Skip to 0 minutes and 2 seconds In this video, we’re going to talk a little bit about product management in general and a lot about analytics around enhancing existing products. Joining us is Kiran Kadambi of eBay. Kiran, thanks for joining us. >> Alex, great to be here. >> Kiran has been here at eBay around five years. He’s a Senior Product Manager there and also a Darden alumnus, I can’t help but mention. So Kiran, at a place like eBay, I’m sure you’ve got a lot of strong, validated learnings, and there’s a lot of existing infrastructure that’s made the business successful. Can you talk a little bit about, a new PM starts at eBay, what would you recommend they do?
Skip to 0 minutes and 42 seconds How do they understand what’s made eBay work and how to think about the product in the system? >> Great question, Alex. The first thing that I’d Recommend a new hire at eBay or any other company for a matter of fact, is to go use the product in one form or the other. Or if you’re in a company where the end uses are enterprises, then go talk to your end users. And that applies to consumer or business. Use the product, get your hands dirty, go through the inner workings of the product, find issues with it. Talk to people about why those issues exist. Is it because it’s a bug in your product?
Skip to 1 minute and 26 seconds Or is it because it’s a government mandated reason that causes it to be there? Is to get the nuts and bolts and until you do that, you really can’t be a good product manager until you use your own product. And second is, what I recommend to somebody who joins new is to go and observe how the whole system interacts. So, the first point that I would advise was to look at it from the outside.
Skip to 1 minute and 58 seconds And the second point, is to look at it from the inside, where go see how the things work. Try to get an understanding of the whole flow. The end to end process. What happens in the good scenario? What happens in the bad scenario? So that’ll be my- >> That’s something you would do through actually just using the product or talking to engineers or technical project managers about the infrastructure itself. >> Correct, so you want to what have the conservation in addition to using the product on your own. Try to read through previous diagrams, or previous documentation. And try to get an understanding of the end to end flow.
Skip to 2 minutes and 40 seconds And what I generally do is, I personally would let things down on a flow chart or create a log diagram at high level to help me understand the different catch points. And, given that you would probably do work which would involve a subset or the whole set of catch points, it’ll be great place to start. >> Yeah, yeah. And one thing we’ve been working on of course, is how you amplify the product market fit for an existing product by extending new features. Can you talk a little bit about how you approach that at your work in eBay? >> Sure, I will give you an example.
Skip to 3 minutes and 17 seconds As one of the project that I worked on the past is how do we improve the eBay market place? So, even the marketplace where there are tons of buyers and sellers who come online, who come on eBay, and who interconnect the transaction. So what happens is, is that sometimes over a period of time for various reasons you have that on site. This is inventory that nobody buys, and nobody even looks at it from such perspectives. So what it comes down to is, how do we look at this problem, and how do we can fix it. So in order to look at this problem, we first need to have a good definition of what the problem is.
Skip to 4 minutes and 5 seconds So that in this example, what does under performing mean? So that could mean different things to different people depending on your vantage point. So the first thing that I had to do was understand how do we define this specific problem. And that meant that, look at different aspects of it, and then decide on how we progress further. >> And how did you do that in this case? >> So, one of the things we did in this case was to look at the listings
Skip to 4 minutes and 35 seconds of the items on eBay, and see their edgings of how long they could be on site. And then, even as a company that’s really big on how we use data to drive a lot of our decisions and analysis. So what I did was look, at our, Our data around search conversion.
Skip to 4 minutes and 58 seconds As well as how many times this basically got shown and search.
Skip to 5 minutes and 5 seconds And then how much of those sightings doesn’t click through? And then [INAUDIBLE]. So we would put all this data and then try to identify what would be a good baseline. So how would we define ageing of a listing? So let’s say that hypothetically, if a listing is one month old, do we see what kind of [INAUDIBLE] do we usually see? And what kind of impressions do we usually see? And similarly extend that to three months, six months, and so on. So the idea is by the end of this, we’ve identified what is a clear cohort group of listings. What are their characteristics? And which we can then form a definition which you can then use to share across our organization.
Skip to 5 minutes and 51 seconds And this kind of helps drive various activities. For example, not just on, why should we even wait to? Can we do a better job at recommending changes? Because eBay is a marketplace and we have all this data. How do we look at other continuing or other items with similar intentions and varied impressions and see what’s different? How do we set. So it can feed a lot of those. So that’s how we approached this and looked at data. >> So you define the under performing listing, set a baseline and then you create some ideas or hypotheses about how you might improve them. >> Correct. >> And then what do you do from there?
Skip to 6 minutes and 33 seconds >> Like you mentioned, one of the things we do is hypothesize and come up with different proposals and solutions, and ideas. And see how they impact different categories or articles. And then what we do is, the next thing that every product manager should do, which is go test them. So, some of the things that I did in this particular use case or example was we ran some AV tests. Where for the test crew, we ended up not showing listings that were over a defined under performing listing threshold that we identified. And for the control group, we let the organic sources show up.
Skip to 7 minutes and 18 seconds What we found as a result of this was that, there was increased conversion for these sellers around for the test buyers in this example. Where items that had a better integration ratio perform better, and it goes without saying. And the thing to observe is that it also helped the other sellers who they have not shown up, otherwise. So let’s say in the control group, as a buyer you used to see items from let’s say 15 sellers on the first page of your search results.
Skip to 7 minutes and 55 seconds For the test group, three of those sellers were not shown because they had stale listings. So that helped in conversion for the whole group of sellers. But also more importantly it helped our buyer experience. So you would be more inclined to come back to eBay, because I got what I want. But then, in my first search I didn’t have to wait through lots of bad listings.
Skip to 8 minutes and 22 seconds >> Yeah, yeah. And you make your conclusion on this test by the performance of these AB results. And that allows you to run a nice disciplined program, where you’re closing the loop on your ideas. Is that accurate and reasonable to say? >> Yes it is. So based on the test result, what we got was we got bankable data. To say that okay, if we go with this proposal and apply the definition of under performing listings at [INAUDIBLE]. And this kind of inflation ratio, then that makes sense. That helps us go back and say that, okay, how do we institutionalize this?
Skip to 9 minutes and 0 seconds So what that means is, how do we provide capabilities across the value chain on eBay, where sellers can focus on this particular value, in terms of how their listings are inflation ratio. And then, optimize their business to make sure that they reach a threshold value, which is category dependent and which would help them convert better.
Skip to 9 minutes and 27 seconds Go ahead. >> I was just making sure I understood. So you solved this problem of getting the right listings in front of the buyers. And then, you took those learnings to the other side of the market to help the sellers who were having those deal listings or just. To generally, show them what you’ve learned in a form that’s actionable for them? >> Correct, so at the end of the day, you want a more efficient marketplace where you want your current ratios to go up. So how would you do that is, if you have data which identifies what matters to the buyers. And if you can feed them back to our sellers to make those specific improvements.
Skip to 10 minutes and 4 seconds And then, what happens is that the seller can focus on building their business, which is to focus on sourcing the right inventory at the right point. And then selling on eBay for the right price to our buyers. And so that the buyers can get what they want on eBay, which makes them come back to eBay and that cycle continues. So that’s what we ended up achieving with this. >> That’s great, fascinating. So for the product manager who wants to do better, to learn how to take this analytics driven approach to product improvement, what are your top three tips?
Skip to 10 minutes and 42 seconds >> So somebody, a product manager who wants to succeed, especially in today’s world were the key differentiator, in my opinion, would be how your company utilizes the data. But everybody collects data. The question comes under how you get insights from that data and make sense of it? So it just comes down to asking the right questions. So I think that would be one kind of the top skill that would come to mind in a data driven world. So asking the right questions and knowing what to make sense or knowing. How do you make sense of all the data that your organization has collected. So there will be something that I would say would sit high on that list.
Skip to 11 minutes and 25 seconds And the second one is, don’t be afraid of data and the dish and that that it takes you. So sometimes, you find that you’re really passionate about a product, that’s definitely has happened to me. Where you pursue, go down a line, and you realize that the data shows, otherwise.
Skip to 11 minutes and 46 seconds >> Yeah. >> Right, so, that I believe is something that you shouldn’t be afraid of. And sooner you do it, the better because you can change sooner and come back with better learnings. So that would be something that is probably the second tip I would give to somebody who wants to come into this role. And the third one is, it’s to primarily make this part of your daily routine.
Skip to 12 minutes and 18 seconds Make it a more hands on experience. I know that a lot of schools are focusing on taking SQL skills. Go ahead and take them. Because what I’ve found is, that similar to the case method at Darden, the more you get your hands on, the more experience you get. The better you get at identifying patterns, the better you get at asking the right questions. And [INAUDIBLE] some good feedback [INAUDIBLE] yourself. >> So make a little extra time to acquire those skills on the job for yourself. >> Definitely. >> That’s great. Well, Kiran, thank you so much for joining us and for these insights about how to run your program with analytics. >> Thanks Alex, thanks for having me.
Skip to 13 minutes and 1 second It’s always fun to come and talk to you.
Kiran Kadambi on structuring problems in product management
In this video, Alex discusses applying data science to improve an existing product. One advantage of an existing product over a new product is the large amount of data that has already been gathered for an existing product. The data can be categorized in four ways: descriptive, diagnostic, predictive, and prescriptive. Think about a product you have worked with and categorize the known data under these four descriptors.
© Copyright Rector and Visitors of the University of Virginia