Wednesday, October 13, 2010

Attacking the Automated Driving Market

Yesterday's post on Google's automated driving announcement attracted some very interesting comments that set me thinking more carefully about how such a technology market might develop.  So our exercise this morning is to imagine that we are an executive in a company whose mission is to develop automated driving technology and to try to become the dominant player in the automated driving market.  My qualifications to engage in this exercise include a decade of sitting in executive staff meetings of technology startups - and thus a pretty good general idea of how Silicon Valley thinks - but do not include much knowledge at all of the auto industry in particular.  So take it for what it's worth.

Ok, so here we go - we are CEO of aDriveTM, let's let in the clutch and engage first gear.

The first thing we would need is to estimate the eventual TAM - Total Addressable Market.  That is, what is the plausible eventual annual revenue of all automated driving companies together?  (Once we have this, we will claim that we will be the market leader with 60% market share - or some similar swag - and that will be our estimate  of how big aDriveTM eventually could be).  In the TAM calculation we are thinking about the situation where the market is mature - let's say in twenty-five years, when this technology is mature and boring and everyone has it (eg in the same way that the PC market is pretty mature now).  Obviously the market will be much smaller now (in the sense that if Google offered the system for sale tomorrow, they'd probably only find a fairly small number of customers).

So, in thinking about the TAM, it seems to me that it breaks down into two segments that are rather different from each other.  The first is a consumer market: selling to people who drive a car now, could perfectly well continue to drive themselves, but for one reason or another would be willing to pay extra to have an automated driving system installed (much as, in recent years, you might have paid to have a navigation system installed in your car).  This obviously requires partnering with existing car companies to engineer our technology into their cars and sell and market it to their new car customers, and it is basically a vanity/convenience sale to consumers.

The second segment is a commercial market: all  situations where someone is currently being paid to drive a vehicle are future opportunities for the aDriveTM sales organization to try to convince them that it would be cheaper and more reliable to fire the humans and use an aDriveTM system to drive their trucks/vans/buses/limos instead (the polite term here is "reducing the total cost of ownership").

For the consumer market, we can guesstimate the TAM as follows - there are 56 million auto sales worldwide in 2010.  Let's figure it will roughly double by the time the automated driving market is mature.  And let's figure that by then competition will have driven the revenue per car for an automated driving system down to a few hundred bucks.  (It will start out as a $5000 option for a Mercedes or BMW, but end as a $200 standard thing in a Tata).  So in round numbers, let's say $200 x 100 million = $20 billion.  Clearly that could easily be off by a factor of two or three, but that's the ballpark.

For the commercial market, we can get some clue by looking at what companies spend on humans now. From the same BLS data we looked at yesterday, we can also find average annual earnings of drivers:

Taking those values, and the number of drivers employed from yesterday, we can compute the total US wage bill for drivers:

$120 billion a year is a lot of money.  And that's not counting benefits, management costs, etc.  Let's multiply by three as a very rough guesstimate for the worldwide driving wage bill (figuring Europe and Asia are roughly equal sized markets to the US). So we end up that roughly a few hundred billion a year are being spent on commercial driving wages worldwide.

Now, maybe the automated systems won't end up capturing as much revenue as the human drivers.  Still, any way you look at it, that's a very large potential market by Silicon Valley standards.

In particular, it's very much larger than Google's current revenues, which their latest 10-Q suggests are running a little shy of a $30b annual rate.  So, just from a financial perspective, it would not be crazy for Google itself to go after this market.  (Obviously, I have no actual information on their plans, but the New York Times suggests an engineering group of 15.  At the $0.25m/year rule of thumb for fully loaded engineers in California, that suggests a burn rate of about $4m/year, and so if the group's been going since some time not too long after the DARPA Grand Challenge, they've put somewhere in the neighborhood of $15m into this project - more than pocket change, even for Google).

Further, I note that - if the technical problems can be fully solved - automated driving systems have some very compelling advantages over human drivers (from the perspective of their owners/employers):
  • Can drive 24x7 with no need for rests
  • More driving done at night will reduce fuel usage and time lost in congestion
  • Reduces the requirement for air-conditioning and heating in vehicle cabs, indeed, most of the cab could be dispensed with, reducing vehicle cost.
  • Doesn't get sick or drive drunk
  • Won't organize to join a union.
Now, standard product management theory says that you don't attack a large immature market all at once.  Instead, you find a small number of niches containing customers to whom your technology is particularly compelling and you go after those.  You use those to a) work out the initial problems and bugs with the technology, and b) increase market acceptance, brand recognition, and the size and capability of your company to the point where you can tackle larger parts of the eventual market.

In this particular case, there are huge adoption barriers in the broad market.  As the New York Times story noted:
But the advent of autonomous vehicles poses thorny legal issues, the Google researchers acknowledged. Under current law, a human must be in control of a car at all times, but what does that mean if the human is not really paying attention as the car crosses through, say, a school zone, figuring that the robot is driving more safely than he would?

And in the event of an accident, who would be liable — the person behind the wheel or the maker of the software?

“The technology is ahead of the law in many areas,” said Bernard Lu, senior staff counsel for the California Department of Motor Vehicles. “If you look at the vehicle code, there are dozens of laws pertaining to the driver of a vehicle, and they all presume to have a human being operating the vehicle.”

The Google researchers said they had carefully examined California’s motor vehicle regulations and determined that because a human driver can override any error, the experimental cars are legal. Mr. Lu agreed.
So there's simply no way you could stick one of these systems in a commercial truck tomorrow and dispense with the driver.

I see two obvious initial niches: the military market, and an add-on extra for high-end luxury cars.

The reason DARPA funded the grand challenges in the first places is because the military is highly motivated to reduce the number of humans in harm's way in Iraq, Afghanistan, etc.  If you think of those convoys of trucks going through Pakistan and Afghanistan to supply the war operation, it would obviously be extremely desirable to get to the point where the trucks were automated, and only the accompanying armored vehicle protection force needed human drivers (who could stop or reroute the convoy as needed).  Legal issues are probably much less of a concern in that context (maybe have a driver just in the lead truck to talk to the Pakistani police when necessary).

If you can sell the technology to the military, you can work out a lot of the bugs and then when you try to sell to others, you can say it's battle-tested in Afghanistan etc...

The high-end luxury niche is going to have to be in the context of the existing legal rules to start with.  In other words, the car is going to have to be like the experimental ones that Google has now - there is complete equipment for manual driving, the driver can take over at any moment, and the driver is fully legally liable for whatever the system does.

Still, ask yourself this: are there enthusiastic techie guys with plenty of money who would love to be able to show off their car that would drive itself to lunch?  No shit there are - Silicon Valley is full of guys just like that (and doubtless a few women too).

Right now, of course, the driver of such a system would still not be able to make a cell call, etc, while the system drives.  Obviously, there would be a long process of changing the laws and regulations to slowly take account of the possibility of reliable automated driving systems (and a single spectacular accident could set the effort back years).  Still, with enough money spent on lobbyists and lawyers, and an increasingly long track record of actual use, I'm sure the legal system could slowly be induced to place more trust in the automatic driving system and give the driver more freedom to be working, texting, talking, or reading, rather than paying attention to the road.

So you establish yourself in those niches, you solve the legal problems, and then you start to move downmarket in the consumer space, and you start to move from the military into the commercial long-haul trucking market.  That is the really big prize financially, as the graph at the start makes clear, but also where you will have to take on the trucking unions.

Some markets will take a long time.  As commenters noted yesterday, in situations where the driver also does other things (like the Fedex man delivering the parcels to doors, or the ambulance driver wheeling the gurney) there's going to be much more difficulty.  Still, even here, there are going to be niches within niches.  Some Fedex delivery/pickup vans mainly go to businesses, and the businesses could be given a discount if they load/unload their own packets (eg with their own automated pallet jacks).  Even partial automation of the truck fleet will allow for a lot of operational efficiencies.  Then, as the decades go by (this is clearly going to be a multi-decade process), other automation possibilities will synergize and allow further removal of the human drivers from the system.  For example, maybe it will take the development of the voice-directed automated gurney before the ambulance driver can be pensioned off.  On the consumer side, older drivers may not accept automated systems - but eventually they will retire from the road and be replaced by younger drivers who have never known anything else.  Maybe the very wealthy will always insist on a human driver for their limos.

Still, absent a major push in the other direction, I see no reason why aDriveTM shouldn't eventually succeed in being a very large company.  The timing is uncertain - maybe it's still a little premature.  But, given that there's obviously massive amounts of money to be made, and that it will be perceived to be technologically cool, when did we ever not do something with those qualities?


Doyu Shonin/Risa Bear said...

So, you've read Player Piano, right?

Stuart Staniford said...

Risa - no, but I ordered it.

Unknown said...

Robot cars. I'm enamored of this eventuality not just for the cool factor. Thanks so much for sharing this in-depth look. I've totally added it to my geeky little notes.

Anonymous said...

I'm curious as to how you intend to (physically) get to the market a product which, by definition, is designed to put Teamsters out of work. Good luck with that. ;-)

You'd better start with automated car-carriers as your first product and hope they don't run down any of the picketers.

kjmclark said...

I added a comment to yesterday's post (sorry - catching up), that applies to today's as well. I don't think the road market is the best one to start with, though when you do, tractor-trailers are clearly the best first bet.

Stuart Staniford said...


Yes, the tractor thing is pretty close. And indeed I think the adoption barriers are much lower.

KLR said...

Player Piano is definitely on the prospectus for any class in Dystopian Literature 101. Unintended consequences of automation, plus Volunteer Fire Brigades.

A bit of catchup: Look Ma, No Hands! A Brief History of Self-Driving Cars. His first example is from 1933, the Wiki article driverless cars sensibly begins thereafter, with a 1977 experiment.

Bets are they'll keep the tuck drivers, with a tug-of-war between their newly increased laxness in the cab and no end of wage negotiations about their utility, with plenty of cash under the table for pols.

Had read that cruise control lead to an uptick in the rate of accidents; didn't find any real data but snopes says that more than a few people were under the impression that it would actually drive the car for them.

Greg said...

Two transition points (well, the same one, really):

Rental cars. No-drive-em rentals should have lower insurance costs, eventually. (The car knows the local traffic conventions and the terrain; tourists mostly don't.)

Corporate fleets, ditto for insurance. And more office work can be squeezed out of employees. "Taking the red-eye" might gain a new meaning.

Also, people may have missed my parenthetical allusion yesterday to automated fork-hoists being one of the "gateway" applications. There's probably about as many of these as there are tractors, and the spaces they operate in can be well-defined: "trainer wheels" for the technology.

KLR: when I first heard the term "cruise control", that's what I thought too. For about ten seconds. Then my knowledge of the state of technology kicked in. People who have lived from the era of horse-drawn carts to seeing the moon landing and 2001: A Space Odyssey would naturally be less sceptical of technology, I guess. Car makers really should have called it "speed lock" or some-such.

Greg said...

Posted too soon.

So: there's a clear line of incremental development here:

tractors: very low third-party risk;

fork-hoists, mining trucks, military all-terrain vehicles: more risk to nearby pedestrians, but the risk is manageable, one way or another;

low-speed road use (self parking, "assisted" reversing - especially with a trailer), special needs ("why should a blind person be denied the right to drive?"): consumer demand;

freeway use;

general use.

The good old profit motive provides all the impetus. Enter the lawyers...

Anonymous said...

We're starting to see automated vehicle parking systems on cars today...


Stuart Staniford said...

Hmmm - I wasn't aware of Mobileye, but they offer an automated lane keeping system... as an option for luxury cars. So that's a baby step in this direction.

Pangolin said...

I think rather than eliminating the driver entirely automated driving will be a means of reducing drivers wages and insurance liability.

I would see the first large scale use of this technology in container ports. Currently hundreds of trucks circle from container cranes to storage yards or rail yards on a constant basis.

If a developer can demonstrate that an unskilled "driver" can be used instead of a skilled teamster half of the labor cost for each unit is eliminated.

As a system accumulates hours of safe operation it will be used on increasingly large numbers of roads. Actual drivers are eliminated as safe hours of operation are proven to exceed human tolerances.

jewishfarmer said...

I realize I'm way late with this, but what struck me as interesting is the market that I see as most likely to absorb this wasn't at the center - the aging population of baby boomers and elders who are terrified of losing their cars and independence. The people who will seek this out won't be luxury buyers, IMHO, but the elderly. I have my doubts about whether it will be widely applicable in the time frame we have, but I don't have any doubts that if it is, the elderly will be primary drivers. Loss of car is loss of *everything* to the American elderly.