During product management interviews, you may get a "sizing" or "estimation" question, where you're asked to estimate some business-related number.
This question tests both your quantitative skills and your ability to break down a complex problem to arrive at a workable conclusion efficiently. It's the kind of question that requires a bit of practice, but it's pretty straightforward after you've gotten the hang of it.
Examples of PM estimation interview questions:
"Estimate the number of mobile apps being downloaded by Americans each day."
"Estimate the number of microwaves being sold around the world each week."
"Estimate the total number of packages being delivered by UPS, Fedex and USPS each day."
"Estimate the total number of photos uploaded to Instagram each day."
A couple of things to note about these questions:
There is a sure-fire wrong way to answer an estimation question: making too few assumptions and jumping too quickly to a conclusion.
For example, a reasonable estimation question to get is "Estimate the number of apps being downloaded by Americans each day". It may be tempting to answer this way:
You've arrived at an answer efficiently, but you made very few assumptions, which really doesn't go a long way towards demonstrating your analytical abilities (and remember, that is the point of the exercise!). Also, with fewer assumptions, each of those assumptions is more likely to contain more assumptions within them, so it's harder for you to justify why you made them.
So how should you start an estimation? First, as with many interview questions, you should pause, reflect and, if desired, ask for additional context. This isn't always necessary for an estimation question if the interviewer provides some context already. But if you feel that some piece of information would be helpful, there's no harm in asking about it.
The starting point should be broad and very likely begins with some population of people. It's useful to know that the population of the world is roughly 7 billion, the population of China and India are both around 1 billion, and the population of the US is roughly 300 million. Here, we see one common trick with estimation questions - you can round somewhat liberally to get yourself "rounder" numbers to simplify the arithmetic a bit. You won't want to round everything or round too aggressively - that will seem like a cop-out - but rounding the US population, which is around 328 million, to either 300 or 350 million is acceptable.
The next step is usually to break the starting point down along some axis - maybe it's people in different age buckets (simple assumption: people are evenly distributed in age between 0 and 100), an urban / suburban / rural divide (⅓ in each), people who are in the workforce versus not (50/50), or something else. Pick a breakdown that makes sense for the question, and then explain why you picked that.
Then, you will likely be making assumptions about the rates at which each subgroup does a certain action, or the percentages of each that have a certain characteristic, or something similar. You may find yourself breaking the subgroups down again, or crossing some off of the list because you don't think they contribute towards the final number you're aiming for. Generally, the goal is to arrive at some kind of estimate of the metric in question for each subgroup, which you would then add up across all the subgroups to get your final answer.
With each step you take, there are two things to keep in mind. The first is to make a reasonable assumption, and the second is to verbally justify why you made that assumption. There is usually a range of assumptions that seem reasonable, but there are also ones that seem obviously wrong. For example, if you assume that 90% of the elderly are posting on social media everyday, that will probably raise eyebrows. Generally, the interviewer is happy to accept reasonable assumptions, though they may push back a bit on some to test your thinking or to suggest a piece of information that they want you to incorporate. But more importantly, you must ensure the interviewer is able to follow you as you go through this exercise. You should be organized enough and clear enough about your logic and calculations that they clearly see how you arrive at each next step.
The general approach of starting broad and breaking things down can take very different forms depending on the question. For example, maybe the interviewer will even give you a starting large number to use. Or, the question might be better answered by starting with a different metric that's not a number of people - perhaps the total number of cars, or cell phones, or videos online, or something else. In these cases, if you don't know a particular large statistic (and chances are you don't), it's a mini-estimation question starting with humans to get to the other large number that would be convenient for the question.
In summary, here are the tips for estimation questions:
Once you've arrived at a final estimate, you're not done yet! You should proactively transition to doing a "gut check" of your answer - "This feels low / high to me because of X, Y, Z reasons". Also, as you're doing the estimation, mark all numbers that are assumptions; at the end, review all the assumptions to point out any that you think are either most unfounded, or which appear to make the biggest difference to your final estimate (arithmetically).
💡 Got a PM interview? Our PM interview drills help get you in top form
Estimation questions are also notorious parts of the consulting interview process. There is a lot of good advice out there about estimation questions in that context, so you may want to leverage online resources for that part of the consulting interview, including [other content from RocketBlocks].
A product estimation question is usually less rigorous than a consulting estimation question. An estimation question in a PM interview is also more likely to be framed in the context of a particular product.
Example estimation question from the RocketBlocks Drills database:
"You're a product marketing manager on the Pinterest team that is focusing on e-commerce opportunity around home remodels in the US. Given Pinterest's heavy design focus and wealth of content that relates to interior design, the executive team is considering whether it should push further into content and commerce around home remodels. The exec team believes that if it could capture a fraction of the lead generation business around home remodels it would be a big boon for their top line revenue story. As a result, they've asked you to estimate the size of the expense on US home remodels each year."
This is a great example for a number of reasons. One, it's framed in the context of a product decision but is estimating the size of a market, so it's a fun mix of product and business. Second, it's about a relatively approachable topic - home remodeling - so hopefully it won't be too hard for us to think about the assumptions involved. Third, as stated, it feels like enough context to get a start on it. Thus, for the sake of this example, we won't ask follow-up questions for more information.
(In a real interview setting, especially an in-person one, you'll likely have a whiteboard to work with. For the sake of this medium, we'll just try to describe everything in writing.)
We want to estimate the size of the expense on US home remodels each year, which I'll interpret as the total amount of money spent on home remodels. Given Pinterest's focus on e-commerce and individual users, I'll assume this is about individual homeowners and their remodeling projects, and not about corporate or enterprise remodeling companies, like a company that might remodel hotels for example. [In reality, you would probably want to phrase this as a question about this assumption, to see if the interviewer wants you to include this alternate userbase too.]
To estimate this, I'd like to start with the population of the US and think about just how many remodeling projects are going on. There are about 300 million people in the US. When thinking about home remodeling, we want to think about households, and there are roughly 3 people per household, so around 100 million households. I imagine that remodeling projects will differ in terms of frequency and size by whether the house is in an urban, suburban or rural locale. I believe in the US, it's about a ⅓ split across those three categories. For the sake of estimation, I'll assume that it's roughly 35 million households in urban and suburban areas and around 30 million in rural areas.
Now that we have a sense of the distribution of households, it'll help for us to think about how many of those households actually own their house - my assumption being that people who own their house are at some point going to think about a home remodeling project, whereas those who rent probably won't, since they don't have as much of an incentive to invest in their house. My guess is that home ownership rates are lower in the cities than in the suburbs or rural areas. Thinking about cities, there are some cities where I think ownership rates are very low - for example, the New York City and San Francisco markets seem pretty dominated by rental buildings, and these buildings tend to have many units, which will further skew us towards the proportion renting. But, there are probably a decent number of other cities where home ownership rates are slightly higher, which counterbalances that. All in all, I'm going to assume that around 40% of households in cities own their home - 40% of 35 million would be 14 million. Home ownership rates in the suburban and rural areas are probably much higher - I assume that most households there own their homes, since that's part of the appeal of not living in the city. I imagine the rate is a little bit higher in rural areas than suburban, since real estate in rural areas is incrementally a bit more affordable. So I'm going to assume 80% of suburban households, or 28 million, own their homes, while 90% of rural households, or 27 million, own their homes.
So we now have a sense of how many households own in each category - but not every household is going to be doing a remodeling project every year. It seems like a fairly rare thing to do given the expense. So perhaps we can assume that, across the board, households are only doing a remodeling project once every 10 years or so. That means in any given year, 1.4 million households in cities, 2.8 million in the suburbs and 2.7 million in rural areas are in the market for remodeling.
Next, what about the price of the remodeling? I have to imagine that these projects are more expensive in cities, then next most expensive in the suburbs, then the cheapest in the rural areas, given patterns in cost of living. I haven't done any remodeling projects myself, so I'm going to take a guess about how much they usually cost. In terms of order of magnitude, the average employed American is making a salary in the tens-of-thousands. If a remodeling project were to cost in the hundreds or single-digit-thousands, I imagine they'd probably take place much more frequently, so it feels like it should be more expensive than that. But if a remodel costs over a hundred thousand, then that feels too expensive for most Americans, especially given that houses generally cost in that order of magnitude too. So I'm going to guess that remodelings usually cost in the tens-of-thousands magnitude, and probably on the lower end of that range. So for now, I'm going to assume that a remodeling project in a city costs $30,000, in the suburb $20,000 and in a rural place $10,000.
Multiplying the number of households in each type of region interested in a remodeling project per year and the average cost of such a project in that type of region, I get 1.4 million times $30,000 equals $42 billion for urban areas, 2.8 million times $20,000 equals $56 billion for suburban areas, and 2.7 million times $10,000 equals $27 billion in rural areas. Adding those three together, I arrive at $125 billion in terms of total US spend per year on home remodeling projects.
Now I want to reflect on how this number feels. To be honest, it feels like the right order of magnitude, but it might be a bit low. US GDP is around $14 trillion - and remodeling is definitely not such a big part of the economy that it would be more than $1 trillion. But, $125 billion would just be around 1% of annual GDP, which feels a bit low given that it's something most homeowners think about with some frequency; I imagine it's a big topic of conversation, and home ownership is a relatively desirable goal in the US. If I had to guess, I imagine that, out of all my assumptions, either the frequency with which people do remodeling projects - once a decade - is too low, or the average cost - dependent on urban / suburban / rural - is a bit low. So if I could look for better data on any of these assumptions, I would probably start with those two. Also, at the very beginning, I decided to focus only on the "consumer" home remodeling market. I imagine there's a pretty decent amount of business happening on the B2B side, so perhaps that would count towards an overall estimate of home remodeling expense in the US, even if that's not Instagram's target audience right now.
This answer dove right in, but started by stating a key assumption - that we would focus on the B2C situation, not the B2B one. This seems reasonable given that the product is Instagram, but it's good that the response circled back to this at the very end to call it out as a potential source of error.
Home remodeling lends itself to starting with the number of homes, or in this case households, in the US. It's relatively easy to go from the population of the US, a number which any candidate should remember, to the number of households (the ‘divide by 3' rule is convenient if you assume the US has 300 million people).
From there, the answer chose to further break things down by an urban / suburban / rural divide and then make assumptions around home ownership levels and remodeling costs in each of those three categories. Dividing between urban / suburban / rural makes sense for this question; there aren't many other dimensions that you can divide households by. In this case, the answer also remembered that another fact we need is how often a homeowner does a remodeling project. This is a useful reminder that almost all estimation questions implicitly have a time range they ask about, so it's important to reflect that at some point in the estimation.
The candidate made a number of assumptions - home ownership levels, average cost of remodeling, average frequency of remodeling, and across the urban / suburban / rural divide. None of these assumptions felt too crazy, and the answer tried to provide a quick bit of explanation for each.
Note that there was a bit of rounding of numbers at the beginning (it's not a coincidence that the candidate chose to split the regions by 35% / 35% / 30% - it makes the math a bit easier), but less rounding towards the end. Maybe there could've been a bit more rounding as the calculation proceeded - but remember, don't be too liberal with it!
The candidate arrives at a final estimation and immediately reflects on it. It's totally fine to say that your answer feels high or low - as long as you state why and then think about which assumption(s) might be wrong.
If this were a real interview, then the interviewer might interrupt at more points in the middle and question specific assumptions. It doesn't always mean they disagree with an assumption, but you should engage with their questions and apply a combination of common sense and user-focused thinking to either push back or adjust your estimation.
Real interview questions. Sample answers from PM leaders at Google, Amazon and Facebook. Plus study sheets on key concepts.