Dissecting Airbnb’s IPO Strategy with Wardley Maps
This week sees the launch of the Airbnb IPO. I thought it would be a good opportunity to dissect their strategy — based only on my reading of the prospectus (S1) and other public information. The obvious disclaimer is in place; This is my perspective and valuation and reflects my personal views gleaned only from public information; you need to do your own due diligence before trading in securities or other financial instruments based on anything you read or infer here! (All data is from publicly available sources such as S1, 10K & 10Q and news articles).
For me the intersection of strategy and valuation has always been of interest. This is my attempt to tie strategy and valuation — after all if strategy is not directed at creating value, what’s the purpose? Thanksgiving weekend proved to be a great time to dig into this analysis since we were pretty much home confined — I spent about 8 hours on this deep dive. The first part was the reverse engineering of the strategy, then some imaginary gameplay and finally plugging the numbers into a Google valuation sheet which is derived from Professor Damodaran’s work. There are some things I would do in a more formal strategy session such as scenario planning and some valuation crowd sourcing activities. I will wrap up by describing how I think the Cynefin framework can be applied to both investments and strategy decision making.
Reverse Engineering Airbnb’s Strategy for IPO Valuation
Literally the last flight I took in 2020 was in February this year to London. Since then it’s been pretty much downhill for obvious reasons. In February we heard whispers of this virus emanating from China but nothing much was thought of it. I try to get to London as frequently as possible to watch my blues Chelsea FC. This year I found a great Airbnb about half a mile from Stamford Bridge. For the first time I also took 2 ‘experience’ tours. One was a day trip to Stonehenge and the other was a half-day London walkabout. These experiences are with local people not professional tour operators, they are small and personal and amazing ways to experience the tourist sights without feeling like an exploited tourist. I remember thinking the upcoming Airbnb IPO should be one to watch.
Well that idea was quickly shelved by Airbnb as they saw the entire market collapse with Covid-19. However they recently announced a resurrection and planned an early December IPO. A couple of weeks ago they published their prospectus S1 Registration Statement which is approximately 489 pages of ‘meaty/sleepy’ information. The first thing I do is convert this to a PDF so I can run some more complex searching. Next you may be wondering if anybody actually reads 489 pages prior to making an investment decision. Certainly I don’t — but I found an effective way to scan for information. There are two categories of information in a typical S1 — ‘corporate speak’ (feel good stuff) and ‘strategic gems’. My goal is to digest and seek out the strategic pieces and try to paint a picture of where I think the company can go. For this I use Wardley Maps.
Wardley maps are a favourite of product managers but find application in corporate and business strategy (and many other arena’s). Simon Wardley proposed a method of mapping the value chain of the business on one dimension (y-axis) and then providing for movement on the x-axis to describe how individual components are evolving. You can read much more on Simon’s blog right here on Medium.
It’s worth mentioning that there is no right or wrong map. A Wardley map is designed to communicate situational awareness and strategic intent — by visualizing how value chains change and evolve. I find mapping useful accompanying my DCF (discounted cash flow model) which is my preferred tool to analyze a stock (as opposed to pricing models like p/e). The map allows me to anticipate change and the corresponding impact on value. It’s also a lot of fun because you play General and can design some context specific gameplays.
Here is my Airbnb Wardley map. It may look convoluted at first glance but let me take you through the story. Then I will describe some of the strategic plays I think Airbnb can exploit.
Wardley maps begin anchored at the top with the core users who are hosts and guests. Everything hangs on these anchors and the map then dives into a so-called chain of needs. Right under the anchors is the aspirational (in my opinion) reason for being — to ‘live/experience-like locals’. The needs then unfold in a hierarchy of visibility to the users. Each component has a need for each other and the relationships are shown with the linked lines (albeit simplified to avoid a mess and highlight significance). Therefore you can see the technology stack intricacies are mainly hidden from direct user view.
Each of the components like ‘experiences’ can tell its own story and its position and movement over time can be discussed and described. Here you can see experiences are the least evolved over long-term and short-term stays. Maps are useful because you can conduct what-if’s and plot a day when experiences are closer to a commodity and anticipate the implications.
The related components have been grouped into pipelines (the elongated rectangles). I use these to group the components which are driving each pipeline. Later I use the metrics as inputs to valuation drivers. For example Gross Bookings Value is expected to grow 29–32% and is driven by the surrounding components including competition. You can visualize pipelines for all the core drives such as gross margin, G&A, marketing, operations and support and product development. Surrounding these are the components that have the most impact on their outcomes. The numbers on the right are the percentage historic ranges as a percentage of revenue. Product development is treated as a capital expense in valuation and is given a ratio Sales/Capital meaning each $2 of revenues requires invested capital of $1.
A map is not a map without movement. Part of the natural supply and demand economy is components move from left to right. Bottom right shows some of the highly evolved components of the Airbnb platform. This is the place where they have built value over commodity components like Amazon Web Services (AWS) and Service Orientated Architectures (SOA). On the other side of the platform are the more custom built components and I have tried to highlight how they are built over Machine Learning (ML) components. A glance on Glassdoor revals 8 of 46 job openings have machine learning requirements.
Clearly ML is evolving rapidly and is probably well into the product section and heading towards commodity for many use cases. I depict this in orange, and have limited the movement to ML (orange arrows) and ‘insurance/protections’ (grey arrows). Of course there are many other components that will evolve, but I think these two have particular significance. ML is pretty obvious and I think Airbnb is reacting well to this. The risk dimension encapsulated by ‘insurance/protections’ speaks to the myriad of vulnerabilities they are exposed to — including taxes, regulations, cleaning, insurance etc. In most cases the S1 documents call out solid reactions. However I wonder if there is an opportunity to commoditize the approach to these risks perhaps with acquisitions.
Next let’s look at some potential gameplays available to Airbnb and then anticipate how these will factor into value drivers which inturn will drive the valuation models.
Gameplays
Excuse the untidy annotations — but that’s the general idea of Wardley maps, to have an engaged discussion (with myself) on strategic plays.
- The most obvious play which we have seen and known about for years is the network effect of these types of businesses. A virtuous circle exists as more hosts and guests sign-up so the value of the network increases and more will sign-up. This is at the core of the value proposition and provides a competitive protection unlike competitors. Currently capturing 2.5% of the serviceable addressable market the upside is still there.
- Clearly the platform can be leveraged and we see this occurring in the acquisitions which are adding to the network. The user experience design in my opinion is excellent and has evolved over the years with continuous improvement. Personally speaking, adding restaurant/meals experiences will round out the platform nicely.
- The most interesting play in my mind: sensing engines as coined by Wardley. Building on the evolving machine learning toolset is the opportunity to make sense of lots of data points around scheduling, locations, pricing, occupancy. This consumption data becomes invaluable for creating new streams of product offerings.
- For me the gap opportunity is creating repeatable bundle. Loyalty programs play into this but it’s more about subscription models. My ‘content diary’ gameplay could accompany this. However I think to truly attain the FANG status they will need to come up with repeatable subscription products.
- My final thought on gameplay is a personal ‘want’ but I think is rooted in logic. I think of it as a ‘travelers diary*’ but at its heart it’s a content play and is directed at organic marketing with a couple of objectives; carving operating margin out of the high 30–35% marketing cost (as a percentage of revenues), increasing network effects, creating supply. I would leave it to the design geniuses at Airbnb to come up with something in the realm of ‘diary plus plus’.
- Operational Evolution — supply demand competitive pressures are naturally causing some of the operations hurdles to evolve and Airbnb clearly leads the lobbying in many of these such as cleaning standards, risk and insurance, regulatory environment, tax collections. The problem is any levers here are not really a competitive advantage for very long.
Value
To value Airbnb I have used a discounted cash flow model — intrinsic valuation — since this offers the ability to link levers to the valuation model (most IPO valuations use relative valuation which simply price relative to comparables and multiples). You can see extracted below how I have mapped a pipeline of components which are grouped around core drives of operating margin. After all, operating margin is the biggest unknown of the Airbnb valuation. Since currently they are in negative territory (not unusually for a young company). Starting at the top GM% is gross margin percentage and you can see it runs at a healthy 75% clip. Then you have 3 core operating expenses G&A (general and administrative), Sales and Marketing, and Operations and Support. Currently each has historically been in the range 13%, 30–35%, and 15–18%. Simple maths says that deducting these from the 75% gross margin leaves approximately 15%. However this is quickly depleted into negative territory by Product Development.
Without going too much into the weeds Product Development should be capitalized and I do that in my valuations but Operating Margin is still negative by about 13% currently. The big question is how much margin will Airbnb squeeze out over the next ten years and how can they do this. Another way of saying how will our gameplays impact margin (and growth). There is no spreadsheet in the world that will give you that answer. It’s also worth looking at direct competitors such as Booking.com which is about 35%, whereas hotels as a segment are about 15% (prior to Covid). Clearly G&A and Operations and Support can be reduced by a couple of points but except for gameplay 6 there is no direct impact. All the other gameplays impact revenue growth and Sales and Marketing. In particular as the business scales there should be opportunity for organic and brand marketing to maybe halve Sales and Marketing which would yield Operating Margins near 30%. That’s highly optimistic so a 25% goal may be a more measured target.
With the analysis I plug in the variables shown below. Financial press reports are claiming the IPO valuation around $35B prices between $45–50 (although over the weekend there is talk of $55–60). I think they will choose the higher level and I model this at $50. The table shows a number of scenarios and I also added for comparison Professor Damodaran’s baseline. Break even is roughly where the initial financial press thought the value would land i.e it’s implied valuation. Since then I have inched it up to the ‘safe’ scenario. Then I took both a pessimistic and optimistic view. Mainly driven by attainable operating margins and capital efficiency. My prediction is it will quickly zoom through to the $65 area on listing day and probably making it unattainable for mere mortals like myself.
For those interested in the valuation model some brief call outs. Cost of Capital is based on a bottom up Beta (0.91) for the hotel industry which is used to calculate Cost of Equity. It is somewhat questionable if Airbnb is indeed completely in the hotel industry. I tried to incorporate the country/geographic risk into equity risk premiums but unfortunately they disclose very little detail except some big regions which I ended up using.
Although they took on almost $2B of debt to get through the Covid crisis at very high interest rates (9.5%, 15.1%). I am assuming the IPO proceeds will drive this down and guesstimate a Ba2 credit rating — to estimate Cost of Debt. For investment purposes including capitalized product development I assume $2.50 of revenue can be generated with $1 of investment. Revenue growth reaches 30% but I have kept next year’s numbers more conservative.
My value of equity includes the dilution effect of a large number of restricted stock units and reduction based on the employee stock options outstanding.
At the time of writing this the IPO price is unknown although they provided a range and I used the upper number of $50-. I assumed all the proceeds would be added to the cash balance. I will update this post when we get confirmation of the final price.
- Dec 7th — we are hearing that the IPO price is now between $56 to $60. I will update my valuation closer to final pricing.
- Dec 10th — pricing is set for $68- (see updated S1); look for ‘wisdom of crowds’ i.e. going live at around midday. I have updated my valuation and put in place a few scenarios which I have called implied pricing. Read this as in watching where the price lands over the next few weeks and then what does it imply for the growth drivers.
For example if the stock behaves similar to yesterdays Doordash listing we will be in the territory of Implied 2. That means next years revenue growth would be 30% (possible) with a growth to 40% in year 5 (aggressive but possible); next years operating margins would turn positive 10% (unlikely and remarkable) and ultimately level out at 28% over 10 years (possible). I don’t see the sales to capital (capital investment efficiency) being a major lever. Its crazy valuation season right now but the wisdom of the crowds will speak…
- Dec 11th — with the price around $145 below are the implied growth numbers. Currently Operating Margins are negative 13, Airbnb grew 42% in 2018, 31% in 2019 and negative 25% in 2020. Personally I don’t see any basis in the intrinsic numbers to justify this, but we may have to wait for the next earnings review in January to come down to earth.
On Decision Making
As you can see in the above analysis there are lots of variables at play in determining the value. It’s quite difficult to understand what each piece of information contributes to the investment decision. Also each of the gameplays has different layers of complexity. Overall a pretty confused picture. To make for a clearer perspective I have leant on the Cynefin framework developed by Dave Snowden. I have included a diagram overlaying the strategy-valuation tools, but will use another post to get into this in more detail.
Suffice to say that we are in the Aporetic state which is aware of our confusion at this early IPO stage. We have some information we have utilized. In the clear area are the accountants financial statements and best practices that go around interpreting these. The complicated state relies on experts to develop effective valuation models where there are linear cause effect relationships. The complex domain is more about developing gameplays and linking these to valuation outcomes. Within this space is running what Dave calls ‘safe to fail’ experiments. My current thinking is around Monte Carlo simulations but there are many other techniques. The chaotic domain is more about scenario planning and developing some unconstrained practices (eg. crowdsourcing valuations) around chaotic events. For example a strategy around reacting to Covid — which Airbnb, although painful, has received kudos for handling and surviving.
Conclusion
Simon Wardley uses the analogy of strategy being akin to chess. I am not so sure. Certainly understanding the landscape as in a chess board is really useful for any business. Unfortunately the gameplays are much more unconstrained in the real world; as are the rules of engagement. Perhaps it’s closer to the world of warcraft, which is why Simon built his thesis on Sun Tzu and John Boyds’s OODA loop. For investment decisions you are looking to tie together ‘your story’ and the numbers. For me story telling is problematic mainly because of the HIPPO effect (highest paid person’s opinion). Whereby the leader in the room gets the prevailing story across the line. Mapping is very useful in depoliticizing the discussion since most people understand that maps are just representations and can be pointed at, debated, changed and even discarded.
Hopefully you see the benefit of using Wardley maps to extract the essence of strategy — even from a 489 page document. Then the map can be used to visualize and tell a story of the evolution of user needs and components. You can then draw sketches over the map and try out some gameplays. For me the real value (excuse the pun) is finding ways to inject these gameplays into the valuation — i.e. how much value will it contribute. I wish it were as simple as a chess match.
