Estimate the number of self-driving cars (SDC) needed for the US?

estimation Guesstimate market estimation Market sizing
New answer on Jun 29, 2020
6 Answers
3.4 k Views
PK asked on Mar 20, 2020
Director of Product at PayPal. Based in SF. Worked at Twitter, Yelp, Salesforce, Houzz in the past.

Clarifying questions:

  1. Are we planning launch Self-driving cars for the first time or will it exist with the current set of cars in the recent future?

    1. Assuming this for the initial launch

  2. Car hailed by a standalone app or by existing tech product?

    1. Assuming via Google maps

  3. Self-driving cars will serve the need for passenger/personal transport and not freight, delivery, etc.

High-level equations:

# of cars = (Peak demand for cars per day)/ (Capacity of each car)

Assuming each car’s capacity is 4 passengers.

Peak demand for personal cars = Demand in urban + demand in rural areas

Let’s calculate TAM for self driving car:

US population is 300M, our target users are in the range 15-65

Assuming the average life expectance of 80 years, the target user count is:

40/80*300M = 150M users

Of this, I am assuming 50% will continue using their personal cars for commuting. Of the remaining 50%, I am assuming 80% will have the financial means to actually request an SDC (Since in 1 out of 8 in the US is considered poor). This leaves our TAM to be 0.8*75M

Of these 60M users, 20% of the US is urban and 80% of the US is rural.

12M urban and 48M rural

For the initial launch, the self-driving car may not get complete attention due to worry about new technology. I am assuming that initially for the first few months the adoption rate will be around 40% in urban cities and 10% in rural cities.

Thus TAM will be 4.8M urban and 4.8M rural. → 9.6M total

Assuming this peak demand for 12M cars over a 24-hour time frame. Also assuming each self-driving car will take 20 mins to complete ride for 1 passenger.

In one hour an SDC can complete 3 rides/passengers. In 24 hours it can complete, 72 rides.

Thus in order to serve a peak demand of 9.6M, we need → 9.6M/72 ~ 150K self-driving cars would be needed for initial launch.

Any feedback on my approach and assumptions?


Overview of answers

  • Upvotes
  • Date ascending
  • Date descending
Best answer
Content Creator
replied on Mar 20, 2020
McKinsey / ex-Interviewer at McKinsey / I will coach you to rock those interviews

Hi Piyush,

I've read carefully all your assumptions, looks very good! In the real estimation within the case probably the second part would be left out (dividing demand by # of rides) because it becomes too cumbersome. You would probably just need to estimate it from the population top town, would arrive at 9.6 M and the interviewer would already stop you there.



Was this answer helpful?
PK on Mar 20, 2020

Thanks for your feedback.

Anonymous replied on Mar 20, 2020

Hi Piyush,

that sounds reasonable to me.



Was this answer helpful?
Anonymous on Mar 22, 2020

I used to work in the Automotive Practice before on the topics related to autonomous vehicles. If you have any further specific questions, feel free to drop a line. Good luck! André

RUBEN replied on Apr 24, 2020
Junior Strategy Consultant, London - Looking for New Opportunities

Hi Piyush,

Looking at the question it seems you want to estimate how many car owners in the US will be early adopters and buy a SDC. However, you start developing the answer very nicely to end in what seems to indicate that you are estimating the number of SDC that will be initially adopted for raid hailing services.

If you want to estimate the later, I would use the other post you have where you solve how many Uber cars are in the US -remember to add Lyft and other minor competitors and assume that the companies would replace x% of their fleet with SDC.

if you want to estimate the former, your structure is very good but you tend to overcomplicate yourself. Start with a solid foundation and be sharp to get a good number fast. For example, the US has one of the highest ownership vehicle rates per 1,000 inhabitants in the world. On average, a household owns two vehicles. Therefore, state very clearly that you are estimating just cars. I guess every household has at least one car, and the average household size must be around 2.5. Thats around 130 million cars.

Then you can estimate the % of early adopters and also take into account that X cars would be replaced that year. So you can also add to your calculation how many people would replaced their dead car with a new SDC.



Was this answer helpful?
Content Creator
replied on Mar 21, 2020
McKinsey | Awarded professor at Master in Management @ IE | MBA at MIT |+180 students coached | Integrated FIT Guide aut


Was this answer helpful?
Anonymous replied on Jun 29, 2020


Your approach is indeed a good one.

If you want to master your market sizing skills I suggest that you should focus on segmentations patterns. You can use the following segmentation for market sizes:

-Demographics (Age, education, income, family size, race, gender, occupation, nationality)
-Behavioral (Purchasing behavior, customer journey stage, occasion & timing,
customer loyalty & interest, risk tolerance, user status)
-Psychographic (Lifestyle, personality traits, values, opinions, interests of consumers)
-Geographic (Geographical boundaries)

-Company characteristics (Industry, company size, number of employees)
-Geography (Geographical boundaries)
-Purchasing Approach (Occasion & timing, customer capabilities, nature of existing relationship)
-Personal Characteristics (Loyalty, risk attitude, user status)

-Demographics (Type of agency, size of budget, the amount of autonomy)
-Geographic (Geographical boundaries)
-Government Tier (Federal , State, Local, Quasi-governmental, International)
-Bid type (Closed, Open)

But sometimes you don’t need to segmentation. Here is an example of case that could be solved with high level top down approach - estimate the size of credit card market in the US:

Was this answer helpful?
Content Creator
replied on Mar 23, 2020
McKinsey | NASA | top 10 FT MBA professor for consulting interviews | 6+ years of coaching

I find it reasonable.


Was this answer helpful?
Daniel gave the best answer


Content Creator
McKinsey / ex-Interviewer at McKinsey / I will coach you to rock those interviews
Q&A Upvotes
121 Reviews
How likely are you to recommend us to a friend or fellow student?
0 = Not likely
10 = Very likely