Joining us is Andrew Kreitzer. Andrew, thanks for joining us. >> Thanks for having me, Alex. >> Andrew is a Darden alumnus, and he was most recently a product manager on the economic graph team at LinkedIn. And he is now the lead product manager at BookMD. Do I have that right? >> That's right, Alex. >> Data science has been really central to your various roles, I know. Can you talk a little bit about what is it like to practice data science, and work with data scientists, as a product manager? >> Definitely, so I've been on both sides of the fence. I started out as working on data science problems for product managers at LinkedIn, developing some core recommendation technology around skills, and then ultimately seeing it go into product. And at BookMD, and previously at LinkedIn, I worked with data scientists on how do we improve existing models, for example, the way jobs are recommended at LinkedIn. That would be something that would be a typical role within product. >> One of the things we talk about a lot here in the course is how to bring in strong inputs to a development process and how to create focus around those. Having done both, can you talk a little bit about the similarities and the differences between the way you bring inputs and the way you the way you drive collaboration on a data science problem, as opposed to software development problem? >> Yeah, so fundamentally the approach to product management remains the same. You are identifying a problem, a potential set of solutions, prototyping those solutions and ultimately with wireframes and sketches, and ultimately seeing whether those solutions solve the problem for users. That what we're trying to do. So in the case of data science and data problems, the difference is that there might be some sort of statistical modeling that's the basis for the prototype. Or you might be prototyping what you think you can solve with data science by coming up with some hacky solution. For example, you might be putting just the email together with canned, curated recommendations, instead of doing a data science algorithm. >> Got it, maybe if you could step us through an example of how this might play out. >> Absolutely, so by taking Coursera courses before, you have a great group of people taking your Coursera courses. So you might have the case where a group of students want to be recommended other courses that will help them in their career objective. So maybe when they filled out their Coursea profile, they said I'm an aspiring product manager, I'm a current product manager, and in order to be successful at my job, I want to learn more about courses that help me do that. So, a product manager would write a user story that says, as an aspiring product manager or as an individual with occupation X, I want to be recommended courses that make me more successful at occupation X. So what you might do is ask an expert, ask you, which other courses would you recommend for a project manager? And then hand curate a list to start. And see whether you get a better performance recommending curated lists like that over having nothing, just a random set of the most popular courses on Coursera, for example. And then once you prototype that and get feedback that it's valuable, you could then apply a data science algorithm to the entire course catalog at Coursera and say, for any given course, what should we recommend to have you be successful at any job? Whether that's a product manager, a data scientist, an operations manager, a sales manager, whatever it might be. >> So we're still essentially trying to find something that's valuable to customer before we invest the effort to build something. >> Exactly, and there's a lot of prototyping tools like R that are the balsamic of data science. So instead of leanly prototyping a wireframe with users, you might rely upon a very basic naive model to do a recommendation, or to do a prediction, and present that to a set of users. >> Got it and one of the things we like to do when we close is talk about kind of a top three list. What are your top three recommendations for the product manager who wants to prepare themselves to be a good collaborator with data scientists? >> Yeah, so three tips for product managers who want to work better with their data science counterparts. The first is, get informed, so read up on blogs about approaches to data science. And dissect some of your favorite products and understand which parts are influenced by data science, whether that's amazon.com's home page or the way the Uber app works. Really dive in and understand it, talk to a data scientist about how they're actually solving a user pain point with data science. The second tip is to align around a metric. So as a product manager, pick a metric that you want to improve. Whether that's conversion or revenue or any sort of reducing churn. And approach that metric, approach the data scientists with that metric in mind, so you can both align from an organizational standpoint, around improving a metric. And the actual method can be up to the data scientist to provide to you. The third tip is be open to bringing data scientists into the solution set early, so once you identify a user need, bring in a team to brainstorm with you so that you don't miss out on potential solutions that have good data science applications. >> Well this is some great perspective from the front lines of this exciting discipline. Thank you, Andrew for joining us. >> Alex, it's a pleasure to be here, thank you so much.
