Exercise 1

Question 1- How many cars are on 101 between San Francisco and Mountain View?

Question 2- Assume you are new to Airbnb and have never booked a place with the service before. You are reviewing a unit’s details (web page) and like all the details of the unit. How can Airbnb increase the likelihood that you book this room right away? Assume that Airbnb has already implemented basic services such as free cancellation, good UX, etc.

You can submit your answers in the comments section of this blog post to receive feedback. You can also see other people’s answers, the feedback they received, and give your own feedback in the same section.

The article how to answer an estimation question in a product manager job interview helps you prepare for estimation interview questions. And the article how to answer a product improvement question in a product manager job interview helps you prepare for product improvement questions.

Here is a list of estimation questions for product manager job interviews and here is a list of product improvement questions for product manager job interviews.


Leave a Reply

8 Comment threads
3 Thread replies
Most reacted comment
Hottest comment thread
3 Comment authors
Rohan AttravanamAnil Punjabibijannabspm Recent comment authors

This site uses Akismet to reduce spam. Learn how your comment data is processed.

newest oldest most voted
Notify of

Here’s my first attempt at answering the cars question.

1. How many cars are on 101 between San Francisco and Mountain View?

That depends on a few factors, time and day being the greatest of them. First, let’s establish a few assumptions. The average car size is around 14 ft. Factoring in a minimum cushion between cars, we can consider the total to be around 20 ft. The distance between San Francisco and Mountain View is approximately 40 miles which is about 200,000 ft.

That means, at 100% saturation, one lane can fit 10,000 cars. Multiplying by 5 lanes gives us 50,000 cars one way or 100,000 cars both ways.

Of course both north and southbound are not always running at 100% saturation. Reality might better reflect a 70% to 80% saturation level during rush hour and a 40% to 50% saturation level during normal traffic hours.

If we assume an inverse relationship between north and southbound rush hour time period, the estimated total freeway saturation is around 110 to 130 percent or 55,000 to 65,000 cars as a ballpark estimate both ways.


Thanks. This structure and approach is very good. Maybe the 20 feet distance between end of two cars is too low given that when cars are driving in a highway, they usually keep some distance from each other. But this is just an assumption you’ve made and you can explain it during the interview. You will also want to ask one important clarifying question during the interview to address this: what time of the day are we estimating this number for? The answer to your clarifying question will help you with making the right assumptions on the traffic on each side of the highway. Great work overall.

Anil Punjabi

Question 1: About cars on 101.

Let’s break down the problem into two segments.
Peak Traffic & non-peak traffic.
Also, carpool vs regular.

Is the question about:
a) Max number of cars anytime during a day
2) Total cars EVER
3) Minimum number of cars on any given day
4) Average number of calls driving between SF & Mountain View.

Also, whenever there is peak traffic in one direction, there is basically free flowing traffic in the opposite direction.

Let’s start by counting the miles first.
Roughly between Mountain View & SF the distance would be about 40 miles.

Let’s start with analyzing just one such route.
Let’s simplify our problem statement to be from SF to Mountain View in the morning at 8am. And let’s only compute the traffic in one direction, the busy direction.

Let’s assume that the entire 40 miles of that route is filled with cars.

One mile = roughly 60K inches.
One car is roughly 200 inches in length.
Let’s add another 50 in front and 50 in back as buffer between cars. So each car takes about 300 inches.

So one mile could hold = 60,000/300= 200 cars in one lane in one mile.
There are 4 regular lanes = 800 cars
So for the entire distance of 40 miles, we would end up having 800*40 = 32,000 cars.

That does seem a bit high, doesn’t it.

During peak traffic we can assume that cars are basically next to each other only for about 50% of the route between SF & MountainView. The rest of the traffic is not as packed. So, each car would take 1000 inches of space instead of 300.

For 50% (bumper-to-bumper traffic zones)
One mile could hold = 60,000/300= 200 cars in one lane in one mile.
There are 4 regular lanes = 800 cars
So for the HALF distance of 40 miles, we would end up having 800*20 = 16,000 cars.

For 50% (NON-bumper-to-bumper traffic zone)
One mile could hold = 60,000/1000= 60 cars in one lane in one mile.
There are 4 regular lanes = 240 cars
So for the HALF distance of 40 miles, we would end up having 240*20 = 4800 cars.

TOTAL CAR IS PEAK LANE AT 8AM = 16000+4800 = 20,800

This is better.

Other ideas to consider:
> Consider one of those lanes to be a carpool lane.
> Consider more than 50% of that traffic to be non


Answer to question 1)

It takes 1 hour to drive from mountain view to san Francisco at about 70 miles per hour. So the distance between mountain view and san francisco is 1 hour x 70 miles / hour = 70 miles.

I have noticed that cars are usually 100 feet away from other on average when there is no traffic. Assuming there is no traffic in the highway now and all cars in the highway are keeping a 100 feet distance from each other, the number of cars per lane would be 100 miles / 100 feet. Given 1 mile is equal to ~ 5300 feet, we can assume the total number of cars per lane is 100 miles x 5300 feet per mile / 100 feet = 5300. There are about 10 lanes (5 southbound, 5 northbound) at any point in the road. Assuming all roads are equally occupied, total number of cars will be 5,300 x 10 = 53,000. Of course this number changes depending on the time of the day. As traffic increases, the distance between cars will reduce and number of cars per lane increases. At certain times of the day, traffic might be high on one side of the highway. In those situations, we will have to calculate the number of cars per side separately.

Answer to question 3)

I would present a new metric that indicates % of people who booked with Airbnb for the first time in that place or neighbourhood and were happy with their experience. A high percentage (a number close to 100%) addresses concerns regarding having a bad experience as a first timer. To calculate this number, I will simply divide total number of first timers who stayed at the place and booked with Airbnb again within 1 year. The nominator can also be total number of people who left a positive review for the host on Airbnb.

I should mention that it is possible that in cases where the number of bookings for a unit is not large, we might not have sufficient data to present this information. In this case, I suggest we look at the ratio described above for a neighbourhood or the whole building (where applicable) to be able to use more data points.


In the how many cars question, to determine cars per lane- shouldn’t we divide 70 miles/100 feet instead of 100 miles/100 feet since the distance between SFO and Mountain view is 70 miles? Am I missing something here?


You are right. It should have been 70 miles per divided by 100 feet.


AirBnB question:
Assume you are new to Airbnb and have never booked a place with the service before. You are reviewing a unit’s details (web page) and like all the details of the unit. How can Airbnb increase the likelihood that you book this room right away? Assume that Airbnb has already implemented basic services such as free cancellation, good UX, etc.

Customer : Somone with an intention to book a place through AirBnB for the first time. How can AirBnB provide them with the right tools/information to make a conversion decision.

Customer needs:
1. Find a place that meets their accommodation needs – Number of people, number of beds, rooms, pets etc (AirBnB does provide all this info).
2. Find a place that has good number of reviews – Again – this feature also exists.
3. Find a place that doesnt give them buyer remorse – As mentioned, they have free cancellation, they are not charged until they actually use it etc

4. Find a place that meets their price need – Now this is where AirBnB can make a difference. Currently, as a new customer to AirBnB – you get to see how much you will pay per night and what the cleaning fee is. However you do not know if this is a good deal or not as there is no comparison to what you would have paid with some other similar listing or through a hotel. A first time user who is budget conscious would have to go to Google or some other travel site to compare rates across hotels and AirBnB.

Solution: Build a compare at capability by web scraping similar listings across AirBnB and other travel sites to let users know that they are saving 20% more as compared to other listings. This would even make AirBnB listers get more competitive amongst each other and the customer will ultimately win.

Risk : It is hard to get comparative price data by scraping sites for similar product listings as all house listings are very different. However, AirBnB has standardized fields of data across every listing that can be initially used to create this price comparison repository.


Hi Nabspm,

I like pain you’ve highlighted and the solution you’re offering. I would have added one step between the need and the solution. I would have presented my creativity and thinking out of the box by listing a few different potential solutions, evaluating them based on certain criteria (e.g. I’ve seen a few people suggest the good criteria to use would be impact on customer experience, revenue, and complexity or cost of implementation), and then making a recommendation. This way, you’re also showing your ability to come up with multiple solutions and to evaluate trade-offs.


And my first attempt at the AirBnB question:

2. Assume you are new to Airbnb and have never booked a place with the service before…

Airbnb will probably have an understanding of their key metrics which correlate navigation behavior, time spent on site, frequency of visits or other metrics with conversion likelihood.

Based on their understanding, they could leverage a tracking mechanism which flags me as a likely candidate for conversion and serves me an incentive by simulating urgency through showing concurrent viewers, views this month or a live view into the booking schedule. They could also serve me an offer as a new customer. A modest discount could be enough help me move from consideration to conversion. A third approach would be to have a non-intrusive live chat option to help answer questions I might have as a new customer.

These different approaches could be A/B tested on site to see what works and what doesn’t.


Hi there

Thank you for submitting your answer.
You have some good ideas there. I think you can improve the structure of your answer by following the standard flow for answering a product improvement question. I’ve included the link for your review.


Feel free to post a new answer to this question. I’d be happy to share my feedback again.

Rohan Attravanam
Rohan Attravanam

Answer to question 2
If a user likes everything there are a couple of scenarios that’ll influence whether they will book the room right away.
1. If they truly have found a great place (subject to their experience with other rooms), at an affordable price. (Budget depends on customer)
2. If they are confident this is a better option than all others out there/they have looked at
3. Where they are in the cycle of the booking. (We can truly influence this, except give them a booking that’s valid anytime in the future. We could restrict it to 3-6 months)

Of these scenarios – instilling confidence is more easily achievable compared to others. So I’ll pick that route and find ways of improving on that aspect.

We could improve the users’ confidence by doing the following:
1. Their friends have already stayed there(FB check-in) and/or reviewed the place or the owner(if they have multiple places) positively.
2. We should provide an incentive to the owner to respond and be active in the reviews. This conversation brings out more light on owner’s personality/friendliness. A good number of ratings and reviews boosts confidence.
3. Include any changes in prices (in the past), and any potential changes in the future. For example, last minute booking has potentially been lower priced or higher priced.
4. Modify the page based on what user searched for and landed on the page. For example, “beachfront” search term should reorder the photos of a beachfront listing to show a view of the beach from the rental. Emphasis on the property being true to their search.
5. Provide other comparable options, like hotel prices, cheapest and most expensive at that time of the year.
6. Emphasis on things to do, walk score, amenities and previous customers’ favorites things about the unit.
7. How many people have booked this place after seeing other places on Airbnb?
8. Include the recommend to a friend score.
9. Text owner for a quick question or have a bot answer some common questions.

Matrix of Impact analysis – Likelihood of booking / Cost Impact / Feasibility

1. High / Medium / Hard (It’s highly unlikely that your friend would have stayed at an exact place. Including 2nd and 3rd-degree friends would improve this)
2. High / Low / Easy
3. High / Medium / Hard (Tough to have owners onboard this idea)
4. High / High / Moderate
5. High / High / Hard
6. High / High / Moderate
7. High / Low / Easy
8. High / Low / Moderate
9. High / High / Hard (Owner might be busy and might not respond. If it’s bot, we need to train the bot to learn from listing, understand NLP and reply back to query)

Based on this – I’d pick 2, 7 and 8 as they have high impact with low cost and easy/moderate feasibility.

Measurement –
We’ll need to test the group of new users who liked the listing (determined may be by favoriting the place) have never booked with Airbnb before. Test/Control and measure the following metrics-
1. Avg. time taken to make the booking by new users
2. Number of sessions/page-views before making the booking