Surges and fleets, and whether your robotaxi ride will be cheap in the off peak


Uber changes prices with demand To serve a city, a robotaxi fleet will need to be large enough to handle the peaks, which is to say rush hour. Today, we handle that with an immense fleet which sits idle most of the time. A robotaxi fleet can be much smaller. Some people suggest ridiculous numbers like 10x smaller but the best estimates I have seen are more in the range of 4x smaller -- that even at the peak, only 25% or so of cars are in motion.

Of course, there will be privately owned cars as well as robotaxis, so the total cars in existence will be larger than this. Outside of cities, privately owned cars will continue to be the norm, in fact.

It is also suspected that some privately owned cars will hire themselves out as taxis when their owners don't need them. Tesla has said they will support this for owners of future full-self-drive Tesla models. I used to be a big fan of this idea, but over time, I have come to believe it will be a minor factor. It's more hassle than people think. I've noticed that as I use AirBNB, the units available are no longer somebody's personal apartment rented part time and mostly units dedicated only to rental. Similar factors will apply here.

One place private hire-out could make sense is at the very top of the peak. The smaller the fleet, the higher utilization it gets. It could make sense to handle exceptional super-peaks by calling upon private cars and offering them a high rate during this high demand time. Instead of greatly increasing the price, as Uber does in a "surge," handle those surges with private cars.

Unlike surges, where there is a limited supply of human drivers that is too small, the entire robocar fleet will usually be available at all times at what might be each car's "standard" price. If you own a car, and it is not offline for maintenance, if you get an offer at your standard price, there is no reason not to take it. What else is the car going to do? Indeed, unless fleets are in a nasty competitive war, many fleet operators will be happy to let another fleet operator use their vehicle, for the right price, when they can't easily serve the customer themselves. Small operators will very much want to do that.

Fleet size -- or more accurately the density of idle vehicles in your fleet -- is what determines wait time, which is a very important competitive factor. The wait time a fleet offers is equal to the customer density in an area divided by the density of idle vehicles in that area. Because vehicles move around and density varies, it can make a lot of sense to sometimes pay a competitor to serve the customer. Many industries do this.

The first instinct is to say, "it's expensive to grow your fleet" to compete, and it is indeed capital intensive. But it's much less expensive than one might think. When cars wear out by the mile, rather than the year, having a larger fleet simply means the cars wear out more slowly. Make your fleet 20% bigger and it lasts 20% longer in time (within reason.)

So the cost of the expanded fleet is just the interest on the capital for the extra life, and the parking cost to store it during off-peak times. Today's interest rates are very low, so the interest cost is not so great. And I predict that parking will be very cheap so that's not large too. One year's interest is about 3% these days. Parking will not be much more than $2-$3/day. So while having a 20% larger fleet requires 20% more capital, the real daily cost per vehicle is much lower and can be readily justified by the competitive advantages of a larger fleet.

In fact, there can be positive feedback. If having a larger fleet means shorter wait times, you will win more business or possibly even be able to charge higher prices, actually making money rather than losing it.

What about off-peak

Recently, a reader proposed that prices might drop a great deal for rides in the off-peak times of the day. After all, with vast numbers of vehicles sitting idle, paying for parking, shouldn't operators cut prices to the bone to win the smaller volume of business? If you have an asset sitting there, and it can instead do work, should it not offer itself out at whatever price still makes money and wins the customer?

This happens in some industries and not in others. Many things are cheaper in the off season and off peak. Some aren't. How will this play out for robotaxis?

On the one hand, a robotaxi ride can be sold with no human labour cost -- the selling part is almost free. It is unclear how much cheap off-peak rides will cannibalize on-peak rides. Some people will be ready to travel at a different time to save money, but that seems more likely for long trips than short trips. People certainly try to modify their commutes to avoid rush hour, but that's to avoid congestion, not to save cost.

A car will have costs to move -- depreciation by the mile, insurance (cost of risk due to moving,) maintenance and repairs and possibly road use fees. Today those are around $20 per hour, but for small electric cars that could drop to about $6/hour. It won't sell for less than these amounts.

A car also has costs to sit still, namely interest and parking. Those are much cheaper, a few dollars per day, or pennies per hour.

If a car cuts prices to grab business in the off peak, this is not found money, it's lost money. If the car is going to last 300,000 miles, consider two choices:

  1. It sells them all at 50 cents/mile, but takes 6 years to do so -- $150K over 6 years
  2. It sells 50,000 of them at 30 cents and 250K at 50 cents -- $140K over 5 years

In model 2, it has offered itself cheap off peak and gotten an extra 10,000 miles of business per year. It needed one year's less parking ($600) and one year's less interest on tied up capital (Also about $600 on a $20,000 car.) So it loses a lot of lifetime revenue for a modest gain in efficiency due to lower utilization.

It is not quite this simple. Some elements of a car will depreciate with years rather than miles. One strong one will simply be technological and aesthetic obsolescence. We know that we toss our perfectly good 2 year old phones in the closet because they don't make the cut any more. If your cars start lasting 10 years, they will lose value even if they aren't wearing out very much.

In addition, there may be companies who are not competing very well, and need to cut prices in order to get enough business to get their cars to even a basic level of utilization. To gain customers, they will need to cut prices all day long, not just off peak. The above math says that a good sized fleet would be happy to let other people wear out their cars on the lower margin off-peak business. On the other hand, top-flight companies need to promise you service whenever you want it.

Some services will charge flat rates rather than billing per ride. For those, it may be hard for customers to compare.

But overall my conclusion is that rides should be only very slightly cheaper in the off hours. In the sense of perhaps no more than 5 cents/mile -- "I will give you a small discount just to save me the cost of being parked."

Will there be surges?

If a company is very short of vehicles, it might raise prices due to demand. But if competitors are not out of vehicles they don't have a reason to. The answer depends on how easily customers can search among different providers (once there are different providers -- that's some time in the future.)

In theory a city's fleet should be large enough to hit the peak demand of its users. And if it isn't, it may be that private cars do hire themselves out when the rate gets high. Private cars, unlike taxis, wear out by the year. For them, marginal profit can be attractive compared to sitting there, and high profit is definitely nice.

There will be unusual peaks. When a stadium needs to be emptied out. Busiest travel days of the year. And, like we are seeing in the Carolinas today, disaster evacuations.

Such peaks, however, especially disasters, have another solution, namely shared vehicles. While the fleet may not be large enough to give everybody their own vehicle, it probably has enough seats to give everybody a seat. Last year I published an article on disaster evacuation. Politically, the idea of surge pricing during high disaster demand is a no-go, even when it makes sense.


I note that right now in places like North Carolina, most of the cars have jammed the streets at once fleeing from an imminent natural disaster. How does one handle this with robocar fleets if they're much smaller than the total population? We can come up with answers I'm sure, but they do need to be thought of at some point.

I link to it at the bottom of the article above. The answers include things like making sure all seats get used, having cars go back to get more people once they have evacuated the first wave, and bringing in cars from other cities. In a hurricane, places 100 miles away are fine -- their spare cars can be there in a couple of hours to do more waves.

Add new comment