(For the last few weeks, I’ve been debuting a series of articles answering a question posed to me by The Ankler’s Richard Rushfield: Will The Irishman Make Any Money? It’s a great question because it gets as so many of the challenges of the business of streaming video. Read the rest here, here, here and here.)
Tomorrow, we start to get data for the Great Irishman Challenge. Well, we don’t, but it will hit screens everywhere as Netflix releases it and presumably places it front and center on everyone’s Netflix homepage. One of our goals with this project is to set our criteria ahead of time, this way we aren’t back-fitting the results to our preconceived notions about Netflix.
Now that we have our models for valuing film explained and re-explained, it’s time to fill in the specifics. Today, I’m going to lay out what we know about The Irishman before it launches. I’ll update some assumptions on the model and revenue streams from some feedback. Then I’ll describe two key inputs for the streaming model—Customer Lifetime Value and Attribution of Subscribers. Plus, I’ll touch on how I plan to triangulate popularity after The Irishman launches, which may evolve as we get more data or potential partners. Finally, I’ll talk about the benefits for this model and how I plan to draw insights from it in December.
Assumption 1: Production Budget
Discussing with Richard, we think this is high. Super high. A pretty good summary of this is Jeff Sneider’s take on his Collider podcast a few months back (episode 11 specifically at minute 54):
“I’ve seen [The Irishman budget] figures from $125, $140, $150, $160, $175, $200 million for The Irishman. If you go on Deadline you can read an article that says the budget is $140 and then two hours later another writer is under a completely different impression and says it’s $200 million. No one is on the same page on the budget for this film. And let me tell you what that means. It means the budget is way f***ing higher than any of you are imagining.”
Gosh, that type of cynicism about PR efforts exactly matches my own. If you hear tons of different numbers that can’t seem to decide how much something cost, well the likeliest option is that it was WAY WAY WAY more. At a minimum, this is a $200 million dollar film. And I’m going to do some scenario modeling up to even $300 million. Which means I’ll split the difference and call it a $250 million dollar movie.
Is this ridiculous? Not so much when you think about it. Consider, what does it cost to get Martin Scorsese, Robert De Niro, Al Pacino and Joe Pesci (out of retirement) on a film set? Especially if you’re buying out all the backend, which Netflix had to do since there aren’t any second window revenue opportunities here. (Which is cool too because it simplifies my model.) If I told you between those four it cost $100 million in talent costs, would you blink an eye? Is $150 million too high? I’m assuming $125 million in talent costs.
Then we can add in the extra production costs. This was a very long shoot. (I saw 300 days somewhere.) And then it was as VFX-intensive as some Marvel movies due to the de-aging process, which also required extra work because initial versions didn’t work. (Sneider lays out this situation with great details in his podcast.) This film required a VFX push to get finished in time for launch at the New York film festival, meaning it ran up tons of overtime. Does that sound like $125 million in costs? Absolutely. If not more.
Assumption 2: Marketing Budget
The Irishman will have two marketing budgets. First, the initial roll out. I looked for estimates online and didn’t find a ton. That said, I’ve seen billboards, online ads and even commercial spots. Which screams definitely something, but less than a franchise tentpole roll out. I’d say it’s probably between $50-$100 million, and since I went high with the production budget, I’ll go low here.
(Also, to echo Richard Rushfield’s “see something; say something” if you know a better number for the marketing budget, shoot me a line.)
Then we have the Oscar budget, which is a little bit harder to disentangle. Already, Netflix has started their awards campaigning, but has specifically tied many of their films together, from A Marriage Story to The Irishman. We know from Richard’s reporting that Netflix likely spent over $50 million on Roma’s Oscar campaign last year. They look likely to beat that again. The question is, will they spend $50 million just on The Irishman, or split it with A Marriage Story? Both are getting rave reviews, and I think Netflix is desperate for a Best Picture win. I’m going to end up calling it about $40 million for The Irishman alone.
Assumption 3: Profit Sharing and other revenue streams
We have a few categories here, so let’s run through them.
Library Value? Yep.
I added library value, assuming the retention model is the equivalent of the theatrical window. Meaning it sets the “price” of the film. Then, we can our theatrical financial model to value library windows. Meaning, the “value” to Netflix for the film after the initial release. To provide an example, Bird Box got most its viewership in the first month, but folks will keep watching it out on Netflix for years. That has a value, which is the “library” value. Using my theatrical model, I’m assuming library value of 25% of first window value. (Specifically, the 10% “digital” revenue for theatrical films is about 25% of the the free, cable, syndicated TV, pay TV and digital second window buckets.)
Box Office Bump? Yep.
If a Netflix film wins a Best Picture, or even gets nominated, that will result in boost in viewership. I’ve seen that for past film and TV series for major awards series. For most Netflix films, this isn’t worth a line in the model, but for this one it is. In this case, I’ll use a 25% threshold of the initial window for the Best Picture bump. This will have a “halo” effect on the library window as well.
Second Windows? None.
Since Netflix films are exclusive to the streamer, every other potential window from home entertainment to licensing to cable channels is a zero in my model. This simplifies our model.
Merchandise? None.
Mobster films don’t really sell a lot of merchandise. Especially brand new films without fan bases or cultural cachet.
Distribution Fees? None.
Since there aren’t second window or merchandise revenue to shield from profit participation, I don’t need to model any Netflix distribution or marketing fees.
Talent participation? Some.
Initially, I didn’t have any profit sharing, but then I got a note that Netflix for super-duper-huge stars did put a bonus system in place for feature films. Basically, if you hit a certain viewership level, you get a 20% bonus in your paycheck. So I’ve updated the model with that assumption in place, assuming that 80 million views (or “one Bird Box”) is the threshold.
Assumption 4: Calculate CLV
It’s crucial have to a good estimate for customer lifetime value. These are calculated fairly often by other people (see estimates here, here, or here). My difference is I don’t factor in content costs because I’m trying to value the content itself. If I did factor them in, I’d be double counting content costs, and that’s a huge “no-no” in accounting.
So here are my inputs for CLV. First, blended average price per month comes from Netflix’s 10Ks. Customer retention estimates come from various sources, including Second Measure. Crucially, though, I have a much lower rate for international because I’ve heard the churn machine is very high overseas. Finally, I use other estimates of Netflix’s marketing spend for customer acquisition costs. All this leads us to:
Small changes in CLV can lead to big changes down the road. In particular, if the average months subscribed starts to lower for Netflix, a lot of current valuations will need to be changed. This means if you disagree with my assumptions, you could really change the CLV estimates and hence profitability. For example, customer acquisition costs by some estimates have doubled recently. I’m using the lower number since this is also about retaining customers, and acquiring new customers is what’s getting tougher.
If I had even more data, I could make separate estimates for new customers versus acquired customers, something I did for Game of Thrones, since HBO had a clear growth trend based on GoT’s popularity. But that level of modeling would just introduce more complications, without actually improving the accuracy of the model.
Assumption 5: Attribution Model
In my series on Game of Thrones, I introduced a term called “Magic Numbers”. For brand new businesses, there is usually a key number in the model that determines how everything else flows. And it’s usually the thing you have the most trouble predicting. In this case, the attribution model is the closest thing I have to “magic number”. It’s the piece the model hinges on. I have a couple of sources for this estimate.
First, I have personal experience with this type of model. I’ve seen it in action. I know the rough numbers for some types of film and how the shape of the curve looks. Fortunately, as long time readers can predict, we know what that shape should look like:
The value of a film in retaining customers is roughly logarithmic, meaning a show that is twice as big as another show isn’t just twice as valuable, but four times as valuable, and so on. This is because as the shows increase in popularity—meaning total viewership—they also increase in the percentage of customers attributed to it.
Moreover, we have a few data points that have informed this. The best estimate comes from Netflix itself. Well, not Netflix per se, but a leak about Netflix. Thanks to The Information, we know that Triple Frontier, a film that cost $115 million and got 52 million viewers…but still wasn’t profitable. Which means that we know our attribution model shouldn’t be enough at 52 million to give Netflix breakeven for Triple Frontier.
Here’s my current working attribution model, and I’m going to keep it the same for territories, but divide viewership by territory based on interest levels.
And how that looks in graph form:
As a fair warning, this is the biggest assumption in my model for The Irishman. They are the “magic numbers” of this series. Netflix could likely tell us how accurate I am; they never will though.
Assumption 6: Theatrical Comparables
One of the things that drew me to the Great Irishman Challenge was the idea that once I built the competing models, I’d be able to compare the streaming-only launches to a hypothetical theatrical release. And not just a limited-theatrical run without data, but a true theatrical run going out to thousands of theaters.
To do that, we need a set of film comparables. I like to do high, mid and low cases. I’ve also included multiple Scorsese films, since I assume it’s correlated. That yields us this data set to leverage for later analysis:
I have some box office data from The Numbers, but haven’t analyzed it enough to draw conclusions. Right now, it looks like these films skew US (about 50% or more viewership on average), but there are some outliers. Moreover, even at the high end, these films top out at about $300 million in worldwide box office. (The Departed did $291 million in global box office.)
The Final The Irishman Model
Phew. I’m tired. With all these assumptions, here’s how my model looks right now for The Irishman specifically.
As promised in part I, I built a “scorecard” for The Ankler, and here are a few of those pieces:
This is worth digging into, and over the next few weeks I will. But crucially I need to explain where I’ll get my last input.
Streaming Viewership (And How I Plan to Estimate)
If customer attribution is the magic number, well the wildcard here is ever getting the data. Remember, Netflix doesn’t tell us how many people watched many of their films. So here’s our plan for how we’ll calculate that:
Step 1: See if Netflix just tells us. They told us for Bird Box and El Camino, but didn’t for Roma. My gut is Netflix won’t tell us, since this film is so long it will hurt completion rates and as a gangster film it may not travel globally. So…
Step 2: See if Nielsen (or another third party) will tell us. I think this is very likely, since Nielsen knows it will be popular so will get a lot of news clicks. We just have to do some adjustments to get global numbers if Nielsen tells us. If they don’t…
Step 3: Use Netflix datecdotes/Nielsen reports and compare them to Google Trends to make our own estimate. I have a database of Netflix datecdotes (see here or Whats-On-Netflix’s roll up here) and we could pull Google Trends data. In full disclosure, I”ve started this process but it’s a bit noisy. If we have to do this method, I’ll show my work.
Key Insight for Today: Could The Irishman ever make money streaming only?
If you’ve followed with me, the probability that Netflix’s latest epic helps them make tons of money seems, unlikely. The key driver there is that they’d need over 80% of current subscribers in the US and then another 60 million viewers internationally. That would be 30 million more subscribers than Bird Box earned, and that still requires a Best Picture nomination. In other words, a monster, monster hit.
I’ll say this, though, before I started I wondered if a film this big could ever be profitable. And now I think it could be. It just has to be Netflix’s biggest release to date in a huge way. Thanksgiving weekend will tell us either way.
