Is it possible to know the number of vehicles deployed, and thus miles per vehicle? In other words, is the flattening the result of fleet redeployment or something else?
The best way to estimate that is to focus on the quarterly Waymo safety reports. There is sufficient data to understand daily miles. I just assume all credible companies with autonomous taxi services strive toward 200 miles / car / day. For example in Q1, Waymo reached about 60,000 miles per day of R O miles. 300 cars is a decent rule of thumb. While this will vary slightly in locales like PHX that are more sprawling I think that is a good guesstimate. The safety reports lag and Waymo does not release mileage for new markets until they reach a level where they can reliably compute accident rates by class. It seems to take them about 10M miles in market to produce a full set of safety projections in a market. This data may be more useful than the CPUC in CA because R O miles with employees and even trusted testers is excluded by the CPUC in certain circumstances. At least I'm my estimation that is probably why the 'rides are dropping or leveling in CA'. The reality is Waymo has bundled a series of challenges into one in order to scale broadly in the US. That includes highways, Zeekr testing, construction area problems, weather edge cases and the recent mandate of dynamic ODDs driven by regulators.
Hmmm. That’s a bit of circular logic. The reason I’m asking about the number of cars is to determine if trips/car are dropping or if they’re taking cars out of the market. Certainly, we can assume if they’re seeing lower demand, they’ll reduce inventory in-market. But they might also be moving cars between markets for strategic reasons.
Not circular exactly. The number of cars in a given market is a derived variable because companies don’t release the data for competitive reason. In extreme cases silly folks create ‘trackers’ but they self-select for superfans so the data is largely useless. Treating all companies as competent (benefit of the doubt) simply assumes all companies to converge to 200 mi/day/car. If they don’t they can probably be ignored. If they do, the normalized car count can be estimated. Picking a value other than 200 is fine I guess. It seems to be an industry rule of thumb associated with competence.
As for the other question trips/car I consistently hear 24 trips/day as a good standard to gauge scaling also. I have seen good data this number shrinks in very sprawling markets like Phoenix to ~20 which makes sense. Highway service will mitigate the numbers of course. Highways ultimately will increase miles/trip and perhaps shrink deadhead.
Good point, that may be a contributing factor but I think it's a pretty small number of Trusted Testers that have access. And then of course since the rides were free, they of course took a ton and skewed the number of 'free Ojai rides' :)
I think seasonality may also have played a part as some others have noted.
The freeway pause reads to me as the finding rather than the confounder. If pulling one road type offline moves both trip count and average trip length, then highway competence is probably what separates a short-hop Uber substitute from something that can actually replace owning a car.
I definitely took Waymo less during the period where freeways were not available but they also have a lot more demand than supply during peak times, so the effect probably contributed but maybe not a ton.
I think it is important to emphasize--and understand--that Waymo is still very much in market experimentation and development mode. Of significant relevance is that they continue to price their services higher than Uber, so their growth is not even an apples vs. apples comparison. And they clearly are more fleet size constrained than Uber in SF and LA markets. Their Ojai vehicles are in "training mode" here in San Diego and for multiple reasons I think that vehicle will benefit their market situation significantly over the next year and beyond. Harry's recent posting that included an analysis of Uber take rates--the key to their financial turnaround--was illuminating not only about Uber, but also provides insight into Waymo's strategic options for ultimate winning the marketplace battle. Uber's ability to push prices for its customers upward provides a large protective tent under which Waymo can effectively compete without overt price competition as it scales its services and absorbs what must be huge implementation costs per new city, if only for the cost of the vehicles. (And training in each market is not cheap, it involves human oversight and technology tweaking.) In addition, at some point, as the CA markets they are aleady in become more mature, spending more on a Waymo trip than the comparable Uber trip becomes less attractive to customers, although I suspect by now that at least 50% of their customer base is not going back to Uber or Lyft. The point is that a focus on a few months of ridership number trends in existing markets doesn't tell us much about Waymo's prospects more generally, particularly given that Uber's approach to achieving profitability is a gift to Waymo to enable it to become established in each of the metro markets it enter. The economics of robo-taxis at scale will kick in over the next 2 to 3 years and our focus on these early days experiences will no longer be relevant, even as they are certainly interesting now.
Great discussion. The market types are bifurcating for Waymo. Seems likely they may convert more Zeekrs by the end of 2026 than all the I-Paces in 6+ years of operation! The narrow CPUC analysis over time is a case of lies, damn lies and statistics I think. The early CPUC datasets included way more details from Waymo but then CPUC overreached. As a result little detail is now provided because of lots of redactions. Unfortunate.
Mid 2025 forward was a big bet from Waymo to simultaneously scale to highways, dynamism in weather, dynamic ODDs b/c of mandates, etc. That is likely the reason for the seeming pause I think.
Authorized to serve in 18 counties in CA likely means broad expansion in the core services in SF & maybe LA and a large increase in trip lengths. Expansion into the East Bay in SF will change the nature of the trips and can mean service at SJC, SFO & maybe OAK as well as regional airports. LA expansion feels like a mess IMO. It should make sense but even having an LAPD car a mile away does not protect riders or the cars. I think they will be cautious to make investments in LA when SD and SAC will be better spent for example. Most of the other cities new to service with Waymo are positioned with heavy dependence on highway driving. The miles per trip will grow I think.
Is it possible to know the number of vehicles deployed, and thus miles per vehicle? In other words, is the flattening the result of fleet redeployment or something else?
The best way to estimate that is to focus on the quarterly Waymo safety reports. There is sufficient data to understand daily miles. I just assume all credible companies with autonomous taxi services strive toward 200 miles / car / day. For example in Q1, Waymo reached about 60,000 miles per day of R O miles. 300 cars is a decent rule of thumb. While this will vary slightly in locales like PHX that are more sprawling I think that is a good guesstimate. The safety reports lag and Waymo does not release mileage for new markets until they reach a level where they can reliably compute accident rates by class. It seems to take them about 10M miles in market to produce a full set of safety projections in a market. This data may be more useful than the CPUC in CA because R O miles with employees and even trusted testers is excluded by the CPUC in certain circumstances. At least I'm my estimation that is probably why the 'rides are dropping or leveling in CA'. The reality is Waymo has bundled a series of challenges into one in order to scale broadly in the US. That includes highways, Zeekr testing, construction area problems, weather edge cases and the recent mandate of dynamic ODDs driven by regulators.
Hmmm. That’s a bit of circular logic. The reason I’m asking about the number of cars is to determine if trips/car are dropping or if they’re taking cars out of the market. Certainly, we can assume if they’re seeing lower demand, they’ll reduce inventory in-market. But they might also be moving cars between markets for strategic reasons.
Not circular exactly. The number of cars in a given market is a derived variable because companies don’t release the data for competitive reason. In extreme cases silly folks create ‘trackers’ but they self-select for superfans so the data is largely useless. Treating all companies as competent (benefit of the doubt) simply assumes all companies to converge to 200 mi/day/car. If they don’t they can probably be ignored. If they do, the normalized car count can be estimated. Picking a value other than 200 is fine I guess. It seems to be an industry rule of thumb associated with competence.
As for the other question trips/car I consistently hear 24 trips/day as a good standard to gauge scaling also. I have seen good data this number shrinks in very sprawling markets like Phoenix to ~20 which makes sense. Highway service will mitigate the numbers of course. Highways ultimately will increase miles/trip and perhaps shrink deadhead.
Isn't the drop in weekly *paid* trips just due to Ojai being free (up until very recently)?
The stats will jump up again now that the Ojai trips are paid.
Good point, that may be a contributing factor but I think it's a pretty small number of Trusted Testers that have access. And then of course since the rides were free, they of course took a ton and skewed the number of 'free Ojai rides' :)
I think seasonality may also have played a part as some others have noted.
The freeway pause reads to me as the finding rather than the confounder. If pulling one road type offline moves both trip count and average trip length, then highway competence is probably what separates a short-hop Uber substitute from something that can actually replace owning a car.
I definitely took Waymo less during the period where freeways were not available but they also have a lot more demand than supply during peak times, so the effect probably contributed but maybe not a ton.
I think it is important to emphasize--and understand--that Waymo is still very much in market experimentation and development mode. Of significant relevance is that they continue to price their services higher than Uber, so their growth is not even an apples vs. apples comparison. And they clearly are more fleet size constrained than Uber in SF and LA markets. Their Ojai vehicles are in "training mode" here in San Diego and for multiple reasons I think that vehicle will benefit their market situation significantly over the next year and beyond. Harry's recent posting that included an analysis of Uber take rates--the key to their financial turnaround--was illuminating not only about Uber, but also provides insight into Waymo's strategic options for ultimate winning the marketplace battle. Uber's ability to push prices for its customers upward provides a large protective tent under which Waymo can effectively compete without overt price competition as it scales its services and absorbs what must be huge implementation costs per new city, if only for the cost of the vehicles. (And training in each market is not cheap, it involves human oversight and technology tweaking.) In addition, at some point, as the CA markets they are aleady in become more mature, spending more on a Waymo trip than the comparable Uber trip becomes less attractive to customers, although I suspect by now that at least 50% of their customer base is not going back to Uber or Lyft. The point is that a focus on a few months of ridership number trends in existing markets doesn't tell us much about Waymo's prospects more generally, particularly given that Uber's approach to achieving profitability is a gift to Waymo to enable it to become established in each of the metro markets it enter. The economics of robo-taxis at scale will kick in over the next 2 to 3 years and our focus on these early days experiences will no longer be relevant, even as they are certainly interesting now.
Great discussion. The market types are bifurcating for Waymo. Seems likely they may convert more Zeekrs by the end of 2026 than all the I-Paces in 6+ years of operation! The narrow CPUC analysis over time is a case of lies, damn lies and statistics I think. The early CPUC datasets included way more details from Waymo but then CPUC overreached. As a result little detail is now provided because of lots of redactions. Unfortunate.
Mid 2025 forward was a big bet from Waymo to simultaneously scale to highways, dynamism in weather, dynamic ODDs b/c of mandates, etc. That is likely the reason for the seeming pause I think.
Authorized to serve in 18 counties in CA likely means broad expansion in the core services in SF & maybe LA and a large increase in trip lengths. Expansion into the East Bay in SF will change the nature of the trips and can mean service at SJC, SFO & maybe OAK as well as regional airports. LA expansion feels like a mess IMO. It should make sense but even having an LAPD car a mile away does not protect riders or the cars. I think they will be cautious to make investments in LA when SD and SAC will be better spent for example. Most of the other cities new to service with Waymo are positioned with heavy dependence on highway driving. The miles per trip will grow I think.