Brad Templeton Home
Competition in the Robotaxi world
In a future world where there is ubiquitous and cheap robotaxi service, the business models of the automotive industry will change. How will robotaxi providers compete with one another, and who will win and lose?
The big question concerns just how many competing robotaxi companies a region can support. Some fear there might be a natural monopoly, leaving only one. Other signs suggest you might get 2 or 3 but not much more.
In San Francisco, the birthplace of all the human driven ride summoning companies, Uber dominates and Lyft takes up almost all the rest. Sidecar, which was one of the pioneering companies, shut down at the end of 2015. Both Uber and Lyft lose money, so it is uncertain whether SF has room for only 2, only 1, or some other number of companies. Of course there are also still taxis (including taxis with mobile apps for hailing) and the traditional black car companies. There is also transit and many carpooling tools which are used by big employers. There are even casual carpools on for the Bay Bridge.
I prepared a spreadsheet outlining some of the economics of operating a robotaxi. In the robotaxi world, I predict a price that drops, after the market matures, to near 30 cents/mile, with modest per-ride fees. This is cheaper than a transit ticket for a large fraction of trips. It's cheaper than owning a car yourself. Robotaxis will compete not just with one another, but with car ownership, car rental and public transit.
Today, if you ask "How do Uber and Lyft compete" the answer is pretty simple. It's almost entirely on price and availability/wait time, with some elements of brand. The companies would also point out they compete for drivers, and how they treat drivers affects the rider experience. The generally don't compete on the quality or brand of the cars, though they offer internal lines which define the class of car you'll get. Robotaxis may model the taxi business, or may embrace other elements, particularly from the car business.
The most obvious way companies can compete is on price. At the same time, there will not be just one level of service. All the other forms of competition below (such as luxury, wait time and more) can just be compensated for with better prices. Single companies will offer different levels of service at different prices.
Nonetheless there will be things that affect the price for companies offering the same service. Those who are able to build their vehicles more cheaply (at the same quality) can have lower costs. The rough cost of a mile in a large fleet will be equal to the total cost of a vehicle divided by the total lifetime miles of a vehicle, plus energy/fuel, overhead, maintenance and storage/parking -- plus profit and a (hopefully small) per-mile share of fixed costs.
Price predictability & ride packaging
Uber's "surge pricing" is famously controversial. By raising prices at times of peak demand, it works to make it more likely you will get a car if you are willing to pay. But people also hate it, because they can't predict the price. (For example, it is risky to commute a long distance with Uber as you might find the cost is unaffordably high due to surge when you want to return.)
Companies will complete on predictability, and they will also compete on price by selling subscriptions and packages, such as "unlimited rides for $x/month." Selling packages will also encourage customer loyalty. Riders who buy by the ride can switch company moment to moment. Subscribers will switch infrequently.
The core reason for surge pricing (getting drivers to drive at high demand times and go to high demand areas) does not apply to robotaxis. The entire fleet will always be available as scheduled (barring downtime for maintenance and recharging.) Surge pricing need only exist during major unplanned demand or at the expected top peaks of demand. (At those peaks, carpooling may be required to meet the demand.)
Another way to compete on price will be with the pricing plan. Today, car owners don't pay by the ride and they do not think of using their car as costing them 60 cents/mile, though it does. They think they buy gas once a week, maintenance and insurance once or twice a year, and a car every 5-10 years (or a car payment every month.) In spite of it all working out to the same number in the end, people like some payment methods more than others.
Rides will also be sold to other than the rider. For example, it could be that a large fraction of robocar rides are provided by employers as a free ride to work, or by stores, bars and restaurants offering free rides to patrons. If this is the basis for a large fraction of rides, the competition will be in bulk sales, where price will probably play a large role as well. Commutes are predictable trips, which can be served by much smaller fleets because you pre-allocate vehicles in advance.
Wait Time (and the many factors that affect it)
One of the most crucial competitive factors, and the one which points most strongly to a natural monopoly is how long people have to wait to get a ride. At the highest level, this depends on how many vacant cars the company has available near you, which depends a fair bit on fleet size. The company with the largest fleet is most likely to have a car close to you and win this contest.
It's not strictly fleet size though. The main factor in wait time is the density of available vehicles in your immediate neighbourhood. The density of available vehicles is actually governed by the fleet density minus that fleet's customer demand.
For example, a company with 500 cars and 300 customers in your area will actually have more available cars than a company with 1,000 cars and 850 customers, and give you a shorter wait time in spite of having a smaller fleet.
Generally, there is a basic wait time (a couple of minutes) below which there are diminishing returns. People may not value greatly getting the wait time down from 50 seconds to 25 seconds. It has been suggested that 2 minutes will be enough for most people. This is particularly true with systems which watch a customer's cell phone and predict the need for rides, allocating or moving vehicles based on the fact that demand is likely.
A smaller fleet can assure short wait times by limiting new customers until it has enough fleet to serve them, or knows it can hire capacity from partners. A company with a suitable customer distribution can offer a high level of service by having a relative oversupply of vehicles. When vehicles wear out by the mile, the cost of this is interest on the capital (around $1.25 per day for a mid-priced vehicle at today's low rates) and parking and travel to parking at off-peak times (variable, ranging from $1-$4/day) and the above-listed risk of obsolescence. Indeed, operators might advertise a deliberately high ratio of cars to customers, with only modestly higher cost, as a competitive advantage. Though there is a cost to worry about when margins get razor thin.
Think of it this way -- in today's world we have zero wait time, at the huge cost of everybody owing, parking, and paying interest on a car. Yet we tolerate the costs of that huge oversupply just fine. Small companies with big fleets can offer low wait times at just a little more cost.
Many rides have some predictability to them. Commuters will tend to travel at expected times. Trips to meetings and airports and other things found in calendars can be predicted and even scheduled. There is very little economy of scale for scheduled rides. You can be a working scheduled limo company with a fleet of just one. (You pass on requests you can't handle to partners.) Many other rides will be predicted by customers a short time in advance of a trip. Many people know they will be leaving somewhere 5-10 minutes in advance if not more. If they are willing to express that in advance, then a company with a small idle fleet can still offer zero wait time. The large fleet that offers short wait time with no advance notice still is more convenient, but not by that much.
This can be further compounded by very simple "AI" in a customer's phone that works to predict departures. For example, your phone might notice you ringing up purchases at a store, or walking to the elevator in a building or shaking hands and saying, "I need to go" to colleagues. Future wearables will have speech interfaces making it very easy to tell your device, "I will want a ride in 5 minutes." In this case, small fleet companies willing to act on these hints can serve such customers as quickly as large fleet companies. Some customers will be hard to predict, but others (like regular commuters, students etc.) will be very well predicted by their phones.
Fleets may also not be evenly distributed. A large fleet company might have the best average density over an entire town, but a niche company might always do better in your neighbourhood, and thus be your choice for trips originating from your home or office, but less so for return trips.
Smaller companies may also cooperate, passing business on to a partner if they know they can't provide quick service. In effect they will aggregate their fleet size while still monetizing their customer relationships.
Wait times will vary during the day, of course. We might find it easy to measure average wait time (which is what depends on idle fleet density) but customers might pay far more attention to things like maximum wait times. choosing a company that has a slightly longer average wait but never leaves them hanging for 15 minutes when they are in a hurry.
Moving empty vehicles to pick up customers has a cost, with estimates suggesting that will be about 10% of miles. Larger fleets will need to do this less, and fleets which do better predictive movement will also do it less. Fleets that have to do wasteful predictive moves (ie. the car did not get used) will suffer a cost. Fleets that anticipate demand and move vehicles well will do better.
Economies of scale in vehicle cost
The larger fleet will get better prices on everything because they buy more. In addition, their fixed costs can be distributed over more vehicles. If the fleet operator is also the vehicle developer or manufacturer, that can make a huge difference since all software and R&D costs must also be distributed over the fleet. Safe to say, a fully vertical developer/operator can't have a small fleet.
Having more vehicles in your fleet can also mean more data, and also more updates to map and traffic data. The more vehicles you have, the more quickly you will encounter map changes and construction zones.
Better cars at random
Wait times can be reduced by swapping in a more expensive vehicle than the customer ordered. You can usually serve a solo customer with a 4 seat vehicle, it is just wasteful. You can serve an economy customer with a luxury car too -- though the reverse is more complex. Generally you will design a fleet to be a bit more expensive than optimal because upgrading is always the easy path. Car rental companies routinely do this already.
Customers may like companies that often give them an upgrade.
Use of private and on-demand vehicles
Today, companies like Uber and Lyft do not own vehicles but rely on driver-owned vehicles. Owners of private robocars will have the option to have them hire themselves out to robotaxi operators when not in use by their owner. The cars will return when needed, or the fleet manager will compensate the owner with a ride in another car if they have sudden need for a vehicle. This might allow owners to make some money from their cars, and possibly afford a better car than they otherwise would. Tesla has announced plans to offer a service of this sort for their owners.
The economics of private vehicle use are less rational. Indeed, it is commonly felt that Uber, Lyft and others may be taking advantage of the fact that their drivers undervalue the wear and tear on their cars and calculate their profit with gasoline as the major cost, when in fact it is depreciation. Owned cars are likely to depreciate more by the year than the mile (unlike highly used taxis) and so owners will think of them as "an asset sitting around that might as well be used," rather than "an asset that depreciates as it is used."
In addition, owners may tend to own parking spaces, and not view parking there as having extra cost. In the future, though, it will be easy to rent out any parking space for current market rates. Robocars can park in your driveway or any other temporary spot, and leave the moment your car is about to return. Owners may also view their own interest costs as part of the cost of owning the car for themselves, and consider hire-out income as gravy.
Owned cars will tend to be larger, if current patterns hold. While most taxis, to be efficient, will be small and for 1-2 passengers, owned cars tend to be larger (4-5 passengers like today's cars) and thus have higher costs per mile. Normally this would push owned cars for use on group travel (more than 2 people) and pooling.
Owners may also price irrationally, saying, "as long as I own the car and don't need it, rather than have it sit there, I will hire it out any any price that costs more than the obvious per-mile costs." That is to say, they might ignore most of the depreciation and just set a price from fuel, maintenance, insurance etc. This could push down prices.
Generally private cars would be in highest demand during rush hour. The corporate fleets will have more than enough cars at off-peak times. In addition, private cars might be closer to a customer than fleet cars, and offer a shorter wait time, even off peak.
While it is Tesla that has promoted a plan to do this, there is a serious downside for electric car owners hiring out a car they plan to use later that day, which is the time for recharge. Electric car owners don't want to receive their car back for the afternoon commute with a lot of the range gone. Fast recharge can help, but tends to reduce battery life. This is not an issue for cars with liquid fuel.
Fleets may exist which hire-out their own spare vehicles on-demand, sometimes at higher prices. It may make more sense for a fleet to rely on more expensive vehicles during times of peak demand or when a premium customer wants quick service. The cost will be higher, but cheaper than expanding the fleet. Both private cars and fleet cars may advertise themselves as available at a price 50% higher than typical, and thus only get business when demand is very high. Competitors may cooperate (this is common in many industries) so that when one is peaking, the other fills in the gap, at a profit. A service that has promised luxury cars to customers may do so by hiring an outside car at premium prices those times it has run out of its own luxury cars. This will be worth it to offer service level assurances to premium customers.
Freshness of the vehicle
The lifetime of a vehicle is an interesting factor. If you can make vehicles last twice as many miles, your cost per mile goes down, but not quite in half. Customers will prefer vehicles that are new, both in years and in miles. You might not be able to charge as much for a 4 year old car with 200,000 miles as a brand new one. This will be mitigated by the ability to replace and update components (such as seat covers) and otherwise renew a vehicle. While vehicles will mostly wear out by the mile, a vehicle that lasts too many years will become obsolete from a technological standpoint and possibly from an aesthetic/fashion standpoint.
You can also make your vehicle more cheaply -- though that means it will wear out more quickly and may have more reliability issues.
Robotaxis won't serve every address, particularly at first. So one big differentiator will be how big a road network they do serve. When people want to go between destinations that the robots can't handle, they will need to either have a human driver (at their own cost or subsidized by the provider) or drive themselves, or transition between a robotaxi for part of the trip and human driving for the other part. As such, the company with the bigger road network will get more customers, and also have lower costs and prices.
They might also have to switch between robotaxis, which they won't like doing on short trips, but it might be the norm when crossing things like state lines if cars are only certified for one state.
Today, the nameplate on a car is the feature valued most by car customers. With ride services, however, the car brand has become less important. When poeple summon their high-end Uber, they don't care if it's Lexus, Mercedes or Lincoln. They care about the rides-service brand, but not in relation to the vehicle they will ride in.
Many product areas, particularly those associated with fashion, style and status, have customers who pay a great deal more than the basic price based on brand and aesthetics. People like to be associated with valued brands. They feel better about themselves buying higher-end brands. They want to be seen riding in certain brands (and luxury levels) and want that even more if they are going to have guests with them in the vehicle.
Companies will compete with brand and style to win demographics. In addition, single companies will maintain multiple brands and levels in order to capture more of the market. It seems likely that the value of brand will diminish (there is, after all, close to zero brand value in traditional hailed taxi services) but it's such a powerful force it can't be eliminated.
Of course, brand also represents all the other competitive features here. Brands will get reputations for the shortest wait times, best prices, best quality cars and more. I refer here to the image and style aspects of brand.
Brand (and luxury) choice can also vary ride to ride. Users may use an expensive brand when they know they will be seen or having guests, and a value-for-money brand on a general trip.
First Mover Advantage
Many players believe that the company that establishes itself first in an area gets a big advantage in establishing their brand is the obvious choice. They believe this so much that people are investing a lot of money in being first if they can. Nobody can be first in every location, but somebody will be first in the prime places.
Others wonder if there isn't a first mover disadvantage too. The #2 or #3 player is often the winner, getting the chance to learn from the first mover's mistakes for free.
There will be entirely aesthetic differences between brands. This can include style and design of vehicles, colours and things that inspire demographic ("tribal") affiliation. This can include the use of sub-brands, with cars affliated with sports teams, celebities and the like.
For many, a car is a form of self-expression. Many in the USA, that is. This is much less true in some other countries and in places like Manhattan. It is interesting to consider if other products (particularly clothing) will take over the self-expression, style and status roles played by cars in the USA.
Luxury & Features
Brand is sometimes connected with luxury level, but on the inside there will be real differences in quality, luxury and features, both between companies and between lines of the same company.
As noted, you can always serve an economy customer with a luxury vehicle, and a small-car customer with a larger vehicle if the service absorbs extra cost. A luxury customer, however, may prefer a lesser vehicle with a shorter wait-time. It will generally be the case that the wait time is shorter on a request for "closest available vehicle" than it is for "closest vehicle of a given luxury level." With most businesses, paying more (or ordering the luxury version) is always better.
Still, there will be customers of a certain wealth who demand a top quality vehicle almost all the time, and want a provider who can do that with good wait times. (Providers can decrease wait times by increasing the density of that particular fleet.)
There may also be differences in safety level. Cars have competed on safety, but it is expected that all robocars should have very high safety levels, making differentiation here challenging. We might see some cars with better crash ratings. It's also possible that we'll see cars that are deliberately less safe than they could be because they are small and light -- particularly solo vehicles which can fit 2 to a lane and perhaps get through traffic more quickly.
Special competition on safety might happen in areas like the safety of children. A car might surround a child seat with extra airbags, or might make one that swivels to put the child backwards to the direction of a pending crash with pre-fired airbags. Such technology is possible for adults, but very bulky. Extreme safety is possible if you want to spend the money (or the package is small.) Consider that race car drivers routinely walk away from crashes at more than 100mph with their 5 point belts, crash helmets, crash cages and other protection.
This is already a field of rapid innovation. This will not stop, and some providers will push radical new ideas in cars. Many customers will be interested in hiring cutting edge cars. In addition, being cutting edge can be a form of expression and style, even if the innovation is not itself yet that valuable!
All sorts of new factors come into play when attempting to put multiple parties in the same vehicle (carpooling or ride sharing.) Here, customers will expect vehicle type to vary greatly and so won't shop based on that. Instead it will be all about wait times, and delays involving picking up and dropping off other poolers -- and the cost discount you get for tolerating this. Here, economies of scale are strong. The big company is more likely to find other poolers going very close to your route, and thus minimize these delays. This may save more than money -- in some cities, pools will get access to special lanes and other privileges, causing people to pool to save time rather than money.
The cost of solo robocar service will be so low (perhaps 30-40 cents/mile) that only the poor would save enough money on short and medium trips to justify much iconvenience from pooling. To have pooling, you will need strong incentives (like carpool lanes or other privileges) and desire "very low detour" carpools. For example, a carpool that literally involves a 30 second stop for picking somebody up at something akin to a bus stop will be tolerable, but not a 10 minute drive going a mile off the best route.
Pooling may be required (or solo rides highly priced) at peaks of demand that exceed the capacity of the whole fleet. (Things like emptying a stadium, emergencies, weather and some rush-hour peaks.)
As noted above, cars may be designed for pooling, isolating the riders from one another for privacy or security.
Search Engines / Aggregators
At present, operators have resisted any effort to aggregate mobile-app taxi services. Many customers would prefer they could just go to a "ride search engine" and get a ride from whichever company has the best combination of wait time, price and other factors. The big player in a town is likely to not want to be in an aggregator, but smaller players may find it is the best way to work together to compete with the big player. Search sites are very popular for booking most travel needs like flights, hotels and rental cars.
Aggregators who get a commission can likely only take a very small piece of the pie, and as we know web search engines take none of it (usually making money in other ways.) However, the highly efficient market aggregators or search produces is more likely to enable competition. Customers shopping through an aggregator will mostly just want the closest car that meets their standards for quality and price.
Notes on costs of waiting and parking
Earlier, I predicted that the cost of idle time for a vehicle and parking would be quite low. This means that it's not that expensive to increase fleet size. Doubling the fleet size doubles the cost of interest on the capital and more than doubles the total storage/parking cost, but these are a small part of total costs. The real cost of doubling the fleet size (with the same customer demand) is that vehicles now last twice as many years, and start wearing out or going obsolete with time rather than miles.
If 1-2 person commuter vehicles cost $15,000 and interest rates are 3%, that's only $1.25/day in interest, which is a fairly minor cost. Think of it as just 8 cents/hour during waking hours and one can see this means there is very little cost to "leaving the asset idle."
I have also prepared a more detailed analysis of how cheap parking may get that forecasts parking may cost just a couple of dollars a day.
Do you have thoughts on other ways services might compete for your business? Leave them in the comments on this blog post.