Following up to yesterday's Tesla autonomy presentation, here is a specific article on the economics and principles of Tesla's proposed "Tesla Network" robotaxi service on Forbes' site.
I've been known for predicting surprisingly cheap numbers in my own article on Robotaxi economics but Musk forecasts a cost of only 18 cents/mile to operate a Tesla in their network, and big profits at a $1/mile retail cost.
You used $/hour several times when you meant $/mile.
NYC Taxi Fact Book says average taxi was 3.8 years old in 2017. They used to average 70k miles/year, but I think that's down a little. Anyway, the math suggests 450k mile life.
Tesla's depreciation advantage doesn't matter if the cars really generate $200k+ of profit over their life. That's a pure land grab scenario where costs are irrelevant because victory goes to whoever gets approval and rolls out the fastest.
Lower acquisition cost matters a lot if competition drives prices down to 30 cents/mile. But the myth of Model 3s appreciating in value goes away in that scenario.
I believe NY taxis are forced out of service after 5 years. They do about 60,000 miles a year so I will correct that number.
15% deadhead is not at all optimistic. Some studies have suggested just 5%. Those are optimistic. NY Taxis are 38% but they spend a ton of time cruising for rides. Robotaxis just go wait and only travel to get to their nearby rider.
Why would a Tesla robotaxi do any better than an Uber driver (who also could choose to spend no time cruising for rides)?
"Most Uber drivers report about a 1:1 ratio of paying miles to deadhead (no passenger) miles." https://www.forbes.com/sites/enroute/2018/06/05/waymos-push-into-robot-cabs-is-so-big-there-may-be-little-room-for-anyone-else/
I haven't seen the 15% estimates, but I bet they are describing an unrealistically idealized world. Perhaps they're assuming a monopoly where all the taxis in the world are owned by a single company?
In the real world, if you sit around in a location with no paying customers, you're giving up lots in opportunity costs compared to driving a few miles to a location with higher demand.
If your incremental cost per mile is $0.18 and your fare is $1/mile (which is much more realistic than an incremental cost of $0.22/mile and a fare of $0.53/mile), then it makes sense to deadhead quite a bit to get out of low demand locations.
(I wonder if it makes sense to have two different figures for incremental cost per mile. One for cost of miles driven empty, and one for the cost of miles driven with passengers. Self-insurance costs per mile will probably be double or more with passengers. Possibly even more than that, as you can drive much more carefully when there are no passengers to worry about harming or annoying. PIP insurance costs when you have no passengers will be zero. Cleaning costs will be significantly more with passengers. Energy and maintenance costs will be slightly to significantly more, depending if you can drive at a more efficient speed when deadheading.)
The studies with 5% are fairly idealized, but most of these studies are based on databases of real or relatively real (the household travel survey) trips. Yes, a good fleet manager does predictive placement of vehicles to where demand is predicted in these models. Some have done tests with perfect (in hindsight) predictors as well as more realistic ones. If Uber drivers are getting 1:1 they are doing terrible positioning, and I expect it also relates to Uber's practice of not letting them refuse rides that go the wrong way, so many take a ride from the airport to the burbs, and just drive back deadhead to not be down in the burbs.
That Gartner researcher's article contains a number of other serious flaws and has no cite for the 1:1 number so I would like to find a better cite on that.
Will Tesla let car owners refuse rides that go the wrong way? Not if they want to get 90,000 miles/year.
The very fact that there's a notion of "the wrong way" with Tesla car owners suggests that its figures will be more like those of Uber drivers and less like the artificial figures you allude to.
$1/mile seems reasonable. Your estimate, which I've criticized, was "20 to 25 cents/mile for the urban solo ride up to 50 cents/mile for the 4 passenger sedan." The biggest problem with setting the per mile price so low is that it would cause demand to go up too high during peak times. Another problem is that it wouldn't leave enough room for all the overhead of running a robotaxi service.
One question is whether or not there are any minimums or pickup fees. If I can pay just $2 to go to the store a mile away and back, that's going to be pretty great. More reasonable might be to add a $1 pick-up fee, in each direction, for a $4 cost to run to the store a mile away and back. Which is still not bad.
The $0.18/mile seems to be the *incremental* cost per mile, probably not including depreciation, and definitely not including any of the costs of running the robotaxi service. It probably doesn't even include insurance, as Tesla will likely be the one that has to provide the insurance (both as the manufacturer and as the operator of the robotaxi service). As such, that seems reasonable too.
I agree with the other poster that 15% dead miles seems wildly optimistic. I'd like to see what studies are giving those figures. Also, keep in mind that Tesla's service, where the vehicle presumably goes back to the owner periodically, is going to have a significantly higher number of dead miles. More on that later.
90,000 miles a year is an interesting prediction. Based on the "total gross profit" figure that is apparently the total number of miles, with only ~45,000 of those miles being used to transport passengers. It seems high, but if you can pick up a fair number of long distance overnight trips, it might be doable. (And that'll definitely increase the number of dead miles if the car needs to drive back to the owner every morning.) One thing that Tesla will need to do in order to make this work is that Tesla will have to provide a 24/7 cleaning service. And if they want to keep that $0.18/mile valid, the cost of that cleaning service will have to come out of Tesla's 25-30% fee. (That's probably the way to go, though, as Tesla can recoup at least some of its cleaning costs by charging the passenger that makes the mess.)
Bringing your car back to you every morning for your daily commute will definitely have a cost, but it really isn't all that unreasonable. Say your car is 50 miles away for 180 days out of the year. At $0.18/mile that's only $1620/year. And that's probably an overestimate of how far away your car will be on average. And if your commute is short enough, maybe you don't need your own car back for your daily commute. If your commute is only 5 miles, and your car is more than 23 miles away, you might be better off taking someone else's Tesla to work for $5.
An owner's use probably *will* be highly predictable. For short trips you can always hire out someone else's car. If you need to go to the store 1 mile away and it costs you $2 to use the Tesla Network, then if your own car is more than 9 miles away you're better off not summoning your own car. If you tend to have to take a lot of long, unpredictable trips, maybe joining the Tesla Network isn't for you.
I believe Tesla's 18 cents did include depreciation, but $40k / 1 million miles is only 4 cents/mile. I agree it doesn't include insurance since they apparently hadn't even thought about liability.
Pickup fee, cleaning service, remote operators, child seats, etc. are all details that Tesla hasn't started to address. The Oppenheimer analyst was being generous when he called their plan "half-baked". They've got time, though. The other stuff doesn't matter until their self-driving is reliable.
Well, as Musk said, Tesla will be liable for a robotaxi crash operating in their network, so the owner will not have insurance costs for that part of their car's life. Tesla will probably self insure, so the cost depends on how good their system is. If their system isn't good, it will be forced to stop anyway.
It is the correct position.
The insurance companies, never pay, the money comes from the insured. Elon Musk has no money, the money comes from his clients. The insurance systems are no more than human solidarity, previously organized, restricted (to those who took out the insurance), and limited (depending on the type of insurance).
In the future RC world, it is logical that the users (beneficiaries) of the SDV, as well as the victims, receive compensation from those who receive the payment (producers of SDV) for the use of SDV. (Self Driving Vehicles).
A human life, a leg, has no price, but somehow we must compensate for the inconveniences that loss can cause.
In the future there will be no owners of SDV, only users. Of course you can pay for the permanent and unrestricted use of one or more SDV, but not owning it. So we can have fair compensation, (not depending on whether the culprit is Bill Gates or similar), nor lost time and expenses in lawsuits and lawyers.
Obviously Tesla will be liable. Anyone who thinks about this 20 seconds knows that. My point is when asked, Musk said "well, um, I guess Tesla will". Up to that point he hadn't thought about it 20 seconds. And it's a crucial issue, possibly a showstopper.
As I said, though, they've got a long time to figure this out. IMHO Tesla is 3-5 years away from robotaxi-level reliability.
Since you have been posting at Forbes, I've noticed your font size is too small. Chrome says your font size is 12px, whereas Forbes is 18px. The Forbes size seems about right to me. Although Forbes has way too many annoying pop-up, slide-over, and ordinary banner ads.
Comments
Doggydogworld
Tue, 2019-04-23 13:28
Permalink
You used $/hour several times
You used $/hour several times when you meant $/mile.
NYC Taxi Fact Book says average taxi was 3.8 years old in 2017. They used to average 70k miles/year, but I think that's down a little. Anyway, the math suggests 450k mile life.
Tesla's depreciation advantage doesn't matter if the cars really generate $200k+ of profit over their life. That's a pure land grab scenario where costs are irrelevant because victory goes to whoever gets approval and rolls out the fastest.
Lower acquisition cost matters a lot if competition drives prices down to 30 cents/mile. But the myth of Model 3s appreciating in value goes away in that scenario.
15% deadhead seems wildly optimistic, IMHO.
brad
Tue, 2019-04-23 17:00
Permalink
NY Taxi life
I believe NY taxis are forced out of service after 5 years. They do about 60,000 miles a year so I will correct that number.
15% deadhead is not at all optimistic. Some studies have suggested just 5%. Those are optimistic. NY Taxis are 38% but they spend a ton of time cruising for rides. Robotaxis just go wait and only travel to get to their nearby rider.
FKA
Tue, 2019-04-23 18:42
Permalink
dead miles
Why would a Tesla robotaxi do any better than an Uber driver (who also could choose to spend no time cruising for rides)?
"Most Uber drivers report about a 1:1 ratio of paying miles to deadhead (no passenger) miles." https://www.forbes.com/sites/enroute/2018/06/05/waymos-push-into-robot-cabs-is-so-big-there-may-be-little-room-for-anyone-else/
I haven't seen the 15% estimates, but I bet they are describing an unrealistically idealized world. Perhaps they're assuming a monopoly where all the taxis in the world are owned by a single company?
In the real world, if you sit around in a location with no paying customers, you're giving up lots in opportunity costs compared to driving a few miles to a location with higher demand.
If your incremental cost per mile is $0.18 and your fare is $1/mile (which is much more realistic than an incremental cost of $0.22/mile and a fare of $0.53/mile), then it makes sense to deadhead quite a bit to get out of low demand locations.
(I wonder if it makes sense to have two different figures for incremental cost per mile. One for cost of miles driven empty, and one for the cost of miles driven with passengers. Self-insurance costs per mile will probably be double or more with passengers. Possibly even more than that, as you can drive much more carefully when there are no passengers to worry about harming or annoying. PIP insurance costs when you have no passengers will be zero. Cleaning costs will be significantly more with passengers. Energy and maintenance costs will be slightly to significantly more, depending if you can drive at a more efficient speed when deadheading.)
brad
Wed, 2019-04-24 10:52
Permalink
Dead miles
The studies with 5% are fairly idealized, but most of these studies are based on databases of real or relatively real (the household travel survey) trips. Yes, a good fleet manager does predictive placement of vehicles to where demand is predicted in these models. Some have done tests with perfect (in hindsight) predictors as well as more realistic ones. If Uber drivers are getting 1:1 they are doing terrible positioning, and I expect it also relates to Uber's practice of not letting them refuse rides that go the wrong way, so many take a ride from the airport to the burbs, and just drive back deadhead to not be down in the burbs.
That Gartner researcher's article contains a number of other serious flaws and has no cite for the 1:1 number so I would like to find a better cite on that.
FKA
Wed, 2019-04-24 17:49
Permalink
I'd like to find a cite for 15%, so that I can criticize it
Will Tesla let car owners refuse rides that go the wrong way? Not if they want to get 90,000 miles/year.
The very fact that there's a notion of "the wrong way" with Tesla car owners suggests that its figures will be more like those of Uber drivers and less like the artificial figures you allude to.
Anonymous
Tue, 2019-04-23 17:44
Permalink
Numbers
$1/mile seems reasonable. Your estimate, which I've criticized, was "20 to 25 cents/mile for the urban solo ride up to 50 cents/mile for the 4 passenger sedan." The biggest problem with setting the per mile price so low is that it would cause demand to go up too high during peak times. Another problem is that it wouldn't leave enough room for all the overhead of running a robotaxi service.
One question is whether or not there are any minimums or pickup fees. If I can pay just $2 to go to the store a mile away and back, that's going to be pretty great. More reasonable might be to add a $1 pick-up fee, in each direction, for a $4 cost to run to the store a mile away and back. Which is still not bad.
The $0.18/mile seems to be the *incremental* cost per mile, probably not including depreciation, and definitely not including any of the costs of running the robotaxi service. It probably doesn't even include insurance, as Tesla will likely be the one that has to provide the insurance (both as the manufacturer and as the operator of the robotaxi service). As such, that seems reasonable too.
I agree with the other poster that 15% dead miles seems wildly optimistic. I'd like to see what studies are giving those figures. Also, keep in mind that Tesla's service, where the vehicle presumably goes back to the owner periodically, is going to have a significantly higher number of dead miles. More on that later.
90,000 miles a year is an interesting prediction. Based on the "total gross profit" figure that is apparently the total number of miles, with only ~45,000 of those miles being used to transport passengers. It seems high, but if you can pick up a fair number of long distance overnight trips, it might be doable. (And that'll definitely increase the number of dead miles if the car needs to drive back to the owner every morning.) One thing that Tesla will need to do in order to make this work is that Tesla will have to provide a 24/7 cleaning service. And if they want to keep that $0.18/mile valid, the cost of that cleaning service will have to come out of Tesla's 25-30% fee. (That's probably the way to go, though, as Tesla can recoup at least some of its cleaning costs by charging the passenger that makes the mess.)
Bringing your car back to you every morning for your daily commute will definitely have a cost, but it really isn't all that unreasonable. Say your car is 50 miles away for 180 days out of the year. At $0.18/mile that's only $1620/year. And that's probably an overestimate of how far away your car will be on average. And if your commute is short enough, maybe you don't need your own car back for your daily commute. If your commute is only 5 miles, and your car is more than 23 miles away, you might be better off taking someone else's Tesla to work for $5.
An owner's use probably *will* be highly predictable. For short trips you can always hire out someone else's car. If you need to go to the store 1 mile away and it costs you $2 to use the Tesla Network, then if your own car is more than 9 miles away you're better off not summoning your own car. If you tend to have to take a lot of long, unpredictable trips, maybe joining the Tesla Network isn't for you.
Doggydogworld
Wed, 2019-04-24 10:28
Permalink
Numbers
I believe Tesla's 18 cents did include depreciation, but $40k / 1 million miles is only 4 cents/mile. I agree it doesn't include insurance since they apparently hadn't even thought about liability.
Pickup fee, cleaning service, remote operators, child seats, etc. are all details that Tesla hasn't started to address. The Oppenheimer analyst was being generous when he called their plan "half-baked". They've got time, though. The other stuff doesn't matter until their self-driving is reliable.
brad
Wed, 2019-04-24 11:02
Permalink
Insurance
Well, as Musk said, Tesla will be liable for a robotaxi crash operating in their network, so the owner will not have insurance costs for that part of their car's life. Tesla will probably self insure, so the cost depends on how good their system is. If their system isn't good, it will be forced to stop anyway.
Alejandro
Wed, 2019-04-24 18:34
Permalink
No more insurance (and their cost)
It is the correct position.
The insurance companies, never pay, the money comes from the insured. Elon Musk has no money, the money comes from his clients. The insurance systems are no more than human solidarity, previously organized, restricted (to those who took out the insurance), and limited (depending on the type of insurance).
In the future RC world, it is logical that the users (beneficiaries) of the SDV, as well as the victims, receive compensation from those who receive the payment (producers of SDV) for the use of SDV. (Self Driving Vehicles).
A human life, a leg, has no price, but somehow we must compensate for the inconveniences that loss can cause.
In the future there will be no owners of SDV, only users. Of course you can pay for the permanent and unrestricted use of one or more SDV, but not owning it. So we can have fair compensation, (not depending on whether the culprit is Bill Gates or similar), nor lost time and expenses in lawsuits and lawyers.
Doggydogworld
Thu, 2019-04-25 07:33
Permalink
Liability
Obviously Tesla will be liable. Anyone who thinks about this 20 seconds knows that. My point is when asked, Musk said "well, um, I guess Tesla will". Up to that point he hadn't thought about it 20 seconds. And it's a crucial issue, possibly a showstopper.
As I said, though, they've got a long time to figure this out. IMHO Tesla is 3-5 years away from robotaxi-level reliability.
Ross
Wed, 2019-04-24 13:00
Permalink
If you can read this...
Since you have been posting at Forbes, I've noticed your font size is too small. Chrome says your font size is 12px, whereas Forbes is 18px. The Forbes size seems about right to me. Although Forbes has way too many annoying pop-up, slide-over, and ordinary banner ads.
brad
Thu, 2019-04-25 21:24
Permalink
Point size where
Where is the 12 point you refer to? On this site? I am using a particular drupal theme, open to suggestions of others.
MPH
Wed, 2019-05-01 12:18
Permalink
MIT Nunes & Hernandez AV Study 2019 labeled DRIVERLESS TAXI COST
MIT Nunes & Hernandez AV Study 2019 labeled DRIVERLESS TAXI COST hood reference for this discussion
MPH
Wed, 2019-05-01 12:26
Permalink
Title for research study
/Autonomous Vehicles and Public Health: High Cost or High Opportunity Cost?/
Add new comment