The product manager interview playbook — Part II: Estimation questions

Estimation questions test your ability to think critically and deduce answers with little to no initial information.

They are an important part of product management interviews since they test your ability to:

1. Break down an ambiguous problem statement into its constituent parts
2. Take appropriate assumptions to tackle the absence of information/data
3. Get to a point of view that is backed up through a logical flow

Practically, these skills come in handy when you are solving strategic problems such as sizing TAM (total addressable market) for your product, analysing competitive performance or choosing between investments to be made for the best ROI.

The key thing to note is that estimations are just that — estimates. There is no accurate answer; but — there are good and bad approaches; and your approach will be the determining factor in the interviewer’s mind.

Let’s take a few examples of estimation questions, and a few directional pointers to think about them.

How many self driving cars would it take to map the US in 1 year?

The end objective of this problem would be to assess the total land miles that would need to be covered, and how many land miles a single self driving car could do in a day (which can be extrapolated to a year); this would then help answer the question of how many self driving cars are required to map the US in a year.

So, what elements would we need to think about?

1. Land miles could be studied by starting off with an assumption about the total land area of the US (in square miles)
2. We can then think about assuming the density of roads per mile; bringing in a factor of urban vs rural areas (assuming urban densities would be higher than rural densities)
3. Based on the above, we can come to a conclusion about the total land miles that need to be covered
4. After this, there is a question of how many miles a self driving car can cover in a day/month/year; we can consider the follow
5. Average driving speed of the car on highway vs residential roads
6. Assuming a certain # of hours per day when the car can be operational (say it needs to charge in the night), we can estimate the total distance it can cover in a day
7. There can be external factors such as weather conditions, road repairs, traffic and accidents etc. which can adjust the net estimate

Estimate how much revenue Airbnb makes in 1 year?

Let us assume that we want to estimate revenue in the US market only, in a typical year (controlling for major external impacts)

So what elements would we need to think about?

1. TAM in the US — so, say we look at the average size of the hospitality/travel industry and multiply it by Airbnb’s market share; this would give us a size of net transactions that are likely to be done in a year
2. Airbnb’s model typically has a commission rate per transaction — so this would need to assess the average transaction price, and the revenue per transaction
3. We can further break this down by Airbnb’s different products for a more accurate estimate
4. Alternatively, we can also consider an approach where we look at the total number of households and the typical number of vacation days per year; from there, we can assume a factor to assess how many transactions are likely to happen
5. To make an estimate more robust, always consider market factors as well, such as business/industry growth rates, seasonal impacts on travel and so on

Estimate Google’s play store earnings?

Start with an assumption — we are looking to estimate revenue in the global market, before payouts are made to developers

So what elements would we need to think about?

a. Number of app purchases made per user on average

b. Typical price paid per purchase

c. Total number of users

The above can be estimated for a monthly basis and then extrapolated to a year

So what data can we start with to assess the above?

1. Well, the total number of users can be a factor of the total number of smartphone devices and the market share that android holds
2. Most users would download apps that are free, so we can segment users and their typical purchases into cohorts such as (50% of app purchases are free, 30% are upto $1, 10% are $3 and 10% are $5)
3. These cohorts can be tuned by market as well (say, there would be a larger paid cohort in developed markets, and a larger free cohort in developing markets)
4. Further, we can consider app categories, their average paid/free ratio, typical pricing and the presence of in-app purchases
5. As a gut check, around 80% of Google’s net revenue is the ad business, so the revenue estimate from the play store should be in line with this fact, when considering Google’s total revenue

Estimate the number of Uber drivers in New York City?

Let us assume that we want to estimate the number of Uber drivers during peak hours

So what elements would we need to think about?

1. The total number of ride requests in a typical peak hour (let us assume there is driver availability to match each ride request made)
2. So, we can start off again with a TAM estimate in terms of total addressable population — this can be factored with age cohorts, % of commuters who use Uber over other modes of transport and so on (will let you think of more factors here)
3. Once we have this, we can think about how many rides a driver can do in a given hour — say for a typical peak hour, we have 100 ride requests, and a driver can fulfil 2 ride requests every hour — which means a net ask of 50 drivers every hour
4. This can be further tuned by assessing how many hours per day an Uber driver would work, the overlap of these hours with peak hours and so on.

You get the idea — I hope 🙂

Here are a few more practise questions:

• How many ping-pong balls fit in a helicopter?
• How much money does it cost Google Photos to store unlimited data for its users?
• How much laundry detergent is used every year, across the world?
• How many locksmith shops are there in the US?
• What’s the annual revenue of Pinterest?
• How much do NYC Broadway plays make per show?
• How many trees exist on Earth?
• Estimate Dropbox’s annual financial costs.
• How many earphones have been sold in the US in the past year?
• How much does a schoolbus weigh?

So, can we synthesise the general approach followed into a framework?

As we observe, the problem is usually tackled with the following components:

1. Clarify your understanding of the problem and the intended outcome of solving it
2. Create a broad structure of how you would want to approach the problem before starting your answer
3. Use knowledge of basic factual information
4. Break down the problem into individual components, making assumptions as required. Always state these assumptions to the interviewer and get a nod before you proceed
5. Estimations can also be done via proxies; example — the number of cars in a city can be tied to the population/# of households in that city
6. Bring in personal reference/experiences; create upper/lower bounds
7. Pivot and check in with the interviewer, if the direction is making sense
8. Finally, gut check/competitor check; see if solution has drawbacks or can be improved further
9. Summarise your overall approach