This episode features an interview with Jaya Kolhatkar, Chief Data Officer at Hulu. Jaya previously served as Senior Vice President at Data and Analytics Platform at Walmart, and has more than 30 years of career experience with companies like Amazon, eBay, and PayPal. On this episode, Jaya talks about using data to improve and grow Hulu, the future of analytics and cloud data management, and much more. So please enjoy this interview between Jaya Kolhatkar, Chief Data Officer at Hulu and your host, Steve Hamm.
This episode features an interview with Jaya Kolhatkar, Chief Data Officer at Hulu. Jaya previously served as Senior Vice President at Data and Analytics Platform at Walmart and has more than 30 years of career experience with companies like Amazon, eBay, and PayPal.
In this episode, Jaya talks about using data to improve and grow Hulu, the future of analytics and cloud data management, and much more. So please enjoy this interview between Jaya Kolhatkar, Chief Data Officer at Hulu, and your host, Steve Hamm.
--------
How you approach data will define what’s possible for your organization. Data engineers, data scientists, application developers, and a host of other data professionals who depend on the Snowflake Data Cloud continue to thrive thanks to a decade of technology breakthroughs. But that journey is only the beginning.
Attend Snowflake Summit 2023 in Las Vegas June 26-29 to learn how to access, build, and monetize data, tools, models, and applications in ways that were previously unimaginable. Enable seamless alignment and collaboration across these crucial functions in the Data Cloud to transform nearly every aspect of your organization.
Learn more and register at www.snowflake.com/summit
Steve: [00:00:00] Jaya is great to see you today. I wanted to start with a little bit of history. So people are set on that. Hulu started by streaming a portfolio of TV series from its media owners, which at the time in the early days included Fox. NBC universal and Disney.
Now it's controlled by Disney and it's expanded to streaming original programming and the live TV service, which lets subscribers watch over 50 channels via the internet. If you would kind of describe the business strategy for the company these days, and also what's the
Jaya: [00:00:37] Okay. So, um, who's not star has always been to provide our viewers with choice and control over the watch experience. Uh, we have the largest TV streaming library. We are available on millions of streaming devices and because we have life. And, uh, streaming, uh, video on demand. We have multiple plans and price points, and we also offer premium, um, services.
So it leads to. Great number of permutations of our products. So there is the ability for our customers to really customize how they watch Hulu. Um, we have hundreds of content partners, uh, who send us, uh, content, which we then collectively show our customers. Um, as we mentioned, we have live TV. Both from local streams, as well as national streams.
And we also have a very large advertising business and be served thousands of different versions of creative ads every day. So choice and control is what our business strategy is to make sure that TV is. Anything you want it to be on Hulu? Uh, so that's our business strategy. And for that business strategy to really work well, it means that we really need to leverage the.
Billions of data in events that inform us about what we know about our subscriber subscribers and how they actually engage with our applications. It allows us to leverage that data, to create the right customer experience for our customers. Uh, it allows us to. Understand what they need. So this could span from understanding and suggesting plants that would be best for our customers.
Um, it means, you know, showing them the content that they. Might want to watch that we think they will want to watch, uh, uh, to the ads that are most relevant to them all the way to just making sure from a customer, uh, selfless perspective, you know, our customer support team, leveraging data to understand what sorts of questions our customers will have for them, and to be able to give them the best possible answers.
So that's generally our data strategy is to allow our knowledge of the data
which is to
Steve: [00:03:26] that's great
you mentioned before that the company has a big advertising business. Could you describe how that business works and, and how you support it with data analytics?
Jaya: [00:03:39] So, um, Yeah. You know, it is, um, a major part of our, well, it's a large part of our, um, revenue and, uh, we actually have several products which, um, which the customers can choose. Some of them will choose the new ads product, but by and large people tend to have our ad supported product as their choice. Um, You know, some of the things that we do is we do things like understanding what customers, uh, would prefer what types of advertising.
So trying to figure out relevancy of, um, of ads for, uh, each subscriber. That's one area that we look at.
Steve: [00:04:20] from the way you described it, I wondered. Does each individual customer Hulu see a different set of advertisements?
Jaya: [00:04:29] They do see different set of advertising. Um, it depends upon the geography. It depends upon the demographic. Um, it depends upon the content that they watch, so they will see different ads.
Steve: [00:04:42] That is amazing. I mean, you know, now that you say it it's obvious, but is that it seems like that's a dream that the adage, it seems like that's a dream that the advertising business has had for, you know, decades. And now it's finally come true.
Jaya: [00:05:00] Yup. Uh, I think that, you know, it's, it's definitely something that if you didn't have the ability to look at, um, customers with a very personalized view, you would not be able to do that. But, you know, we have the ability to do that. that we build an offer our brands.
Steve: [00:05:22] Yeah, that is really fascinating. Thank you for that.
Do you, actually ask them what they're interested in
Jaya: [00:05:29] No, we don't ask them. We kind of, uh, derive it from some of the content that they watch. Also what, you know, we have segmentation and demographics that allow us to think about what products would be the most suitable for them. That is very similar to what most advertisers do. But I think the. Looking at content, uh, to understand what a subscriber might need, um, is, is something that is unique to Hulu.
And given that we have so many years of experience, we've kind of been able to build algorithms around that.
Steve: [00:06:08] Oh, interesting. So you mentioned a couple of interesting projects you've worked on. What's the most recent, interesting thing that you've launched your, your most recent innovation with data?
Jaya: [00:06:21] So some of the things, um, I think like one of the areas that I'm most proud of and we're like spending a lot of time actually building that out even further is, you know, typically people use data for insights. You know, to, for marketing and for content. And so on, we've been doing a lot of work recently to see how we can leverage data insights, to build new unique products and continuing on, you know, the advertising area, um, a year or so ago.
Um, we had come up with customer insights with said, you know, 40% or more of our subscribers, uh, actually binge more than one series every month. And by binge, I mean, they watched three episodes or more of the fame series in one sitting or one session and interesting. Um, but our product team started thinking about, is there a way for us to use that, to create.
Innovative, um, ways of advertising. And so they created something called the bench ad. And what it actually does is it, we use algorithms to predict who's going to bench and then be offered the ability to our various brands to, uh, sponsor a binge ad free episode. While they're bingeing. So it's something that actually gives both our advertisers as well as our customers.
Um, You know, benefit and, uh, allows us to create a very unique way of advertising that these brands will not have anywhere else. Um, another example was like, we found that people paused while they were watching shows. And, you know, we also talked about leveraging that. Time, when they have pause to see if there was anything that we could do from an ads perspective.
So we launched what we call the pause ad, which is very non-intrusive, but it's very cute and funny. Um, and, uh, you know, gives, uh, uh, the brands await to have a different persona, if you will, uh, in the way they advertise. So both of those have been very successful. We've been talking a lot with our.
Application folks, you know, when we can predict whether someone is coming to, whether they're going to continue watching a show that they already are familiar with, or they're going to come in and discover a new show. And how do you leverage the fact that you can predict these things where.
Understanding data patterns. How can you leverage this? How can you create new products for that? So those are kinds of things that, you know, we've been spending a lot of time thinking through, building out. And as you can imagine, the data load is immense and weaving through and teasing out patterns that can allow us to build these products.
Sometimes it's very challenging, but that's, that's fun.
Steve: [00:09:33] No, that sounds like great stuff. It just occurred to me that, you know, it's, you know, much of advertising on television is storytelling. There are characters there's storylines with bingeing, you could actually have maybe a series of ads that someone so that it kind of develops the characters develops the themes.
Has anybody talked about that or is anybody doing anything like that?
Jaya: [00:09:56] I'm sure they've talked about it, but I'll be sure to mention this in the ads product meeting and claim it as my own, um, Uh, but I think we do have a storytelling aspect within Arbinger today. Um, and you know, Sometimes you could have sequential ads as well , because we know it's three episodes.
You could, you could, um, you know, we build out an ad which takes you through, um, minimal ad load to start off and then add free loads at the end. So, you know, uh, those ad product team. it was a pretty smart bunch of people and they've leveraged our data insights
Steve: [00:10:38] Yeah. You know, it really is interesting. This is something that my wife and I do. We, we get onto a, uh, uh, series and we'll watch two or three a night, you know, and kind of like, we'll go through a season like that pretty quickly. And you are, I mean, I find that I'm practically stuck to my seat.
Especially, I mean, with a really effective shows where you just can't, it's a page Turner, essentially, you want to get, you want to get to the next show, it really is, uh, uh, a new way of consuming entertainment. And it seems like, it seems like the ad industry and your, your company are really kind of rising to the occasion with figuring out what to do with that.
And I, I think there's probably a lot of interesting new things coming down the pike as well.
Jaya: [00:11:18] And very frankly speaking, like what we are thinking about is not just monetizing it, but making it a good experience for our customers, right? Like not having ads adored in an episode is it's really. And great customer benefit in, and you know, our brand partners benefit from it as well. But the drive is usually towards figuring out how we make the experience better.
Steve: [00:11:48] Yeah. Yeah. Yeah. Gotcha. Gotcha. Now, um, Hulu is a still snowflake customer, and I'd like to understand how you're using the cloud data platform.
Jaya: [00:11:58] Absolutely. So, um, um, when I joined two years ago, um, we had just begun testing, um, snowflake, um, one of the first things that we as a team did was to bring together all of our subscriber data and we call it subscriber three 60, um, into one place. One of the reasons for doing that was.
Over the years, different people had different definitions of what a subscriber meant. People who ran the ads business only wanted to pink, that subscribers who had the ad product for truly subscribers. They didn't really care if any, about the others. So you had a lot of these pockets of, um, folks that were using different definitions.
So we wanted to bring all of this data into one place. And then allow people, the ability to analyze this data, derive insights from them really easily. And so that was the first project that we used to flake for. Um, we had about 120 uh, model data elements, which came from multiple sources that we combined into one snowflake database.
Um, and because of subscriber three 60. We were not only able to make our data more actionable. Um, we made it more accurate, um, and we made it more accessible. Um, it is just a much easier way to access data than, you know, our traditional hive queries. And it allowed the business to really. Understand how to leverage the data, be able to create actions out of the datas.
So that was one of the first areas that we actually started using snowflake for. And, um, That was the starting point. We've had multiple three 60 products that we've built on snowflake, and it's allowed the company as a whole to start thinking about the business holistically. So our marketers think about not just the subscribers, but they think about advertising and content and the product it's allowed us to do things holistically.
So that's one of the ways that we've been,
Steve: [00:14:15] So you combine all the data together and all of the functions and business elements of the business can all get into that. So now for the first time, they can actually combine their data with, with data from others that really gives them much more of a, like a. Three dimensional picture, or you said subscriber three 60.
It sounds like they know that sounds like a very smart thing to do. is there one particular, uh, kind of insight that you started getting that w that you could talk about a little bit, that, uh, we'll show us kind of the benefits of integrating the data in that way.
Jaya: [00:14:55] I'll give you an example, a very recent example and something that I paint, um, you know, it takes us up a slight notch. in November we launched, uh, we are part of Disney now, so I, um, You know, call them us as well. So we launched, um, Disney plus in the vendor. And with that launch, we also launched a product, uh, which bundled all three applications.
So Hulu Disney plus an ESPN plus, and we call that the bundle. Um, so we launched the bundle in November with the Disney plus launch and the hypothesis going in was that, you know, If one services good, then three services bundled at a much cheaper rate is even better. Right? So we went in with that thought I'm pretty blind, right?
We were running in three different silos. We were running our own thing. They just need plus was running their own thing. And so very quickly we realized that we really needed to understand how these. Bundled subscribers were behaving all three, um, applications. And how do you then understand that and what do you do about it?
Um, so we, we were able to leverage uh, snowflake product to actually come up with this data exchange where we could share data across the three platforms. Um, and it allows us to share it. This data without really worrying about, you know, shedding data, uh, and having it leave our walls, it's done anonymously, but it allows us to look at every single bundle subscribers engagement on Hulu versus Disney plus versus ESPN.
And, um, that has allowed us to understand. Hey is our hypothesis. Correct. And was it good for all three, a and B? Um, is there some subscribers, this benefits more than others? And if we knew that, how would we actually market to them? Um, and so what we've realized is, and what we found is that the bundle is.
Pretty much great for all three services. Um, a bundle of subscribers are stickier. Um, you know, they're more engaged children, less. Um, and what we've seen is in the groups of subscribers where one service is not fully meeting their needs. So let's say, um, E. Adult with no children on the Disney surface may or may not have, you know, all of their content needs met for that group of subscribers.
Having the bundle is really great because we do very well with. Males, Um, and so that is a lot of content that they are able to leverage. So in that case, you can see that one plus one is actually three in our case. , um, and then you can build in marketing strategies that allow you to target these particular customers with it.
Very. A targeted messaging, like making sure that they understand the kinds of content that they get when they sign up for the bundle. So, so, you know, like having these kinds of things that we can tie together, um, and not just within our own world, but within the larger Disney world gives us a lot of capabilities that we did not have before.
Steve: [00:18:34] Yeah. Yeah. Now you're combining the data you're putting together a bundle and you're combining the data from Hulu Disney plus an ESPN plus now. Before snowflake would that literally have been possible to do that kind of thing and that kind of
Jaya: [00:18:54] It would have been really hard. I mean, you know, I pink, um, We would probably have to use a third party. You know, I've done this in my previous lives. It has been, it has been pretty difficult. Um, this was much, much easier and more seamless than I've done in the past.
Steve: [00:19:16] Yeah. Yeah, that's great. Great. Now I want to ask you to put on your visionary hat and look into the future maybe five years or more. How do you see data changing the entertainment streaming business over these next few years?
Jaya: [00:19:33] You know, the streaming business is in a period of great change. Um, and I think one of the things that we'll determine who is successful is, uh, the streaming service that understands its customers really well and figures out how to. By build content that really resonates with them. Um, and for that, I think we're really going to need to have, um, the underlying infrastructure for them to be able to leverage, uh, this understanding of the data and of the content and how it interconnects.
Um, uh, across the product, a product and tech, and that's something that I paint, uh, over the next five years. Uh, you will see where you will see it. In the newspapers, you're going to see more about content, producers and content and things like that. the apparent fights will be over that.
But, uh, I think the battle will be won or lost by the, the streaming company that really understands the customer needs and, provide, uh, solutions for those needs.
Steve: [00:20:48] Right, right. That makes a lot of sense. Now in your career, you've worked in a wide array of industries. You've been at Walmart, Amazon, eBay in the banking industry and the insurance industry. So you've had an incredibly broad array of experiences as a PR as a data professional. How do you see data transforming the world of business more broadly?
Jaya: [00:21:12] uh, some of the industries I've been in have been, you know, users of data for a long, long time. The, you know, the banking industry has leveraged data very well for 50 plus years. Um, I think what we are seeing more and more is creating new businesses that are based on the fact that we can, uh, leverage vast amounts of data and create products out of it.
You know, like 15 ish years ago, the Google Facebook. Twitters that really would not be possible without their ability to crunch enormous amount of data and create a customer experiences or customer, uh, products that were, you know, useful to their customers. And now ubiquitous, right? Like I don't, I can't see myself.
Living without Google. Um, you see that in the case of, you know, a lot of these gig, uh, economy type companies, it wouldn't be possible for them to really create those businesses without data. So my part is that. You know, as technologies get better as our ability to consume larger amount of data gets better as our algorithms, um, our ability to do, um, to really look at data with a three 60 perspective, I think that will create new products and services, things that we may not have imagined before.
Right. I mean, there are many things that, um, Even I, as a data professional would not think were possible 10, 12, 15 years ago. So I think we'll see a lot of advancement. I hope I'm on the healthcare space. That's where I see a lot of folks talking about, uh, you know, how data can really transform, uh, an industry that has been pretty siloed has been pretty, um, uh, Pretty conservative about leveraging data because of all of the privacy concerns around data and so on.
But I think as we've gone through the pandemic, and as we trying to think about, you know, if we had data, would we be able to do contact tracing much more easily? How does one do that? Um, you know, how do you create those kinds of products? That's, that's an area that I think is pretty right.
um, for innovation.
Steve: [00:23:44] yeah, yeah. You know, I'm, I'm also working on a, I'm a freelance writer and I write in various fields, information technology and healthcare and medical science, especially. And I interviewed a woman today. Who's, uh, A professor at Yale and head of geriatrics at Yale new Haven hospital, which is right down the road from me.
And she was talking about something basically, you know, they have electronic medical records. Now the whole medical industry was based on this idea of the dot all these specialists, knowing how to take care of this disease and this organ in your body. And they never really. Looked at the patient holistically are there.
And they never looked at the patient as it essentially a consumer or a customer of theirs who might have their own opinion about things. So there is actually this big movement in healthcare right now, uh, around, you know, surprise, surprise asking the patient what they want. And how they want to live and, and coordinating, sharing data between the specialist and the internist and coordinating care in that way.
And it's just like, you know, it seems so obvious. And for someone like you coming up through these industries where the focus has been for so many years on, on the three 60 view, and it is coming to healthcare.
Jaya: [00:25:08] It does come into healthcare. I, uh, I have a patient at, um, Stanford and they've done it fairly interesting, um, set of work around. Trying to put all of your visits, all of your diagnosis, you know, , your prescriptions, things like that in one area, uh, where, you know, you have access to your records.
Uh, it's just easy when you go to a specialist for them to take a look at other things, um, that might be pertinent to. Whatever it is that you've gone to see them that, you know, if you're, uh, if your PCH hasn't written that up somewhere, they can go in and look at other records in the past to see if they can leverage that.
And that I think is that's kind of the way I think we will go out in the future. Uh, and I see that, you know, in developed nations, I think. You know, that's definitely something that we need to do, but I see that, um, in, you know, well, not really third world country, I don't call it India third world country, truly, but it is, uh, uh, to a certain extent that, um, you know, over there.
everything is so disorganized that there's a lot of folks that are trying to figure out how to connect all of these dots, especially for rural India, where I think there will be, you know, It would be one of the best things that they can do for, uh, for rural India to see how healthcare can be connected so that people that really don't have any access or don't have the education to, um, tell the doctors, uh, about what.
Uh, a previous doctor might have told them or trying to explain their, um, symptoms and things like that. Those are areas that I think are also, I'm very ripe for innovation.
Steve: [00:27:04] , yeah, I look forward. I mean, it seems like the world is in terms of technology, in terms of business, innovation is really in kind of a golden age. Now. Obviously we have lots of other problems surrounding us COVID and other stresses, but it, it does seem like technology is.
Really in a position to help deal with dresses, but also deal with some of these opportunities. I, you know, I didn't ask you about artificial intelligence. Is that something that you're really using aggressively at Hulu at this point?
Jaya: [00:27:38] Yeah. If it does scale, you know, we cannot really do things without machine learning. Right? Our recommendation engine, our search, a lot of our marketing, um, campaigns. All run off of, uh, machine learning. Uh, we do a lot of things and just understanding, uh, leveraging machine learning to understand, um, indicators too.
You know, some customer event like churn, for example, or reduction in engagement or things like that. So we use it fairly, um, fairly aggressively, I think. Um, you know, it's, it's just, again, A tool that you use, um, with, uh, with a lot of understanding about the customer and about our business. So we've been fairly judicious
about how we use machine learning.
Steve: [00:28:32] So it's not like a magic pill you use for
Jaya: [00:28:35] No, we don't. uh, in fact, like even a recommendations, which typically people, you know, hand over to the machine, we've been very, very, um, in some sense, conservative, but in some sense, we know that our editors have really deep knowledge , of our content and they have. Deep knowledge of, you know, the industry in general, they're able to leverage these side guys moments.
And so we have a combination of editorial and machine driven, uh, recommendations and personalization. So, um, we've been um, very thoughtful about not having black boxes,
um,
, that we are not able to interpret and understand. What's happening.
Steve: [00:29:19] I think that's very smart. And, you know, I think the terminology that's being used most commonly is augmented intelligence or AI, augmented intelligence to make it clear that you need these, these smart analysts, their insights are going to are going to be critical. And yeah, I see you can't just have this black box that kind of turns it into a commodity or just turns it into a machine.
You also have to be also have to double check the machine to make sure that the, the, the recommendations or insights that are coming out of it really are sound
Jaya: [00:29:54] Yup. This sound and you know, like you've got to, you know, you've got to understand how you train the algorithm. You have to understand, you know, are you, are you. So precise that it is, you know, it's not the best thing to do as well. So how do you add a little bit of generalization into the model? The other thing that I think, um, you know, you talked about analysts that understand what the algorithm is doing, but I think you also need business people who believe in the power of data.
Like at Hulu, I think I've been just amazingly fortunate that I have business partners that, you know, believe that data can make the processes better and are able to lean into operationalizing. A lot of the data insights that we provide to them. Uh, even these algorithms, I mean, you can build them and they can sit on the shelf unless someone actually says, yep.
I know how to use them. I know. How it can make a difference to my business. So I think an educated business partner
is, you know, more than half the battle won. If you have that.
Steve: [00:31:13] Now that's, I think it's sometimes called data. Well, it's it's interest, but there's also data literacy. I mean, Duke do companies like Hulu actually kind of train executives to understand data better and the power of it and how to access it.
Jaya: [00:31:29] I think one of the reasons that we created this role was to make sure that, uh, the data team, if the business teams weren't as savvy, that we brought this learning, if you will, or training, if you will, with us to help our exact more than our exact right, the people that are on the ground that are actually running the business on a day to day basis.
To educate them on how to leverage data and you know, when they needed to use that intuition, uh, uh, knowledge of the industry and how they supplemented with data, sometimes they lean into the data more. So. Uh, because they may not have experienced when we're doing new products and things like that. So just, just trying to figure that one out, um, and help train, um, our business partners.
That was a big, uh, part of my job when I joined
Steve: [00:32:32] I just wonder about executives. I mean,
there's all this mythology about the executives and how they make decisions. It was kind of their gut instinct. A lot of people do trust their gut, but, uh, you know, there's all this knowledge, these days about implicit bias about all sorts of things and that you can, in a sense, you almost can't trust your gut.
you see out of an evolution in the way
Jaya: [00:32:57] Okay.
Okay.
You know, as we think, uh, uh, as we say, you know, augmented intelligence, you can also think of it. Uh, having all of these data insights is augmented instincts, right? Um, they're able to look at areas or where they instinctively think. Is that I decision that able to see with insights, whether that is indeed the right decision or not right.
There could be alarm bells that ring that. Nope. That's not the right thing. All the data suggests that that's not the right thing. I think having that. Availability of those data insights is good for them. Um, you know, who am I to say, you know, some months instinct isn't great. Um, all I can do is provide, um, insights that allow them to, you know, either be comfortable with their decision or, uh,
you know, take a second, look at that decision.
Steve: [00:34:04] Well, that's really interesting. It's almost like they might start with an instinct and then just have the, the be conditioned now to check it uh, and then only move ahead with it. If they really find a lot of evidence to back it up.
Jaya: [00:34:17] Yeah. And if you don't want someone to just who buy the data, right? Like you're bringing people in with a different perspective and different knowledge of the industry. So you want to leverage both and you've got to try and find ways to do that. Mmm. And I think for me, it's more fun when, you know, the data insights are used with that instinct and gut than it is just as try new things.
Uh, just based on data.
Steve: [00:34:52] Yeah. Yeah. The Jaya, I think you've used the word fun. At least two, maybe three times in our conversation today. So then I really enjoy the kind of the spirit with which you go at your job. You clearly have a good time at it. And I think that's something that everybody can learn from as well. And I want to thank you so much for your time today.
You know your stories about how, how you proceeded and, and the role you have, and about the insights that people in the company are getting and the sharing between Hulu Disney plus tinea, SPN. Plus, I think those are all really valuable insights that a lot of people will be interested in. uh, So, uh, thank you so much for your time today.
Jaya: [00:35:37] No. Thank you. So. I really enjoyed talking to you.