I'm here today with Rakesh a PM at Microsoft to do a estimation mock interview Rakesh do you mind introducing yourself to those watching yes thanks Kevin it's great to be here hey everyone I'm rkes I'm currently a PM at Microsoft I work on the Apple ecosystem mostly uh previously I've been at companies such as Google as a TPM and SpaceX and a few other YC companies but I'm looking forward to today's interview so let's get into it the question I love to ask you today is estimate the internet bandwidth needed for a thousand graduate students
the question I have is like there are many types of graduate students right there are people within the medical field there people within computer science the people within liberal arts and many other fields and the question itself is going to be dependent on what type of students we choose so for the scap scope of this question at least would you say that it's fair to assume or choose Liberal Arts students as the category that we want to pursue can you tell me more behind that rationale yes um this is just anecdotal um most anecdotal kind
of thing but I think when I was at undergrad University um a majority of the students at the University of Michigan were labor Arts Majors I think it encompassed at least 40 plus percent where you had all the other Majors combined into more smaller kind of Aggregates so the reason why I chose that was every you go you see a level Arts major it's mostly because the courses are so vast and popular that many students love the program so just wanted to choose that as a good kind of majority point at least sure sure and
I think you might be also alluding to the level of usage of one's computer you know when you're more inside of a classroom setting for the majority of your time and also doing a lot of computer related tasks such as writing Etc versus I think you mentioned earlier in the medical field you might be actually interacting with patients is that also a factor that is also a factor I think you need a is spot on like for medical students you may have an issue where people or the students themselves may not use their mobile much
actively but maybe rely on iPads for writing down notes or kind of scribbling down a few Concepts but where in computer science what I've observed is that mostly students would rely on their laptop or desktop to code and kind of test out and problem solve and uh in other majers you may have neither you may just have more mobile usage it really depends on kind of student Persona there but I would say Leb Arts is a good average of all these scenarios so I think that's a good starting point now with that being said I
think the thing I would want to kind of categorize is kind of the platform so we have I think I kind of mentioned these but we have desktop we have mobile um we also have other kind of platforms like iPad but for the sake of this question at least I would like to consider desktop and mobile since they're most commonly used and uh given the scenario that not many people have iPads not many of these students may have them so that'd be fair sure yeah I think it's okay to standardize amongst desktop and mobile and
we don't have to consider F fancy things like iot devices or anything like that so desktop and mobile it is perfect with that being said um I think the next thing I would like to do and start off and break down is the day-to-day schedule of the average graduate student that'll help us kind of understand how many hours do they actually put into these uh put into online usage and how many hours they spend on offline usage that'd be fine yes so let's assume as an average human being student or a non-student the person sleeps
eight hours per day so that gives us around 16 hours of active awake time let's say in these 16 hours that the person sleeps uh sorry the person spends some time eating breakfast getting ready having lunch having dinner so let's take an hour for each maybe three hours so almost as a leite both with 13 hours um for the sake of kind of recreation and hobbies let's kind of remove another three hours since people may want to go to the gym or kind of do other activities maybe have play some sport if not do some
other activity which is individualistic to that person so I'm going to remove another three hours from that that gives us a total of 10 hours for us to work with so I think the interesting scenario here is that uh classes and how much time do classes take per day um so I'm going to assume for the sake of this exercise we're working we can do either the weekday or the weekend because they both have different structures but for the sake of this exercise I would propose we would do their normal school day would that be
fine with you sure yeah we can always um talk about that at the end where potentially we can make different estimates and stuff like that perfect yep for the of this question moving ahead with the school day let's assume Maybe I know that some students take a higher course load per semester so they may have maybe six hours eight hours of classes I remember I had semesters where I would spend a majority of my time in classes but I also have semesters where my friends and I used to spend maybe not that much time maybe
four hours in classes so why don't I take the middle ground here and say six hours of class work per day which I'm actively in class that'd be fine as assumption so if classes are about 1 hour long or maybe they're different in your experience in graduate school you're assuming that around a graduate student is taking about six classes per day um I would not assume six classes per day because um the typical at least to my anecdotal experience classes were really long when you take a four semester four credit course um they typically range
between three-hour seminars where we have series of like mini group discussions all these other activities but they usually take 3 hours I think 1 hour is too small for for uh credit course but um let's assume each one is about three hours so that's like two courses per day different kinds rotating depending on the DAT would would that be fair sure sure okay that gives us six hours of class time and four hours of let's say some study time because we already to have taken out Hobbies we all we already taken out kind of day-to-day
uh usage like uh eating socialization Etc so let's say 4 hours of active study time to prepare and get ready for these classes would you say that's fair that's fair okay okay now comes the interesting part how do we categorize these in terms of desktop versus mobile so the thing with classes is that most professors at least most likely to my anecdotal experience would not want people actively using their mobile devices in class unless they have really good reason to um and we have an UCare scenario where people kind of still use their mobile phone
kind of actively and just put it on the side or do whatever they want but they're not actively kind of fully engaged on it so I'm going to remove that Ed scenario and consider that people have their laptops open at least for a majority of these six hour sessions because they're taking notes they're referencing something most people have their laptops open but may not actively use it so would you say a good active usage here would be four hours Within These six hours of classes and two hours of just nonactive passive usage where the active
laptop is in standby would you say that's a good assumption yeah I think that's reasonable and feel free by the way to just assume things and I will jump in if I have questions around where that rationale is coming from perfect now with that freedom I'm going to assume that four hours of Active Desktop usage out of those six hours and two hours of standby and the remaining four hours I would um want to assume since it's preparing and studying uh I would say it's a mix of both mobile and desktop usage so I would
say two more hours of mobile usage per day and I would say 2 hours of laptop usage so that gives us a total of 6 hours per day of active laptop or desktop usage and two hours of mobile usage I would be curious about the mobile usage right anecdotally speaking just looking at screen time numbers they're far above 2 hours probably for most folks how do you square that and how would you consider that or yeah would you push back I'm I'm curious where that two hours um is how you're justifying to two hours here
that's a good question I would say on the lower Spectrum uh a typical student could have a SC ER where they they're not actively checking their phone and sitting on their phone they could just use it in between bus rides use it between kind of coffee braks for a few minutes and that all adds up to two hours of a time sometimes they have a quick thing they want to check on their phones or kind of a video they want to watch for a quick relaxing break so I like I I see where your question
is coming from and I do agree on the point so why don't I do this why don't I keep the two hours as active meaning that people use it on active basis but remember the 3 hours I kind of subtracted at the beginning which was mostly hobbies and Recreation I did also want to consider since you mentioned it that sometimes people rely on their mobile devices and depend on many social network platforms as recreational activities but not everyone so why don't I kind of push that a bit more since we're definitely seeing more screen time
to saying let's add another two hours of those three hours so that brings up the to Mobile hours per day to four and I think that's a good total screen time per day not considering the too too many extreme cases for the typical grad student do you say that's fair okay great I think that sounds much more fair awesome so now that gives us a total of 10 hours per day on active internet usage for both devices combined um with that being said I think the next step is for us to break down the activities
associated with each device and we can clumb them together but I want to kind of lay out my kind of perspectives on these activities would you agree perfect so out of the six desktop hours what I've kind of came up with and assumed is let's say the student uses three hours of the desktop or the laptop time to kind of watch videos regarding the lecture while inside kind of class or maybe not in class maybe spend some time time just watching informational videos which are related to class that is and I would say the remaining
3 hours we would be spent on taking notes kind of writing up a few papers drafting up a few ideas and other miscellaneous activities checking your email Etc would you say those are fair assumptions okay and for the mobile side I think this is where it gets interesting I would say two hours of miscellaneous activities this could be something like texting um maybe you're responding to a few friends Etc and two hours of active social media feed which is pretty maybe content heavy this could be any anywhere from watching a YouTube on a video on
YouTube which is played in high quality or something even as watching a few Instagram reals or Tik Tok reals I'm not sure if Tik Tok is still available but um something of that kind would you say that's fair that's fair now I think the next step is to kind of make an assumption of how many what's the bandwidth of these individual activities and kind of use that as a data point for us to get started I would say for the desktop scenario just 3 hours of browsing and maybe surfing the web could take anywhere from
one mbbs per kind of per kind of uh would say the interval of time and this time let's say is kind of one minute so per minute I would say 1 Mbps for kind of miscellaneous activities like drafting up a paper Google Docs slides or whatever kind of or operational tools you're using maybe another one Mbps for the sake of Simplicity and for I would say the mobile scenario I've estimated the high quality videos maybe something like a high quality video on YouTube or other social media platforms could be five Mbps per minute and the
other thing about um social media which is kind of um like you see high quality re sometimes you don't you scroll down and watch a few high quality pictures I want to take maybe not five but somewhere around three Mbps would you say those are fair assumptions okay now with that being said let's do a few some math so I'm going to walk you through it so I would say for the first person of the desktop which is which is 3 hours so that's 3 hours * 1 Mbps that's equal to 3gb for browsing as
well we have another 3gb for mobile considering random social media videos pictures reals we have three hours of that sorry we have two hours of that so that's two times two hours times 3 mbbs that's equal to 6 GB per day and we have high quality video so that's 2 hours of high quality videotimes 5 MBS that's equal to 10 gigabytes per day so how much does that give us if we sum all these together so we have three + three that is six total for desktop and we have 6 + 10 which is 16
total for mobile that's roughly equal to 22 GB per day per student on average in this scenario and I think the first and primary top most question was how about okay you got it for one student but what is it for a thousand students so that's 22 gigabytes times 1,00 per day which is equal to roughly 22 terabytes per day I think this metric gives us a good average considering the assumptions we made and the methodology we use used but I do want to raise um some concerns and a few risks here but before I
do so do you have any follow-up questions to that did the math make sense um open for your feedback yeah I think the math made a lot of sense and great that you recalled the a thousand student prompt I am curious where you think and maybe this leads to the next portion you're about to cover but where do you think that some of these um where do you think that some of this estimation could have gone wrong or could have been improved I think that's a good question I think this goes back to the sanity
check so I think the first and most important thing are the assumptions that we take in mind uh I think the the scope of considering legal Arts students as the average kind of graduate level student is kind of maybe not the best way of estimating this question I would say if I had more time and if I were an actual pm investigating this issue I would take a random average of students so I would take 100 medical students I would take 100 L art students I would take a 100 maybe chemistry or psychology students and
so on so I get a good distribution of these random people and I would actually dive deep into some of these personas and maybe have surveys to understand their typical behavior and maybe average all that together to and remove kind of the standard deviation so we have a a better understanding of these different students in real sense that gives us a more accurate representation of a good student base so that's something which I would definitely improve if I had more time and the second thing which really concerns me here in this estimation is the Assumption
of how people use their time I've seen students I know in this example we took 6 hours of desktop and 4 hours of mobile I've seen scenarios where people may not really depend on desktop maybe not spend more than two three hours per day but I've seen scenarios where people would spend a lot of time on mobile and the same other way around I was seen people maybe a group of students in a particular field refrain from using mobile too much and spend a majority of their time on desktop take computer sign students for example
a lot of their time spent on coding and problem solving preparing for interview use um testing out their code Etc so you don't see many of these people kind of interact with their mobile device too much as much as compared to the desktop so just that one example can give you a lot of info on the type of variations of how people use their mobile desktop so I would say once again if I had more time I would kind of have a better representation and uh kind of use that to classify how to measure these
uh thousand graduate students and the final thing if those two made sense I'll stop there before you have any questions before I talk about the third thing I would improve no please go ahead go ahead yep the third thing I would improve is the actual measure of data so I think at the beginning of the conversation I mentioned how we were assuming this is in the United States I do understand that depending on the location of the University the graduate program students have different behaviors they have different sleep skill they have different Customs to what
they're accommodating to um and even the minor factors like weather could affect a lot about how students spend their day and even how much sunlight is coming in in to the day how much how much like nighttime is during the day there's so many um dependent variables here to consider so I would consider all these into cons consideration when making these uh large claims and would actually get a accurate data point on how much bandwidth on average is taken per activity in much more critical detail I know we made assumptions about one mbbs on for
active surfing I think it may be more um I made that basic assumption but I would get those numbers really accurate so we can have a really good concrete mathematical uh and u datadriven uh way to estimate this so I would definitely improve on these three to four points if I had more time thanks for KES so why don't we move on to Let's consider the mock closed let's move on to the feedback portion do you have any quick feedback besides the stuff that you mentioned earlier for yourself and then I'll dive into the um
what an interview interviewer rubric feedback might look like for you I think I would have spent some more time on asking a few more clarifying questions I think I have done a decent job but would have focused a bit more on kind of getting those really crisp and clear and maybe I would have spent a bit more time on kind of laying out my methodology of selecting and choosing a few assumption based numbers on how people spend their day I know that for the sake of this exercise and for the sake of kind of Simplicity
of it me wanting to get a more of a average figure in the way uh I chose a few app assum numbers so definitely spend more time thinking about my thought processes doing so um I think I would also consider a few more edge cases so I know that the numbers here are really simple and straightforward but what I would do is consider edge cases whereas what about if students or these graduate level students are in a time of study or in a maybe end of semester exams or maybe something like summer brick that greatly
affects um a lot of these estimations and I would consider okay perfect yeah so I actually did thought you did a really great job especially on the clarification this question is deceptively simple and so I think you hit on some of the major points which I'll get to in a second things that I thought you did well Rakesh so first of all great job scoping down the U problem and making it a lot more manageable right and you did that in the clarification section you talked about Liberal Arts students I might have just kept it
a little bit more General right like something about um students who spend a lot of time on their laptops or in the office as opposed to out of the office I.E lawyers doctors etc etc so that might have been a little bit too Niche but overall still fine um the devices that one uses and the scheduling was all super clear as well so I appreciated that number two I thought you did a really great job rolling with the punches for when the you when the interviewer pushes back and when you get to more advanced level
interviews that is going to happen much more frequently and I thought you did so with a really positive attitude and was able to come back with a coherent answer and then number three I thought you did a great job in this section where you called out where the estimation could have gone wrong and that's definitely a more advanced um estimation tactic right where you are proactively telling the interviewer hey you know this isn't perfect if I had XYZ data this is what I would be looking into obviously no one's going to know a lot of
this off the top of their head and I want to call out that most people are not going to know bandwidth and megabits per second and what is a sort of reasonable MVPs so that could have been something um someone clarifies at the beginning of an interview as well no one is expecting you to memorize these numbers okay so that those are on the positives on the things that you could have improved so this is a more minor presentation point where and this is something a lot of people struggle with but you have a bit
of upspeak which means that your tone goes up at the end of your sentences so for example you would say mobile usage desktop usage in general when you lift the tone at the end of the sentence it sounds like a question and it's not as confident sounding so I'd practice Downs speake more again a lot of people face that issue I do that sometimes too so don't feel don't feel like it's um something that you know is something you can't overcome right everyone does this number two is try to be more proactive with things in
general right so whether it's giving I think I was the one who prompted you to say hey where could these things go wrong or I think in general it's like hey like I'll jump in if I have any um concerns and then you still consistently ask does that sound good with you which is a good technique to check in with the interviewer but maybe you can do so by saying hey I'm going to move on to the next section if that's all right with you or just tell me what you're going to do next instead
of just stopping and saying hey does that sound good with you right so it's already great that you're checking in that's definitely necessary but do so in a more proactive manner either by calling out some risks and considerations and or just moving the interview along the reason that's important is as a PM you want to be seen as a leader you want to be seen as someone who can rally the crowd and rally the conversation so that would help with um an interviewer being more confident in your skill sets for this role and then the
final thing I would say is this question can be very um convoluted and the interviewer can lose the structure very quickly and so I would recommend some sort of visual asset to go along with this type of question of course we're doing this virtually in person there's white boards there's paper stuff like that but to the extent that you can either screen share and or you know lift up your piece of paper that you're working on and show me the mathematical equation you're drawing out that would help me follow along a little bit better because
I will admit once it was the middle section and we're making all these assumptions etc etc it's very easy to get lost but that said overall Rices I thought you did a fantastic job and I really appreciate you coming in to chat with us today any last thoughts you want to share
