.
I certainly appreciate the comments above, misguided as they may be, because they bring up a very good point. Is Sirius XM's increased churn from 1.9% to 2.0%, discussed on the Q3 call, cause for concern? I don't think it is, and I'll explain why.
First, how does Sirius XM calculate churn? From the most recent 10-Q:
Average Self-pay Monthly Churn is derived by dividing the monthly average of self-pay deactivations for the quarter by the average self-pay subscriber balance for the quarter.
This sounds simple enough. Take the deactivations from the self pay pool, divide by the subscriber balance, and you have churn. Many assume that a stable churn means proportional deactivations to the subscriber base and that assumption is generally correct, but there are problems here in placing too much emphasis on a single tenth of a percent movement in this metric, which I will explain below.
First, it must be understood that churn is based upon two numbers that move independently of one another. Self pay subscribers are subscribers who have elected to pay for the service of their own free will, and this number generally increases over time. The rate of increase is influenced by the conversion rate of promotional subscribers, as well as the size of the pool of promotional subscribers. Already we can see that self pay subscribers is a moving number affected, generally, by new car sales. But because the number is only taken after conversion, increases in new car sales are not immediately added to the pool of self pay subscribers. This lag can be as little as three months, or as long as 12 months.
This is important point one to understand. Increases in new car sales do not "convert" to self pay subscribers for three to 12 months; thus, the self pay subscriber total is effectively lagged.
Second, it must be understood that a self pay deactivation is taken as one churned customer as soon as that customer deactivates. There is no lag time here. If a customer deactivates on August 1, that customer is considered "churned" immediately, and is not subject to the three to 12 month lag time that a new subscriber is subject to before conversion.
And that change in time, or delta T, will cause problems for those looking at churn as a simple data point.
Consider the following scenario, and its effect on that churn figure.
Sirius XM customer Joe owns an older vehicle. Joe is in the market for a new vehicle, and purchases a brand new 2012 Ford Fusion with a paid Sirius XM trial for three months. Joe has now sold his old vehicle, cancelled his previous subscription, and is still a Sirius XM customer, but is now under the paid promotional period. Three months later when the trial period ends, Joe converts to a paying subscriber.
Let's look at how this impacts churn in the month that Joe buys the car:
- Joe cancels his old subscription, adding +1 to deactivations.
- Joe is still a subscriber to Sirius XM, but does not add to the paid subscriber total, and is actually removed from the paid subscriber total.
See the problem? In the month that Joe purchases this new car, Joe adds to the numerator (self pay deactivations) and subtracts from the denominator (self pay subscribers) to double effect.
While this is not a problem if the rate of new car sales is completely flat, this becomes a problem as new car sales ramp up to greater levels. Because of the lag effect caused by the promotional period as more subscribers sell old cars and buy new cars than in previous quarters, the rate of conversion lags the rate of sales. The problem compounds as more and more individuals subscribe to Sirius XM, because more and more subscribers will be selling old cars to buy new cars.
Thus, as auto sales increase, the churn rating for Sirius XM should likewise increase proportionally due to this lagging effect. Investors should fully expect Sirius XM to struggle to keep churn ratings at 2.0% or less as auto sales increase over the next several years, and as more individuals sell or trade in vehicles and switch from self pay to paid promotional trials.
Another very big issue with churn is the fact that it is reported in one tenth of a percent increments that are rounded. Why is this an issue? Because one tenth of a percent is an absolutely huge number of subscribers, and it may appear that an increase of 0.1% in churn means that 19,000 of the roughly 19,000,000 self pay subscribers are now churning out of the service. This, though, is the problem with rounding numbers.
How large can the discrepancy be between 1.9% and 2.0% when rounding is taken into account?
Consider that the lowest a reported 1.9% can be is 1.85%, which would be rounded up, and the highest 2.0% can be is 2.04999%, which gets rounded down. Now consider the fact that churn can be 1.94999% one quarter and be reported as 1.9%, and churn can be 1.95% the next quarter and be reported as 2.0%. Given these rounding issues, the rough difference between the two reported numbers (assuming a stable self pay base of 19,000,000 for simplicity's sake) is from 1 to 38,000 self pay deactivations per month. That's an absolutely huge range, and shows the problems with churn being expressed in rounded tenth of a percent increments.
And the final issue with the churn metric may be evident from the example of Joe. Churn is not a measure of individuals who choose to cancel because they are not satisfied with the service. Because all self pay cancellations for any reason are included in churn, it includes an entire collection of individuals: some who choose to cancel because they are not satisfied, some who choose to cancel for financial reasons, some who choose to cancel because they switched radios or cars, etc., etc. When I cancelled on my Sirius Stiletto radio and activated my Lynx in January? I was added to churn for Q1, yet I am a very satisfied Sirius XM customer.
Because there are so many moving parts with the churn metric, it should be understood by investors that movement within a range is to be expected. My personal expectations are from 1.8% to 2.0% churn as reasonable, and for churn to possibly increase as time goes on through a natural effect of a larger subscriber base, compounded with the effect of lag time caused by increased new car sales on the base of established customers. I would not be surprised to find that churn increases to 2.1% sometime in 2013 or 2014 if increases in auto sales and increases in Sirius XM's subscriber base continue.
Investors should absolutely not be concerned with a reported increase in churn of 0.1% from quarter to quarter, or even year over year. A change of greater than 0.1% may be a valid reason to question "why?" but a change of 0.1% on a rounded figure is negligible.
The actual churn number that caused an increase to 2.0%? 1.954%, or 5/1000ths of a point from 1.949.
That's less than 1,000 subscribers churned away from being listed as 1.9%. Given the increase in auto sales seen this year, it falls within the realm of "completely and absolutely reasonable and acceptable." And as Seeking Alpha user hernje points out here, the churn rating for the most recent month, September, came in at 1.82%, or what would be reported as 1.8%. It's an average, folks, and it fluctuates.
So spurn that concern over Sirius XM's churn. There's little to worry about on a 0.1% variance.