The Role of the New SaaS Model in Our Robot Taxi Future. (Part 1 of 3)
The sophisticated technologies, the sheer number of AV suppliers, and the concentration of ride-demand across far fewer ride-hailing platforms are the three primary factors driving the relevance of the Supply-as-a-Service (SaaS) model to robot taxi commercialization over the near-term.
This is Part 1 of a three-part series covering the critical role and execution of what I call the Supply-as-a-Service (SaaS) model in the commercialization of robot taxis. This Article 1 starts by explaining why the SaaS model will be so critical in the near-term commercialization of robot taxis. Part 2 provides a framework and detailed strategic and operations considerations to guide the execution of the model. In Part 3, I apply the framework and considerations of Article 2 to predict how the model will be executed along critical and complex dimensions of the model.
In this article, we have already established that the robot taxi is the best commercialization strategy for an autonomous vehicle. The next question is then Who will own the fleet when everyone wants to avoid this “asset- and operations- heavy” business model? It turns out, what I call the “Supply-as-a-Service” model plays a significant role in answering this question.
Essentially, the two most powerful types of players in the robot taxi industry are the “autonomous vehicle (AV) suppliers” and “demand-generation platforms,” which I define as:
- “AV suppliers” to refer collectively to both “AV tech suppliers”, who are creating technology systems needed for making a vehicle capable of driving autonomously, and vehicle manufacturers with whom these AV tech suppliers are working to put an AV on the road.
- “Demand-generation platforms” are the ride-hailing platforms such as Uber, Lyft, Gett, Ola, Grab, etc. who have critical assets in 1) a large customer base of riders, and 2) an array of sophisticated technologies, products, and operations that are focused on maximizing ride-demand and revenues therefrom.
In a few cases, such as Uber and some other OEMs, the AV supplier also plans to be the demand-generation platform; however most AV suppliers are not demand-generation platforms nor vice versa. Many are separate entities today and — I argue — will be for years if not indefinitely.
Let’s be clear and state the fairly obvious but important point. Manufacturing a car at scale is a unique capability that few companies have. Ultimately OEMs will be needed to design and assemble the vehicles, but there are many other players who can provide the rest of the tech and assets needed to bring a robot taxi service to market.
Below, I examine why both the AV supplier and demand-generation platform either can’t or doesn’t want to vertically integrate into each other’s space, and hence why they will remain separate entities and employ the Supply-as-a-Service model over — at least — the next 3–5 years.
Why AV suppliers won’t create their own ride-hailing platform [in the short-term].
Replicating all the platform product and operations would be inefficient use of capital and time. While creating the traditional ride-hailing product and business competencies needed for operating a ride-hailing platform is certainly within the technical capabilities of most autonomous vehicle teams, who are solving one of the most challenging computer science problems of our generation, AV teams are smart to realize that creating a ride-hailing platform would require significant capital and — maybe more importantly — time. The question is not “Can you build the tech and operations?”, it is “Should you build them?”. Surge-pricing, dispatcher, fraud, demand-estimation, supply-optimization, etc. systems are complicated technologies to develop, but — with enough time and money — they can find the talent and build them. Heck, Ridecell and BestMile are two companies that offer off-the-shelf white-label software stack that a new entrant can use to rapidly build — almost overnight — these and many more downstream layers in the tech stack. But this tech isn’t the real issue. The real impediment to starting a new ride-hailing service is the cost of operating an inefficient network, which I describe in more detail below.
Network would be too small. Keeping ETAs low is critical to a good rider experience. Low ETAs can only be achieved reliably at certain scale, which depends on the physical size and demand density in the operational domain served by the robot taxi. Note that there is both a time and geographic dimension to demand. It is obvious that ETAs in a given market will be longer if the fleet is too small, relative to demand. Less obvious however is that ETAs would be long also if demand volume and density are low. This is because the low level of demand would cause an AV supplier to put few vehicles on the road, and this would then result in long “deadheads”. These long deadheads lower ride demand in another way as well. Because vehicle miles and time cost money, the vehicle supplier must recoup the cost of these deadheads through higher fares, which will lower demand.
Because today’s global ride-hailing demand is concentrated on just one or two regional platforms, the demand density each platform is large enough to provide for efficient fleet utilization. The more efficiently a fleet is utilized, the lower become its per ride unit costs and the lower its ride-fare prices can be while still being profitable — an obvious competitive advantage.
This does not mean an AV supplier could not be successful in launching a new platform; rather it simply means that the new entrant would have to incur the same costs that the incumbent platforms have already paid to reach their size and scale. This is a big disincentive, and some new entrants may lack the ability or willingness to invest such massive amounts of capital** to absorb these inefficiencies. For example, in getting a market to scale, all the ride-hailing platforms burned through significant amounts of capital incentivizing drivers to be online even when supply exceeded demand. They smartly did this because they know demand is primarily a function of ETA and price of the ride, and the more cars on the road, the lower are ETAs. A new ride-hailing platform would have to endure the similar — although not as extreme — costs of artificially inflating supply side of the network provide a good rider experience, which then leads to a growth in the demand side of the network. Thus getting the network effect’s flywheel spinning.
No trip data. Without access to reasonably accurate trip demand data, an AV supplier will make suboptimal fleet and infrastructure (e.g. parking, charging, repair facilities, etc.) sizing decisions, giving a better-informed competitor the opportunity to deploy the right-sized fleet and infrastructure and then capture market share. Without trip data, these massive and sticky capital expenditure decisions would be made in an uninformed manner, creating significant risk of suboptimal or even unprofitable operations, wherein the fleet is underutilized and/or demand is unfulfilled. With the exception of Google’s Waymo, no AV supplier has access to trip demand data.
‘No trip data’, like ‘Network would be too small’, is more of an obstacle than a barrier to market entry. It doesn’t prevent the new entrant from entering, it just potentially slows their market share growth and lowers profits. A new entrant can of course enter a market without trip data and then slowly learn where and when there is ride demand, but there is a problem with this approach. Fleet and infrastructure acquisition decisions are long lead-time decisions that must be made relatively far in advance of deployment because it takes so long to manufacture a car and install the infrastructure. As a result, a new entrant employing a strategy of conservatively entering the market with too few cars and then incrementally adding more cars to the fleet leaves the entrant vulnerable to a better-informed competitor who deploys faster and grabs market share sooner.
Speed to market-saturation matters. Since the existing ride-hailing platforms provide the fastest and most capital efficient means of executing on a given robot taxi market (e.g. Chicago metro area), the first AV supplier to market can saturate the market and create significant barrier to entry that deters new entrants. When an AV supplier goes to market, it will consume a variety of finite resources that will make it difficult for a prospective competitor to procure and compete. Such finite resources include real estate (parking garages, repair facilities), curb space, service vendor capacity, etc.
In addition, if an AV supplier is already adequately serving the needs of a given share of the market, it leaves little room for a new entrant to win that share of the market. As a result, the AV supplier who gets to market and scale the fastest by partnering with a platform stands to create significant barriers to entry for competing AV suppliers.
Why ride-hailing platforms prefer the supply-as-a-service model.
Asset-lite business model. Ride-hailing platforms enjoy an extremely flexible asset-lite business model and will avoid — at all costs — an asset-heavy business model of owning and/or maintaining their own fleets, which is not their core competency (software and building consumer-facing services are). Owning vehicle fleets would require a platform to deplete cash reserves and divert significant cash from other growth initiatives — unless they could find an asset-backed lender to finance the vehicle acquisition.
Also, by partnering with multiple AV suppliers, the platform can potentially move faster than a competing platform to saturate the market with supply and grab market share.
Lower overall costs of autonomous vehicle supply. There are much fewer ride-hailing platforms than AV suppliers, providing the platforms with leverage to negotiate favorable terms with competing AV suppliers.* Many platforms, such as Uber and Lyft, have already publicly announced plans to work with multiple AV suppliers. I suspect they do this not only because each supplier has limited near-term supply but also because this enables them to negotiate attractive pricing and terms and to minimize the risk of being vulnerable to any one supplier.
Lack of technical know-how. To put an AV on the road and give fully autonomous rides requires operational and technical competencies that most platforms lack. Mapping operations, vehicle data downloads and storage, updating on-vehicle software, flashing/updating hardware’s firmware, vehicle refueling, vehicle maintenance, and remote pilot operations, are just a few of the competencies that the AV supplier uniquely has and that the platform would be unwise to build.
Final Thoughts
Combined, all these forces mean that ride-hailing platforms will be able to negotiate with most AV suppliers to provide their vehicles as a service.
In Part 2, I dig deep into the strategic and operational considerations that will drive how the supply-as-a-service model is likely to be executed between AV suppliers and ride-hailing platforms. Part 3 outlines how I see these critical considerations being executed between AV suppliers and platforms.
Follow me on Twitter (@jasonpdoran) for more thoughts on startups and our urban mobility future. This post was also published on Medium at @jasonpdoran.
Part 2 — Executing A Robot Taxi Supply-as-a-Service Model
Part 3 — How the Robot Taxi Supply-as-a-Service Model Will Be Executed
*To build a network the size of Uber or Lyft in the United States would cost in the hundreds of millions, primarily in rider incentives, which competing platforms will also use ferociously to compete for riders.
**At last count, the State of California has autonomous vehicle testing permits to over 24 teams, and there are even more. I estimate closer to 35 teams worldwide.
