Tuesday, 29 May 2018

New valuation metrics for tech companies, and Spotify

There is a very old saying in markets - that the four most dangerous words in the English language are 'this time is different', and that is particularly true when the bulls start advocating for the use of new valuation metrics over tried and tested favourates, as prices outrun anything remotely resembling what can be justified by the old. This sort of soothsaying is long discredited for good reason - is has often accompanied major market tops, and been revealed to have been folly in the cold, sober light of the ensuing bust.

However, with that context and caveat firmly in mind, I believe a sound argument can nevertheless be made that the economics of select (and by no means all) new-era software businesses really are fundamentally different to old-world (non-digital) businesses, and therefore that their financials need to be interpreted - and their stocks valued - with these differences in mind. And when such differences are properly understood, it becomes clear that certain businesses (again, not all) can in many instances be revealed to be reasonably priced despite the fact that they are losing prodigious amounts of money.

At this stage in the cycle, where tech stocks are already relatively expensive, the forgoing may have more utility in cautioning against shorting certain apparently bubblesque names, rather than justifying going long. Many smart VC operators and other tech enthusiasts have long figured out the foregoing, but many traditional value investors have not. Some of them (e.g. Einhorn) are well into the process of blowing up their funds shorting these sorts of names as a result.

The fundamental difference with software/online/app-based businesses, as compared with traditional businesses of yesteryear, is that the internet has removed nearly all of the traditional barriers of time and distance, which has resulted in businesses being able to instantly target a global market with little to no marginal costs of additional product/service delivery. This is because many online services are, for the most part, delivered by a computer algorithm where the marginal costs-to-serve is basically zero, and incremental gross margins are therefore close to 100% (in practice, somewhat below due to marginal web hosting and other variable technology infrastructure costs).

In addition, many of these industries are 'winner takes all' networked businesses where the first businesses to get large and dominate a given industry is likely to monopolise industry revenues and profits long into the distant future. Economies of scale is not a new concept, but the economies of scale for such businesses can be unprecedented in magnitude.

What this means in practice for a software business in the early-stages of S-curve adoption is that revenues and gross profits are apt to grow exponentially, whereas support costs (e.g. technology and back-end systems development) are likely to grow more linearly. This divergent geometric trend in revenues/gross profit and support costs means that loss-making business, provided they can continue to scale revenues exponentially, have the potential to reach a point where profitability literally explodes upwards. This payoff profile is fundamentally different to old-world industrial or consumer businesses, where revenues and costs scale much more proportionately to one another (as market access can generally only be grown incrementally over time in the form of physical distribution networks - e.g. by opening more stores). In addition, for the latter businesses, high levels of profitability were required finance the significant capital expenditures and working capital investments required to expand market access to perpetuate growth, whereas in many cases, technology businesses are 'capital light' and face no such barriers.

Furthermore, a significant fraction of development/support costs need to be incurred upfront (software development, as well as customer & merchant acquisition), and are also incurred largely independent of the level of revenue/gross profit being generated (and in the case of customer/merchant acquisition costs, they are generally a function of new customer additions, whereas revenues/gross profit are a function of cumulative customer additions). Consequently, a fast growing SaaS or online platform business, for instance, may be booking significant operating losses, but provided the unit economies are attractive (LCV/CAC - or lifetime customer value over customer acquisition costs), and the business has the capability to exponentially scale into a large addressable market, current accounting profits can actually be a highly misleading indicator of the value of the business.

A better framework, in my submission, than current multiples of revenue, earnings, or book value, is to instead value the business using the following approach: (1) estimate the total revenues/gross profit that could one day be generated if a realistic proportion of the total addressable market is captured; (2) estimate what the support costs will likely look like at this point in time, relative to that level of gross profit, and hence estimate the overall potential profitability of the enterprise at this latter date; (3) estimate a conventional terminal PE multiple that is likely to prevail at this point of time when the business is approaching maturity; (4) estimate how long you think it is going to take to reach that point, and discount that estimated terminal value back to present value at a realistic discount rate; (5) adjust for any interim capital raisings necessary to fund the business to positive FCF; and (6) perhaps most importantly and frequently overlooked, probability-weight the above outcome.

By way of example, let's take a business like Spotify (SPOT US). The stock has a market capitalisation of US$28bn; net cash is US$0.5bn, and the company is expected to generate US$5.2bn in revenues this year; US$1.3bn in gross profit (25% gross margin); and US$400m in operating losses (i.e. SG&A costs are about US$1.7bn). On conventional valuation metrics, this stock looks very expensive, at 5x forward revenue, with the enterprise making consistent and sizable losses. However, alternative metrics suggest that there is still a plausible value-based bull case for owning the stock.

SPOT has an extraordinarily scalable business and is emerging as the a clear market leader in an industry where the long total addressable market could prove to be a not immaterial fraction of the entire world, assuming paid streaming eventually becomes a widely-adopted consumer service, and the means by which most consumers access music. Furthermore, absent record label collusion, the company's 25% gross margins could quite well prove sustainable, as the bulk of the company's cost of sales are royalties paid to record labels/artists, and those key record label suppliers are major shareholders in SPOT. Furthermore, as SPOT's size and dominance grows, having access to SPOT's distribution platform will become as important to the record labels as it is for SPOT to have access to those record labels to attract listeners (SPOT could come to drive the bulk of music industry revenues, and with a large installed user base, being dropped by SPOT could result in a severe loss of revenue/listening market share for the relevant label, which would risk artist defections and a loss of discovery of new artists).

SPOT currently has about 75m paying users, but as the market leader in an exponentially growing industry - online music streaming - let's hypothesise that they can get to 1bn paying users long term at maturity (about 13% global population market share). At a bit under US$10 a month, or say about US$100 per person per year, that is a potential revenue base of US$100bn, and at 25% gross margins, a potential gross profit pool of US$25bn. How are SG&A costs likely to scale during this growth? They are likely to grow at a much slower rate given the tremendous economies of scale associated with many of SPOT's support costs. Let's say they grow 3x in size, to say US$5bn. This would leave SPOT with operating profits of some US$20bn, or say about US$15bn after tax.

At maturity, the stock would likely trade at about 20x earnings, given the business' global dominance and capital-light earnings, suggesting its market capitalisation could reach as high as US$300bn, or 10x the current level. Furthermore, with the company currently being US$0.5bn net cash, and losses declining, it seems unlikely major further equity capital raisings will be needed. As a counterpoint, we will ignore any positive cash flows that might be generated prior to our terminal year.

How long will it take to reach 1bn users? Let's assume 10 years, and lets discount that back to present value at 15% pa. That would suggest a present value of that US$300bn terminal value of US$75bn.

However, we also need to probability-weight this bullish outcome, because it is far from 100% assured. Spotify is competing with heavyweights such as Apple Music and Amazon, and it is possible SPOT will be unable to differentiate itself sufficiently long term - particularly if Amazon starts giving away music for free to its growing army of Prime members; signs all labels and develops comparable playlists and user algorithms that counter any customer 'lock in' SPOT might be able to secure. Apple is less of a threat as the majority of global smartphone users are Android users, although Apple Music may well secure a meaningful share of high-value Apple users.

If we ascribe a 50% probability to success, and for simplicity, a 50% probability to wholesale failure (i.e. no intermediate states of lesser success), we would derive a valuation of US$35-40bn, which is in the broad ballpark of SPOT's current market cap. One can play with any of the above variables, adjusting the probability of success; terminal user estimates; the timeframe needed to get there; the discount rate; and terminal operating margins, to estimate valuations. If I run various scenarios, I reach the conclusion that SPOT is approximately fairly valued, and may in fact still be somewhat cheap.

However, SPOT's exact valuation; investment merits; or whether it will succeed is not really my point. My point is that the levers of the stocks' valuation have exceedingly little to do with current earnings, and far more to do with the probability of success, and the size and timing of the potential payoff if things work. This highlights why markets are not paying a lot of attention to short term profitability for these types of stocks, and are instead (appropriately) paying far more attention to trends in paying users and gross profit margins. It is not that trends in short term SG&A support costs are irrelevant; it is that they are a far less important driver of long term valuation outcomes. This is sometimes why we see tech names deliver outsized share price gains in response to favourable subscriber growth trends (e.g. Netflix earlier this year), even if short term earnings miss expectations.

Understanding this framework is important - particularly if one is inclined to short scalable tech businesses that seem expensive on current earnings metrics. Einhorn is one such investor, and he really needs to figure this stuff out as soon as possible and close his disastrous 'bubble basket' tech shorts, which are proving ruinous to his fund. The fact that he does not understand the above is evident not only in the names he has chosen to short (and the considerable money he has lost doing so), but also his comments in his quarterly letters, where he has expressed bewilderment as to why these stocks continue to go up despite missing quarterly estimates. He is entirely missing the point.

Whether one is optimistic or pessimistic and is inclined or disinclined to go long SPOT or other high-flying tech aristocrats is a different issue (I am not long or short), but one should be extremely weary of going short these names, in my view.


LT3000








11 comments:

  1. Great article. XRO is one i have long struggled with. It didnt make any sense on traditional valuation metrics when it was $10.. and it still doesnt at $40! But obviously i have been missing something, like you say, it costs the business next to nothing to add an additional subscriber, and once they are on the platform they are fairly sticky, the lifetime revenue generated per subscriber is key. Once the subscriber level hits a critical mass the business basically runs itself.

    ReplyDelete
    Replies
    1. Hi Pete,

      Indeed - I looked at XRO at NZ$20 and almost bought some, but couldn't quite bring myself to pay 5x sales. In hindsight & on reflection it was probably a mistake (and it will probably be a mistake not to buy SPOT as well).

      XRO illustrates the points in the article well and I considered using XRO as an example as well. XRO's losses reflect the expensing of upfront customer acquisition costs, yet lifetime customer value is a robust multiple of CACs because customer retention/lock in is very high. The losses were therefore illusory & actually just a function of accounting policy. If you spend $100 to make $500 over the next 7rs, that's called an investment, not losses. If you expense the full $100 in year one you will book losses. But that's like writing down all your capex to zero in year one. It makes no sense.

      The company highlighted this clearly by providing its lifetime customer value metric (number of subs x ARPU / churn). Tracking this metric (along with LCV/CAC) was the correct way to value this business (and still is).

      Cheers,
      Lyall

      Delete
  2. Hector Mustache10 July 2018 at 01:14

    Very true, Shorting is not necessary in my opinion. Plenty of money to be made on the long side.

    Since you follow Einhorn and took a look at Brighthouse Financial, it tanked further significantly since your last valuation.

    ReplyDelete
  3. This comment has been removed by a blog administrator.

    ReplyDelete
  4. Regarding the Spotify example. 1bn users being 13% world population market share (1.0/7.6). That is a true fact, however, half the world live in poverty. So let’s half the world population who could potentially afford $10 per month. 7.6/2=3.8. Let’s takeaway another 1 billion (completely made up figure) for the people who either can’t afford $10 per month, would rather use that for something else or who don’t like music which gives a generous market of 2.8billion users. Then Spotify has a market share of 36% (1.0/2.8) – what about Google Music, Apple Music, Amazon Music and others? Surely they will control more than 65% market share? Competitors may drop prices in an effort to win market share.



    These “modern metrics” are good enough to tell a potential investor a good story. It all ends the same.

    ReplyDelete
    Replies
    1. I don't necessarily disagree with you that 1bn users is an aggressive assumption. There's certainly no margin of safety built in to using that as one's base case (or even bull case). The example was used mostly to make a point about scalable tech businesses. I don't own SPOT & agree there are things that can go wrong & growth (& margins) that can disappoint. Cheers

      Delete
  5. You're double-adjusting for risk of failure by using a 15% discount rate and a 50/50 probability weighting of success/failure. Surely a music subscription customer with recorded preferences is quite sticky and given costs are almost entirely variable, this company must be less exposed to systematic risk than the average. Your discount rate should thus be closer to 7-8% than 15%!

    ReplyDelete
  6. thanks for writing this up - great post. Biotech investors (for the better) are forced to think this way.

    my only question is highlighting the point Jamie is making. Assuming your initial assumptions are reasonable, Is the "double" adjustment a bit too harsh?

    Thanks.

    ReplyDelete
  7. Thanks for sharing these great tips. Tech companies need to continue improving their products and services if they wish to keep their customers.

    ReplyDelete