Netflix: Cooked? Episode 2
In this blog, we further examine how Netflix accounts for its spend on content and conclude that it's neither conservative nor consistent with its accounting policies.
We cannot yet update this chart for the numbers released by Netflix for 2020, as the preliminary earnings release does not include the work in progress breakdown and we shall need to wait for the 10-K. We have, however, calculated the life without making the adjustment for content in production, which is the red line in the chart above, which continues its steady upward trajectory. The content life continues to lengthen, indicating that earnings are bearing an ever lower percentage of the cost of content production.
The bulls would argue, and we agree, that it’s important to note that there has been a significant increase in subscriber numbers and to acknowledge that this could have the impact of lengthening the life of content. Until we have the 10-K, we would not know if there had been any change in this respect, so we shall assume that Netflix has not changed its policy on content life. There would be no reason to lengthen it further in our view.
Studios were closed during lockdown last year, so Netflix would likely have been unable to produce the content targeted in its original budgets. The line “Additions to content assets” was $11.8bn in 2020, down from $13.9bn the year before – this statistic increased by 7% in 2019, by 33% in 2018 and by 31% p.a in the previous 5 years.
We would expect with a slowdown in the new content spend that content life would turn down not up, as the spend trended closer to amortisation ($10.8bn in 2020 up from $9.2bn). Content spend is still above amortisation, in spite of the cut in spending, and the red line in the chart is flattening marginally (rate of change >12%,7.2%,4,5%).
Now, post the initial 2020 lockdown, all the studios are trying to catch up on content production at the same time. The industry was already crowded, fuelled by Netflix’s appetite for every increasing volumes of content; hence we expect that the cost of production will increase further; our on the ground “research” indicates that production costs are being inflated by
– the need for regular Covid tests for all participants
– the requirement to have sufficient staff in reserve in case of Covid fallout, especially for scenes involving large numbers of extras
– as studios frantically try to catch up on their backlog, demand for skilled operators has increased and supply is being curtailed by Covid fallout, so price has risen. Certainly it’s anecdotal, but one stunt man told me he had had his best ever week in January ‘21
While all this is anecdotal rather than factual, it makes sense and we should expect cost per hour or similar metrics to be higher in 2021 than in 2019, likely significantly so. There will probably have been some impact in 2020 also.
So the latest figures continue to support our contention that Netflix’s accounting for content is less than conservative and that the numbers in its accounts are inconsistent with its accounting policy. What else have we learned since we last published?
I list below comments from Netflix IR in response to our last blog; recap our prior comments on the policy; discuss how I might follow this up if I were still employed in the hedge fund world; and conclude with some thoughts from David Einhorn.
NETFLIX’S RESPONSE
We were surprised and pleased that the Netflix IR took the trouble to respond to the assertions in our previous blog. They made a very fair comment that the tone of one of our comments could be interpreted wrongly which we duly clarified. But the response was essentially to repeat the statements in the accounting policy and to tell us that we didn’t understand. It was a similar response to countless others this analyst has heard in a long career talking to companies about their financial statements.
They highlighted that licenses are typically shorter deals so that as the proportion of original content increases, the life would naturally lengthen. I am not sure that it’s possible to adjust for this as the split is not disclosed and our argument is that
the trend is upward, which could certainly be influenced by the owned vs licensed content split, but in which case why not disclose that or at least give some guidance to an analyst like me
the absolute period is longer than would be possible if the policy were being applied – this is arithmetic, and I doubt that we have got this wrong
even the stated policy could prove aggressive, and why not be conservative in the assumptions you make?
ACCOUNTING POLICY
We reviewed the accounting policy in respect of content in our first blog on this subject. We summarise that commentary in the Appendix below, but our conclusions were that
1 the policy is highly subjective
2 the numbers do not stack up with the stated policy
3 the disclosure that there had only been a write down in Q4 2017 is surprising given the breadth of the content and the risks inherent in its production
OUR CHECKLIST
Having reached a conclusion that there is something suspect or at least odd about the company’s application of its accounting policies, the question then is how to investigate further. In a long only fund, such a discovery might be sufficient for the analyst to abandon research. In a long-short fund, particularly when looking at special situations as I used to, I would ask a number of questions in order to ascertain how significant this was:
1 look at the results – margins and returns on capital – in the context of peer group
2 make a common sense assessment of the spend
3 look at competitors’ amortisation policies and rates
4 review spend KPIs and compare with history and with peers to determine if the spend looks sensible:
a. what is the content per subscriber, and vs history and vs peers?
b. how much content do similar companies carry on their balance sheets?
c. How long does the average subscriber need to stay before Netflix earns the content spend?
5 in the case of content, we might review the spend relative to the viewer engagement
6 are there any trends in original production vs licensed content which might indicate future problems? For example, if a small proportion of the spend but a high proportion of viewership is on licensed content, and competition is increasing, will that push up the price of licensed content and hence overall spend, without any increase in volume or quality of output?
We would like to be able to produce an exhaustive analysis of the media sector and content costs per subscriber globally, but we are not media specialists, we don’t have any of these stocks under coverage, and it would take a huge amount of time and effort. In this respect, we are in a similar position to the private investor, so what can we do?
1. We can look at the annual content cost per subscriber.
2. We can look at the overall level of spending against peers, and ask if the numbers make sense.
3. We may not be able to produce global comp tables, but we can pick out one or two stocks and compare the results and see if they make sense.
4. We can look for evidence from the press, and particularly the specialist entertainment press, as to whether the spend on blockbuster productions has attracted comment in the industry.
5. We can look at Netflix’s spend in the light of conventional broadcasters – even better if they split out the cost of sport and news, neither of which Netflix carries, and see if it is consistent.
6. We can study Netflix’s viewership and see if its customers are watching Friends or The Crown – that may be a clue as to the appropriate amortisation period.
7. We can look to see if Netflix is increasing the proportion of spend on its own productions (it is) which would naturally result in a longer amortisation period.
We have only started this work, but as a taster we found a blog post relating to the first point, the annual content cost per subscriber. One issue here is that very few of the terrestrial broadcasters are directly comparable. The New Moon Capital Blog produced the following data in October, 2019, so it’s a little out of date but still relevant:
-
Netflix has ~150 million subscribers and intends to spend 15 billion on content. This translates to ~$100 per subscriber.
-
Disney, including Hulu, has about 45 to 50 million subscribers, and spends about ~$10 billion on content, once sports right are removed. This translates to ~$200-220 per subscriber
-
Amazon Prime has about 25 to 30 million active subscribers and spends about ~$4.5 billion on content, about $150 to $180 per subscriber.
The data is presented in the chart, and of course the revenue generation effect is quite different, having been higher for Netflix than for others. But annual spend is a slightly misleading metric, as it’s cumulative spend and total invested in content which matters. But this would suggest that the Netflix spend is not unreasonable, although it says little about the accounting.
Source: New Moon Capital Blog
Further work on this and on the other points are on our to do list, and we shall return to the subject in the coming months. We thought it was worth revisiting post the 2020 results to see if anything had changed and it doesn’t appear so, in spite of the fall in the spend. Our central point remains:
The company needs to produce a consistent stream of attractive and even blockbuster new content in order to sustain its subscriber base, particularly in the face of an increasing number of alternatives. Its cash spend of $12bn ($14bn) better reflects the economics than its GAAP financials which continue to appear to us inconsistent with its stated accounting policies – that is our real concern.
WHY IS THIS IMPORTANT?
When we first published this, someone asked why was this relevant – why not just focus on cash flow? Good question, and the simple answer is I like to have a few valuation metrics. For Netflix, EBITDA ignores the spend on content, treating it similarly to capex, which seems unconservative. Sentieo uses an EBITDA number of $15.5bn, but to assess EV/EBITDA, it uses the IBES adjusted EBITDA figure of $5116m (I am not clear how they derive this), to give an EV/EBITDA multiple of 48.4x.
I would use an adjusted EBITDA metric to reflect the content spend which I estimate at $2.9bn, so Netflix is on sub 100x this metric. EV:Sales is another measure I like (and its recent popularity is likely attributable to the number of loss-makers in the US market) and on this basis, Netflix is quoted at a shade under 10x. Steep, but an argument could be made that this is not unreasonable.
The issue on content amortisation is that Netflix is quoted by Sentieo as being on a P/E multiple of 88.9x, falling to 41x two years out – which appears much cheaper than on my adjusted cash flow measure, but I believe that the earnings number being used is artificially inflated by aggressive content amortisation accounting. And earnings-based valuation measures are an integral part of my process which includes an element of triangulation.
The idea behind these blogs, therefore, is simply to explain my approach – the intention is to illustrate the process in order to help people evaluate this for themselves and apply the techniques more broadly. I am now in the education business, not at a hedge fund. I am not trying to value Netflix, simply to explain how I think about the valuation. Hopefully, it helps some people.
FOOTNOTE
We shared the original blog with David Einhorn of Greenlight Capital, a well-known Netflix sceptic, who not only kindly responded but very generously allowed us to quote the question he asks about Netflix:
“The data we see suggest that for many shows the viewing is extraordinarily front loaded. I have often asked groups of people a series of questions:
Raise your hand if:
1. You have access to NFLX
2. You watched NFLX original programming (as opposed to licensed content) in the last 3 months.
3. In the last 3 months, have you watched NFLX original programming that was more than 1 year old at the time you watched.
In large groups….about 90% are yes to 1, about 60% are yes to 2, and about 5% are yes to 3.”
This of course is hardly conclusive evidence, but we remain of the opinion that the Netflix content accounting policy looks aggressive and the data remains out of line with its stated policy. If Netflix is cooking the books, that would be troublesome, and would be surprising given its stellar operational performance. But it would certainly affect valuation perception and we therefore plan to return to this, to continue our investigative points above and to review the extra information in the 10-K.
Don’t miss the next exciting instalment!
APPENDIX
ACCOUNTING POLICY
In case you missed the first Netflix blog, let me just go through again what we found in the accounting policy, with our comments in italics:
STREAMING CONTENT ACCOUNTING POLICY
The first paragraph is clear:
“We acquire, license and produce content, including original programming, in order to offer our members unlimited viewing of entertainment. The content licenses are for a fixed fee and specific windows of availability. Payment terms for certain content licenses and the production of content require more upfront cash payments relative to the amortization expense. Payments for content, including additions to streaming assets and the changes in related liabilities, are classified within “Net cash used in operating activities” on the Consolidated Statements of Cash Flows.”
We would expect original programming to have cash upfront as Netflix finances production – that is quite normal. It’s unclear why this would apply to content which has been licensed and why there would be more upfront cash than amortisation, given that you would expect them generally to pay for a 5 year license in 5 equal instalments which would surely exactly match the amortisation, if it were being done on a straight line basis. Overall, there may be an element of upfront payment where Netflix is financing the production of licensed content, or where the payment is made at the start of the year and the contact year spans two financial years, but broadly license payments should be closer to amortisation over time.
If Netflix were using accelerated amortisation for licensed content, the cash spend would be lower than amortisation in the early years and higher in the later years and over the whole portfolio you would expect it to balance out.
“We recognize content assets (licensed and produced) as “Non-current content assets, net” on the Consolidated Balance Sheet. For licenses, we capitalize the fee per title and record a corresponding liability at the gross amount of the liability when the license period begins, the cost of the title is known and the title is accepted and available for streaming. For productions, we capitalize costs associated with the production, including development cost, direct costs and production overhead. Participations and residuals are expensed in line with the amortization of production costs.”
This seems straightforward. Note that the accounting bookings are done on a per title basis.
“Based on factors including historical and estimated viewing patterns, we amortize the content assets (licensed and produced) in “Cost of revenues” on the Consolidated Statements of Operations over the shorter of each title’s contractual window of availability or estimated period of use or ten years, beginning with the month of first availability.”
Again this seems logical and the maximum period is 10 years
“The amortization is on an accelerated basis, as we typically expect more upfront viewing, for instance due to additional merchandising and marketing efforts, and film amortization is more accelerated than TV series amortization.”
More upfront viewing is logical and accelerated amortisation is more conservative so that seems sensible. I would question why a film would be more accelerated than a TV series, although I do not know the viewing figures. Although a tiny fraction of series might have decades of life, on average I would have expected a 5 year old movie to have more life in it than a 5 year old TV series, but lets assume they know what they are doing and I don’t (obviously).
“On average, over 90% of a licensed or produced streaming content asset is expected to be amortized within four years after its month of first availability.”
This clearly is not happening, as we saw in Part 1 of this series, “Netflix: Cooked?”
“We review factors that impact the amortization of the content assets on a regular basis. Our estimates related to these factors require considerable management judgment.”
Most estimates require considerable management judgment, but it’s not a phrase you read very often in an annual report. Refreshing that Netflix are so honest, or perhaps the auditors insisted on the wording. That there is a risk of error requires management to be more conservative – it does not appear that they have been sufficiently conservative.
“Our business model is subscription based as opposed to a model generating revenues at a specific title level. Content assets (licensed and produced) are predominantly monetized as a group and therefore are reviewed at a group level when an event or change in circumstances indicates a change in the expected usefulness of the content or that the fair value may be less than unamortized cost. To date, we have not identified any such event or changes in circumstances. If such changes are identified in the future, these aggregated content assets will be stated at the lower of unamortized cost or fair value. In addition, unamortized costs for assets that have been, or are expected to be, abandoned are written off.”
Netflix has disclosed two impairments that we have found. The Q4 2017 earnings call disclosed that there had been a write off in that quarter (“related to the societal reset around sexual harassment”, presumably House of Cards post the Kevin Spacey revelations) and in the Q3 2015 call Ted Sarandos Chief Content Officer disclosed that “Well, we’ve not had content write-downs”. I find this quite surprising – in the media business, you are bound to have a few flops.
From the 2015 10-K: “In the third quarter of 2015, the Company changed the amortization method of certain content given changes in estimated viewing patterns of this content. The effect of this change in estimate was a $25.5 million decrease in operating income and a $15.8 million decrease in net income for the year ended December 31, 2015.”