Estimation questions, like many other PM interview questions, are tough because they are so open-ended. There are so many paths you can go down that it’s often difficult to get started. Additionally, it can be hard to explain your process or do quick whiteboard math. In this lesson, we’ll see some strategies to counteract all those problems.
Most estimation questions will probably ask you to estimate something you have no clue about. Even after coming up with a structure and breaking down components, you might still struggle to come up with some final number. Here are some tips to break out of that rut in the interview. An example we will use throughout to demonstrate is “how many cars are there in Seattle?”
An excellent method to estimate a number is to figure out another related and known number. Instead of trying to just guess how many cars in Seattle, we can hypothesize that the number of cars is directly related to the number of families in the area. This latter quantity is definitely easier to derive a number for. The self-driving car example in our course shows this method in practice.
Once we have our related quantity, all we need to do is scale that number up or down. For example, if we assume there are 300,000 households in Seattle, it doesn’t mean there are 300,000 cars in Seattle. Each household may have, on average, 2 cars. So that means there are 300,000 households * 2 cars = 600,000 cars in Seattle.
In essence, what any method to estimate is trying to do is relate some unknown quantity to some known quantity and move from there. One of the most fail-safe options is to relate the unknown quantity to some known quantity in your life. Working with the Seattle question again, you might know that 50% of your friends have cars. Working from here, you can estimate that 40–50% of Seattle is probably around the same age as your friend group. The other 50% of the population might have more or less (and this is up to you to assume). If you assume that the population of Seattle is somewhere around 1 million, this would give you at least 1 million * 50% of the population same age as friends * 50% of your friends have cars = 250,000 cars. Then, you would want to estimate how many cars the other 50% of Seattle has (the group not similar to your friends). Obviously, this doesn’t seem as accurate as it could be, but it’s a starting point. From here, you can adjust your assumptions.
When all else fails, start with the extremes. What’s the realistic upper and lower limit on quantity X? You can then whittle down these limits through the course of the interview, ending with a much narrower range or even a concrete number.
For example, we know that the extreme lower limit on cars in Seattle is 0 and the extreme upper limit is probably around the population size in Seattle — around 1 million. Even that can give you a pretty good estimate. If you take the average of the two numbers, you get 500,000 cars in Seattle. The actual number is closer to 430,000. But, in the interview, we don’t care about correctness as much as we care about your process. So definitely whittle down your range with relevant assumptions.
Ask clarifying questions -
I'd like to lay down the high level data points that'll be used to solve this -
And then each of those can be broken down further as the next step.