Inaccurate Churn Formula

Format for Printing

Format for printing

Request Reprints


By AceInMySleeve
July 9, 2004

Posts selected for this feature rarely stand alone. They are usually a part of an ongoing thread, and are out of context when presented here. The material should be read in that light. How are these posts selected? Click here to find out and nominate a post yourself!

Having participated in the Yahoo boards for some time, I must say the return on investment there is quite low. The participation level on this board appears to currently be a bit low, but the quality high, and it would be my hope that we could turn it up a notch. Netflix presents an interesting investment story, not just of the recent year's performance, but I expect also in the year to come.

I have a long position in this stock and feel comfortable with it. I am so comfortable I want to spend some time actually railing against the company in a pet topic of mine. It was picked up recently in a Herb Greenberg article (not as a result of my doing) but otherwise has received little attention.

This is the startling inaccuracy of Netflix's churn formula. One that significantly understates the true absolute value of churn, and in addition, particularly glosses over the impact of high growth on churn. I have attempted to contact Netflix and discuss with them these findings, as has Mr. Greenberg and Mr Von Der Porten (Herb's informant as it were), and one must conclude that they intend to ignore all inquiries. An error in strategy I do believe. Along with the recent terse snubbing of the popular fan site, Hacking Netflix, I am drawn to believe there is a failure to live up to the transparency and accessibility otherwise suggested in the company's financial and corporate communications to the public.

So let me get to the details of this particular issue. Netflix's churn formula is unprecedented in my research. There is not a single churn calculation method used by all subscription-type companies, but there is no other company I could find that uses this particular one. Since a new trial subscriber that fails to convert to a paying subscriber is counted as a churned subscriber, Netflix seeks to include all new trial subscribers in the denominator of their formula. There should be no ideological problem with this goal. However, in implementing this goal and attempting to remain simple, they over-simplified. There is a flaw in the formula which leads to increasing inaccuracy as true underlying churn increases and/or growth rate increases. Netflix has both a high churn and high growth rate. The conclusions I have found are two-fold:
1) An accurate measurement of churn shows an increase from Q403 to Q104, not a decrease.
2) In absolute terms, the churn measurement is roughly 20% inaccurate to the low side for Q104.

I've struggled with how to present these facts in a readily digestible manner without extended math contortions. The intuitive explanation of the problem is this:

All gross sub additions are treated precisely the same as subscribers that carryover from a previous quarter (look at denominator). In reality, these subscribers continuously join the company over the quarter. The inaccuracy comes from treating all new gross subs as having had 3 full months in which to churn away, whereas in reality, the average is much closer to 1.5 months.

The bottom line suggestion, to retain simplicity and significantly boost accuracy would be to change the churn formula from:

Cancelled Subs / (Starting Subs + Total New Trial Subs)


Cancelled Subs / (Starting Subs + Total New Trial Subs / 2)

The denominator in the second formula much more accurately represents the weighted average subscriber count. Weighted average subscriber count is frequently the denominator in other companies' churn formulas. I can quote one here for reference from DirecTV's quarterly SEC filing:

"Average monthly subscriber churn represents the number of  DIRECTV U.S. subscribers whose service is disconnected, expressed as a percentage of the average total number of DIRECTV U.S. subscribers. It is calculated by dividing the average monthly number of disconnected DIRECTV U.S. owned and operated subscribers for the period (total subscribers disconnected during the period divided by the number of months in the period) by average DIRECTV U.S. owned and operated subscribers for the period."
Note that this method does not allow anyone in the public to readily calculate the figure themselves, but it stresses the use of average subscriber counts.

Let me now supplement this discussion with a churn figure comparison:
Netflix Value My Value
Q403: 4.76% 5.46%
Q104: 4.67% 5.62%
Change: -.09% +.16%

And note that these figures are roughly 20% different. Additionally, these are both discrete formulas that use linear approximations of curves over 3 month intervals. Continuous formulas both increase the absolute error rate, and the increase of churn from Q403 to Q104. I think the added complexity of a continuous formula is not required to make the point however.

Think also about how intuitive and understandable these results are. Q1 represent enormous sub growth (50% gross adds compared to 34% in Q4). Since we know that trial subs convert only at a 90% rate, one would expect churn to increase in this situation.

Now before concluding this outrageously long post I'd like to mention a few things:
1) Churn is used subsequently by analysts to compute such things as average sub life time. The difference suggested here goes from 21.4 months as suggested by Netflix's formula to 17.8 months for mine. This is a material margin of error for valuation exercises.
2) Netflix's CFO has previously reaffirmed that the churn calculation is sufficiently accurate. It is my opinion that he has mistakenly compared Netflix's reported churn for all subscribers to the true value of churn for just paying subscribers. This is a mystery only he can answer.
3) Since the error introduced comes from the combination of high churn and high growth, both of which are intended to decrease for Netflix over time, it is my belief that they intend to ride it out.

Well, I will conclude here. There is more that one could talk about but it's a start. It does not change my position as an investor but I won't deny very much wanting to see Netflix confront and fix this issue.

Become a Complete Fool
Join the best community on the web! Becoming a full member of the Fool Community is easy, takes just a minute, and is very inexpensive.