Prediction Time: Forecasting the Effect of Netflix’s Price Increase on US Subscribers

Netflix moves the PR needle. Even I jumped into the Twitter maelstrom to generate clicks based on their two announcements last week, especially the decision to increase prices on US customers.

The problem, for me, is that Twitter, as a medium, is really bad at digging into numbers. It isn’t Twitter’s fault; spreadsheets just don’t really fit. (See my last big analysis article for another debate taken off-Twitter.)

As a result, a lot of the “debate” on Twitter devolves into “this is good” or “this is bad”, with some anecdotes thrown in and the occasional Twitter rant. The fun thing in the #StreamingWars2019 is we’ve all clearly taken a side and this war will only end with all our heads on pikes. (I’m rereading Game of Thrones/ASOIAF in preparation for April 14th and George R.R. Martin ends lots of events with that outcome.)

We can do better than Twitter debates. Today, I want to make the subtext of all the discussion on Netflix text. I want to change the terms of the debate around Netflix by moving into concrete specifics. Strategy is numbers, right? 

That means putting our predictions into quantitative terms. I described my process for this regarding M&A back in July and my series on Lucasfilm. So here’s the question:

How will Netflix’s price increase in 2019 impact US subscribers in 2019?

The results will come in when Netflix announces their annual/quarterly earning in January 2020. For the record, Netflix currently has 58.5 paid memberships at the end of Q4 2018, among three tiers of pricing. Over Q1 and Q2 of this year, they’ll increase prices $1 to $2, raises of 13-18%. 

I’m going to walk through my process to make a prediction. First, I’ll explain why I’m predicting customers in 2019, not other financial factors. Second, I’ll evaluate what we know and some good and bad ways to look at the problem. Third, I’ll talk a bit about the data and finally make my prediction. Feel free to leave yours as a comment on this article or in my Twitter feed.

Stating the Problem: If the number of subscribers who leave is lower than 18%, it’s a win.

This is the simplest of simple microeconomics that Netflix is practicing here. If you raise prices, but the units sold (in this case, customers) decreases less in percentage terms than the price increases, you make money. (Assuming no increases in costs.) Since this is digital and each additional “unit” sold has a marginal cost of zero, that math works. (Note: this is still an “assumption”. If you continue to need a larger and larger content library to woo subscribers, well then our magic “marginal costs is zero” isn’t actually true.)

economics model

Source: EconomicsHelp.org

Like the “value creation” model, the above chart is the simplest explanation of price and supply and how they interact, but it is woefully incomplete. Many, many other variables ultimately impact the number of units sold or customers who subscribe.

Yet, as rule of thumb, it works. The number, therefore, to watch out for is the subscriber growth or decrease. If Netflix decreases its subscribers to 55.6 million paid subscribers, that’s a 5% decrease. Since that is still lower than the 18% price increase, the move made financial sense. Thus, the terms of the debate change to, “will Netflix customers grow, slow or halt?” Here’s the past 7 years of subscriber numbers (paid, US):

subs from earnings reports

Predicting the Effects: How Many Subscribers will Drop from Netflix?

There are a couple of ways to try to triangulate this number, but let’s start with how not to do it.

The Bad Prediction Method: Using yourself as a data point.

Many people when discussing TV or film use themselves as the ur-example of a customer. I saw multiple people say on Twitter something along the lines of, “I use Netflix all the time. I don’t care about a $2 increase. Ipso facto, this doesn’t matter.”

Now, if you are a representative sample size of America, then congratulations. This analogy works. (Also, I have a ton of other questions to ask you. Like who will win the 2020 election? You should know.) If instead, you are a single data point, then we need something else.

The Trust Method: Believe in Netflix’s army of economists.

Netflix has an army of PhDs with mountains of data looking at prices. A lot of them are smarter than me. (They have PhDs and I don’t. That’s easy math.) So they have models from economics and statistics they can leverage to estimate the “price elasticity” of their customers. They’ve likely modified them by territory with different content loads to understand how pricing helps or hurts subscriber growth. So they’ve built these models understanding what the decay could look like from a price increase. Understanding this, we could say that Netflix is fully aware of the consequences and choosing to raise prices for clear strategic reasons, and leave the debate at that.

If we wanted to “trust in Netflix”, we’d say they know what they are doing and leave it at that. That said, as Kevin Drum recently wrote, we should never just believe companies. Moreover, even successful companies make awful decisions. Just believing in the corporation doesn’t work for me. We need another method.

A Better Method: Surveys!

The much rarer, but more useful than the two methods above, is to ask customers. Since the Netflix announcement, we’ve had this in spades. Here’s results from surveys from Hub Research, Streaming Observer and The Diffusion Group asking how customer felt about the price change: 

survey results

Using the simple math, since 9, 3 and 8% are lower than 18%, well this is a smart decision.

The problem with surveys is customers aren’t great at predicting their actual behavior. People are notoriously bad at self-reflection. This could go both ways. In some cases, people may be angry over the increase, but do nothing. Or they may be fine with it, but still end up cancelling when the bill gets higher. Also, the surveys above still have relatively small sample sizes (two surveys combine for 1,000 together.) Moreover, in some cases, the respondents may not even be the true decision-maker. So this is a starting point, but not the end point.

(By the way, the gold standard is a conjoint study, but that’s expensive. But that’s the best way besides experimentation to gauge price increases impacts on customers.)

My back-of-the-envelope method: Look at MAUs or Annual Users

So we can’t use ourselves, we can’t just trust Netflix and surveys are useful but limited. Unfortunately, to really make a prediction here, I’d need to know one more piece of data. And I bring it up because stock analysts should really insist that Netflix releases this piece of data so they can better understand the financial health of the firm. That data is…

What is your Monthly Active User base, by month?

I’ve tried to guess this before. I assume it is high (in this article, I suggest above 50%). There has to be a floor around this because of the global viewership numbers they’ve released on select shows. I’d take the 50% minimum and add about 10%, so the floor of MAUs is 60% of Netflix customers in the US.

We can also guess at the top range. Mainly using the “Netflix tells us good news rule”. Is Netflix MAUs on average above 90%? If it were, they would definitely tell us. If it was above 80%, that’d still be good and they may tell us. If it’s below 80%, I don’t think Netflix would. So that gives us a nice range of an average MAU between 60-80%. If that range seems large, well, see point three below. 

I turned the focus to MAUs because this short circuits the “I would pay for this increase” logic from above. Drop outs from Netflix aren’t going to occur from active users who stop using Netflix, it will come from people who use the price increase as a reason to quit subscribing to a service they don’t regularly use. If you can’t imagine a world that includes people like that, well, that’s why we have surveys in the first place.

So my plan is to use my MAUs, and then make a range of outcomes regarding drop out rate based on that. (Say of non-users, 10-25% will drop out.) However, I do want to look at some qualitative/narrative criteria that could impact where my judgement lands. 

The Caution: Netflix may know this will hurt subscriber growth, but Cash Flow!

Netflix forecast $3-4 billion in free cash flow losses (in their accounting) in 2017 for the fiscal year 2018. They ended up at the low end with $3 billion, and are forecasting $3 billion in cash losses again next year, calling this as the high water mark. Notably, that forecast comes with this price increase in mind. Consider this…

screen shot 2019-01-22 at 3.36.55 pm

In other words, based off assumptions that nine months of price increase at an average of $1.50 per customer (which could be a pinch high, with some customers downgrading), you get roughly a $820 million dollars in extra cash. Since Netflix is forecasting another $3 billion in free cash flow losses in 2019 (per the last earnings report), that means without the price increase, you would see FCF losses climb to $4 billion in 2019. In other words, they need this price increase to keep FCF losses under control.

Unlike past years, more and more market observers are watching this cash burn. By finally getting control of losses (holding steady at $3 billion) Wall Street will likely stay on the Netflix bandwagon. If losses had increased to $4 billion, would investors have the same patience? If your FCF losses increase by a billion dollars every year like clock work, it gets tough to say it won’t just keep increasing forever without serious pain.

This would be the biggest argument against the Netflix PhDs and finance guys who approved the price increase. Netflix may really, really want subscriber growth–and so does Wall Street–but are willing to take a decay in customer growth simply to shore up the losses. This may be a protective move for the stock price, sacrificing US growth for cash, while focusing on international gains.

Netflix could be in for a year of flat subscriber growth with this move…and they know it…but they had to do it anyways.

Some Data Lessons.

Before I make my predictions, I want to explain the difficulty in making this prediction. Consider this the analysis of the data set before us, and it applies to everyone, including people at the company.

First, this is a classic “small sample size” problem.

Frankly, Netflix doesn’t have as a large of a sample size for their economists to leverage as I made it appear above. Sure, each year gives you a quarterly subscriber number. So 2008-2018 is 11 years by 4 quarters, or a sample size of 44. Except, the initial growth really isn’t representative. They transitioned a lot of people over from DVDs. We could look at monthly data, which would multiply it by 12, but that’s probably too granular for how often Netflix changes prices. If we just went by price changes we are at a sample size below ten. So no matter what model you build, you won’t have a lot of data to leverage.

Second, there are multiple variables to consider, in addition to price.

The problem with those fancy models is they may not have priced in the value of all that content correctly. If the model makes the assumption, “We can replace Disney/Warner streaming content with originals” and it turns out their originals aren’t popular with people (though still popular with critics), then the double whammy of price increases and content loss could lead to a drop out. Moreover, new services are entering that will compete for share of wallet. All these factors should work together to make us very uncertain here. 

Third, be humble with your predictions.

So for everyone—myself included—who is willing to make a prediction: don’t! Or do so with a huge grain of salt. Your “90% confidence interval”—hat tip to Kahneman, Tversky, and OB professors everywhere—should be really large. Again, if the goal is to make a prediction that is right 9 times out of ten, then make the range very large. 

Finally, my predictions.

I have two. First, many of the people debating this on Twitter will not actually make a quantified prediction. They will say something like, “This price increase will not materially effect Netflix’s subscriber growth.” With a prediction like that, nothing is quantified. Even if they are wrong, how could I prove it? There is so much wiggle room they could easily debate the facts.

Still, I think we’re smarter when we at least try to quantify our predictions. It forces us to think through the issues. So as foolish as it may be, I’m going to make a prediction. For the upper range, Netflix saw a real world subscriber increase of 11% in 2018. It was 10% the year before that, so I’m making 10% my upper interval on my prediction. In other words, I see a best case scenario where the price increase doesn’t hurt subscriber growth.

For the floor, this is tougher. And this is where I worry about being too confident in my prediction. Again, it is very, very hard to be right 90% of the time. We’re often too over-confident. So I can see an upper growth of 10%, but if the price increase back fires, along with all the content leaving, I could see a scenario where Netflix loses 10% of their customer base. Yikes, that sounds high, but again I want to be right 9 times out of 10 here. Here’s a table with potential % drop out rates to judge for yourself:

2019 subscriber range

Now to the prediction itself. Assuming a 80% MAU rate (on the high end of my range), with a 10% “willing to drop out” gives me 2% growth. That sounds right and gets me right to about 60 million customers. So there you have that prediction.

Prediction: 2% growth in subscribers, up to 60 million customers at the end of 2019.

90% Confidence Interval: 52.6-64.3 million customers in the end of 2019. (That’s subscriber decreases of -10% to 10% growth.)

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: