Interview Question

Associate Product Manager Interview



Google want to change youtube from having one image for videos to having several to better indicate what the video is about(5 images was used). How would you achieve this

AnswerAdd Tags

Interview Answers

4 Answers


Talked about extracting images as specific intervals based on the lenght of the video and the number of images needed. Also mentioned video analytics and how major changes in frames could signify a scene change and take an image from that point

Anonymous on


Split the video in 5 segments that overlap, but barely. In the first segment, find the frame that differs the most from the previous one and save it as a thumbnail. Do the same for every segment, just make sure that in successive segments not to pick a frame that might have been already picked (i.e. from the overlap area).

Andi Mullaraj on


As a PM you don't worry about the nitty gritties of the code, like they are talking about above. I took a PM approach to solving this issue. Devs are the ones that will come up with the specs such as AJAX blah blah back end. A PM looks at it from a high level: First, you need to make sure that the feature is worth releasing. How does it affect current customers (advertisers) and users. Do tests so that you have data driven results. Here are some tests that are run before a feature is released: 1) A/B testing - measure clicks on ads, share, exits, e.t.c. Since they want to make a video relevant with more images, in this case exits will be more relevants metric to measure. You can check if advertisers are now getting more ROI. 2) Analytic systems to measure how users interact with the new feature 3) Logging sytems, dogfooding, and even forums, and mentions. All the above gives you data to make an informed decision. Say you decide to go ahead, metrics have improved. You need to estimate impact Business impact - how does this affect ads, do you want to have ads have more images too? Engineering impact - you need to think about storage- in this case its marginal storage cost Now that you have estimated the below, how do you go about releasing it: Here are options: - Invite only -release to specific market say North America first. Iterate and find bugs and improve - Have a fail back strategy - how could you revert if this do not work out as possible,

Anonymous on


BACK-END (Server-side): Simpler than this. You divide the length of the video (minus 20%) between 5. This 20% is given as 10% buffer at the beginning and the end, as it would be a mistake to use as thumbnail pictures the beginning or the end because they can be black, credits or unrelated titles. These divisions represent the 5 thumbnails. They are processed during the video processing phase, and appear shortly after the video is uploaded. FRONT-END (Client side): Now let's explain the way the user would see this in the interface. My approach would be that when you go over the video thumbnail (main thumbnail, number 3) with the mouse, an unobtrusive popup (ideally coded in AJAX) pops up and displays 5 small thumbnails. This can also be done by activating an animated sequence of thumbnails in the placeholder for the thumbnail once you go over it with the mouse.

Antonio Bustamante Mirayo on

Add Answers or Comments

To comment on this, Sign In or Sign Up.