Pretty comprehensive approach - Only thing which was missed imo is Competition. The question was "Uber" Drivers - There are other competitors like Lyft. So the actual number will be lesser.
The 50% of population (leaving out < 20 yrs and > 60 yrs) needs to be applied to SF's population. SF's total population is 1M. So I'm wondering whether subtracting 1M from 3.5M is correct. It should have been 0.5M subtracted from the 3.5M leaving 3M for South Bay. The final number won't be affected much though.
Another possibility: do segmentation before TAM sizing based on age distribution. Because, you can assume more young people to live in SF while more family people (aged 40-60) could live elsewhere.
The data I would like to have access is 1. Age group of population 2.Number of people owning a car I would have given answer in terms of ranges and best & worst case scenarios.
feedback could include some other ideas of approaching this estimate, such as a % of vehicles on the road that are uber vehicles, or using published trips data from Uber to work backwards and estimate # of trips. I don't think we are necessarily looking for a number, but more on the approach to get to that estimate.
Few more thing I would do. Anything wrong? Are we including Uber Eats delivery drivers or just Uber Rides? According to me "peak" hours would last for 2-3 hours not just 1 hour. So after calculating the no of rides, I will split them equally across 2-3 hours.
A 1:1 ratio for teh nuber of ride to no of riders is a bad assumption. A rider would be taking at least 3-4 rides a week (depeding on the age group breakdown) . Taking a monthly rides approach is a better idea here imo.
Could someone please help to explain why assuming age distribution to be normal would yield a 3.5m for people aged between 18-60 from the total 7m population? Thanks!
Ages 18-60 would roughly be population set of 40 yrs which is half of the total population with 80 yrs life expectancy ( if normal distribution is considered). So it would be half of 7M which is 3.5M
As an ex-Uber driver, my answer would be. When it's peak hours and there's a red color on the Uber apps heat map. Drivers should be a lot. I Thank you...
would you factor in Covid exodus approximation from Bay area (Tech workers) into the target rider population (Demand) to calculate the Drivers (Supply)
Hi Bhavya! It would be helpful provided that you have other pieces of information to help with your estimation e.g. total drivers in the market x uber market share = uber drivers. Hope this helps!
Thanks for the example! I like when you imagined the rate of South Bay rides/hour might be lower due to it being less densely populated (therefore requiring more drivers per rider there)
I would rather leverage the data from the Uber itself, to understand the patterns and perform a predictive analysis. Because I feel these numbers from the data will be more accurate compared to the assumptions made with outside number. This would be my approach as a product manager.
When asked this question in an interview, I could either solve it in 10 minutes like she did with so many assumptions OR I could go into detailed user groups (like students, workers, special events/concerts) and consider trips per hour divide them into casual users/power users and do a weighted average --> This would take about 30-40 minutes. Isn't this type of a detailed approach better for interviews?
You will mostly get 5-7 mins for a question like this in an interview. Also its not about getting to right answer but to see if you think on your feet.
I think that the assumptions were not reasonable and open to debate. Assuming a data point doest mean that one has to make it super broad. I would have done a normal distribution of the age groups and then for each segment assign a % of users who would use uber on a Friday evening.
The big issues is that home girl is assuming that demand will stay consistent throughout the entire day. Also she couldve asked a clarifying question on whether or not if it was servicing for a holiday or something
She removed the population under 18 and over 60 coz they cannot be drivers but then used the # 3.5M for calculating riders. She got confused on the actors here
she forgot to tweak the number of people take rides per hour during covid time. since most of them would do WFH and not take uber ,hence 200k people will take ride per hour during covid.
Little surprising that she said people age 60+ are too old to be familiar with the technology. The way she phrased that plus her assumption around that statement is somewhat offensive.
Don't leave your product management career to chance. Sign up for Exponent's PM interview course today: bit.ly/3wS9B2P
Could also add point on market share for Uber here as well. Not all of the population will use only Uber.
this was the big gap in the answer
Exactly. Came to comment same but saw your comment.
Pretty comprehensive approach - Only thing which was missed imo is Competition. The question was "Uber" Drivers - There are other competitors like Lyft. So the actual number will be lesser.
Watching this helps me understand just how much an algorithm reflects the thinking and personality of the person writing it
not really great assumptions. 475K rides pr hour? why is everyone moving every hour?
The 50% of population (leaving out < 20 yrs and > 60 yrs) needs to be applied to SF's population. SF's total population is 1M. So I'm wondering whether subtracting 1M from 3.5M is correct. It should have been 0.5M subtracted from the 3.5M leaving 3M for South Bay. The final number won't be affected much though.
Another possibility: do segmentation before TAM sizing based on age distribution. Because, you can assume more young people to live in SF while more family people (aged 40-60) could live elsewhere.
The data I would like to have access is
1. Age group of population
2.Number of people owning a car
I would have given answer in terms of ranges and best & worst case scenarios.
The SF population of 1 million should have been substracted from 7.5 million. Correct me if I am wrong.
feedback could include some other ideas of approaching this estimate, such as a % of vehicles on the road that are uber vehicles, or using published trips data from Uber to work backwards and estimate # of trips. I don't think we are necessarily looking for a number, but more on the approach to get to that estimate.
Anybody know what software she Shalong was using to type down the points?
google docs
Few more thing I would do. Anything wrong?
Are we including Uber Eats delivery drivers or just Uber Rides?
According to me "peak" hours would last for 2-3 hours not just 1 hour. So after calculating the no of rides, I will split them equally across 2-3 hours.
Why does she start assuming at first when you should be asking questions first or clarifying what is it that he wants to do more specifically and why?
cant it be the no. of uber app installations in that area to make it simple?
Curious if the question is shared ahead of the interviews?
Commenting to follow.
Definitely not
Can someone answer where she got the number 8 from?
A 1:1 ratio for teh nuber of ride to no of riders is a bad assumption. A rider would be taking at least 3-4 rides a week (depeding on the age group breakdown) . Taking a monthly rides approach is a better idea here imo.
Could someone please help to explain why assuming age distribution to be normal would yield a 3.5m for people aged between 18-60 from the total 7m population? Thanks!
Ages 18-60 would roughly be population set of 40 yrs which is half of the total population with 80 yrs life expectancy ( if normal distribution is considered). So it would be half of 7M which is 3.5M
@@suchitrapalat I turned on the subtitles and saw she actually said uniform distribution...
As an ex-Uber driver, my answer would be. When it's peak hours and there's a red color on the Uber apps heat map. Drivers should be a lot. I Thank you...
😂
would you factor in Covid exodus approximation from Bay area (Tech workers) into the target rider population (Demand) to calculate the Drivers (Supply)
Would it help to consider number of total peak hours, and competitor market share/uber market share?
Hi Bhavya! It would be helpful provided that you have other pieces of information to help with your estimation e.g. total drivers in the market x uber market share = uber drivers. Hope this helps!
Thanks for the example! I like when you imagined the rate of South Bay rides/hour might be lower due to it being less densely populated (therefore requiring more drivers per rider there)
I would rather leverage the data from the Uber itself, to understand the patterns and perform a predictive analysis. Because I feel these numbers from the data will be more accurate compared to the assumptions made with outside number. This would be my approach as a product manager.
Hi Imraan! Thanks for sharing your approach!
475k rides per hour? Sounds reasonable :) Lots of missing parameters to drive the estimation, subpar performance i'd say
Does someone know the real number?
The age range is wrong, she’s saying no under 18 just for corporate. In reality a lot of rides for under 18 year olds.
When asked this question in an interview, I could either solve it in 10 minutes like she did with so many assumptions OR
I could go into detailed user groups (like students, workers, special events/concerts) and consider trips per hour divide them into casual users/power users and do a weighted average --> This would take about 30-40 minutes. Isn't this type of a detailed approach better for interviews?
You will mostly get 5-7 mins for a question like this in an interview. Also its not about getting to right answer but to see if you think on your feet.
this is so gooddddd
I think that the assumptions were not reasonable and open to debate. Assuming a data point doest mean that one has to make it super broad. I would have done a normal distribution of the age groups and then for each segment assign a % of users who would use uber on a Friday evening.
Really helpful
The big issues is that home girl is assuming that demand will stay consistent throughout the entire day. Also she couldve asked a clarifying question on whether or not if it was servicing for a holiday or something
awesome
She removed the population under 18 and over 60 coz they cannot be drivers but then used the # 3.5M for calculating riders. She got confused on the actors here
I think she was saying they cannot ride alone (under 18) or won’t use an app (over 69) - so riders either way
she forgot to tweak the number of people take rides per hour during covid time. since most of them would do WFH and not take uber ,hence 200k people will take ride per hour during covid.
I believe her assumption was "pre-covid times" so that would not be a factor in this case.
@@dhruvmatta99 maybe.
Little surprising that she said people age 60+ are too old to be familiar with the technology. The way she phrased that plus her assumption around that statement is somewhat offensive.
Uber corrupção
I must say that's a lot of assumptions that you made.
I don't think that's how it works though.
Nothing personal.
Assumptions is exactly how estimation questions work
@@pranamdaga8716
Assumptions must be made on relevant data sets.
Like the data set of car owners in the Bay area.