I was recently asked to “guess the number of images clicked at the Eiffel Tower in a day” in a job interview.
I started writing the problem down right away and structured it like any other estimating problem. segments of the general public who might take pictures there, the size of Paris, the number of tourists there on any given day, etc.
The interviewer questioned me about whether there might be a more accurate way to estimate this halfway through my response. After giving it some thought, I came up with a suggestion that seemed entirely illogical to me. I proposed that we look at hashtags associated with the Eiffel Tower on Twitter and Instagram. Count the posts and then modify it depending on additional suppositions. The interviewer nodded in agreement, and we moved on to talk about how to use social media posts to gather the necessary data, the sample size needed for a hypothetical survey, etc.
My ability to look in the correct places for the answers and my statistical analysis skills were put to the test by what initially appeared to be a general estimation challenge. Is this a superior way to estimation questions in general given that this is the basis of product management—gaining insights from the data points that are readily available? to propose a strategy rather than come up with a number?
The approach rather than the quantity is always the focus of this kind of discussion. According to my past experiences, I would say yes to applicants who could execute the strategy you initially used well. However, I would always be more enthusiastic about candidates who were able to employ workarounds and original hypotheses to get to the solution. Ultimately, you should approach these issues in whatever way makes you feel most comfortable, but never be afraid to take a chance (as long as you can defend your reasoning)!
@SamanthaYuan, totally agree. You can directly calculate an exact number using the social media hashtags method, at least according to your initial hypothesis. This is helpful, but it necessitates the same risky set of assumptions as any other estimating question.
I do, however, believe that looking at “knowledge people are offering for free online” is a fantastic place to start for some study as a PM. Why hold focus groups or interviews when consumers are freely giving their opinions on TikTok, Instagram, and Amazon?
Absolutely, I’m a fan of questions that have “book correct” way to answer them, then a multitude of different ways to arrive at the same answer.
- If they can’t find a starting point, then that’s a no-go for me.
- If they give the “book correct” answer, then I’ll assume they’re a good candidate for a junior PM role.
- If they can come to a quick estimate, then triangulate to reasonable answer then I’ll assume they have enough experience under their belt for a more senior role.
Estimation questions work well because it puts PMs in a position that they’ll frequently find themselves in… where they’re presented with a problem that is seemingly impossible to even begin. How they wind their way through to an answer can tell you a lot about how they approach the unknown.
Yeah so this is something that comes up in consulting interviews a lot. It’s called “finding a proxy”. Often times what you want to measure is not feasible (too timely, costly, etc.) so instead you find something else that could function as an easier substitute and still a provide a roughly accurate estimate/answer to the question you’re asking.
Yes, it is. Typically, that question is looking for ways in which you’d guesstimate using trustworthy data points that are most readily accessible (even though there is technically no right answer). It also assess your capacity to deal with insufficient data to solve ambiguous questions.
The best way to go about it is:
- Take two approaches (or proxies) and pick one with a solid reasoning to do so.
- Segment the problem
- Take round figures and perform quick and dirty calculations.
- For bonus points, see if you could convert it into a $-value. Eg: of you’re trying to sell something to people clicking pictures at Eiffel tower, the total addressable market size is x number of people → $y.
It’s a super cool approach and I appreciate the write up. I wouldn’t have even thought of this. That being said, if you brought up the IG approach right off of the bat, it wouldn’t satisfy the original question which was to provide and estimate right then and there. I’m curious to hear people’s thoughts on if that matters, or if it’s more the approach.
@KaranTrivedi, It’s definitely the approach that matters. In today’s world, I would argue that if you aren’t looking at these free-for-the-taking information sources, you’re leaving money on the table, so to speak. For example, if you’re building an app, any given app store has thousands of reviews of your competitors’ apps, spelling out in specific detail exactly what they’d like to see in similar apps. If you don’t pick up that free info as a foundation for your own research, what are you even doing? All good novel research starts with a literature review!
In this specific case, the number of hashtags doesn’t equal the number of pics snapped, not even close. Maybe only 1/10 pics snapped are posted. Maybe 3/10 are from previous days or even years. Maybe only 4/10 people taking pics actually post on IG at all. But, it does give you an actual directly verifiable number to start with, which does give you a leg up on many interview-style estimation methods, and shows awareness of modern sources of consumer information.
Another immediate thought I had in mind when thinking of alternative way is to see if there’s data on number of entrance tickets sold of Eiffel tower and may be * 5
For those that are great at these sizing exercises, is this something that takes a long time to learn or is it something that you feel was already an extension of how you thought about problems before?
It’s kind of an isolated skill set tbh. There are times where you have to make estimates on the fly but you always have the luxury of the internet. The good news is that it is easily practiced at almost anytime, anywhere. Wherever you are, just think of these problems in your head and how you would break them down and solve them. For example when commuting to work I would think about how many cars are on the freeway right now, or how many food trucks are there right now, etc.