How data powers Oscar Health’s mission to transform Healthcare

Tarush Aggarwal (00:00)

Welcome to another episode of People of Data, where we get to highlight the wonderful people of data. These are our data leaders from various industries, geographies, and disciplines. Discover how companies are using data, their achievements, and the challenges they face. Today on the show, we have Thiavani Palaniapan with us.

‍

She spent years helping companies make smarter decisions by using data in meaningful ways. From leading teams at Oscar Health to her long career at Wells Fargo, she's all about solving problems, improving customer experience, and building strong teams. Welcome to the show, Thevani.

‍

Theivanai (00:38)

Thank you. So excited to be here.

‍

Tarush Aggarwal (00:39)

Same, are you ready to get into it? Let's do it. So for those of you who don't know, what does Oscar Health do? And how does data play a role in helping Oscar Health do what it wants to do?

‍

Theivanai (00:42)

I'm ready.

‍

So, Oscar Health is all about trying to make healthcare affordable and accessible for all. And really when I say all, I mean trying to get to everyone who needs healthcare. So, a little bit about maybe the US healthcare system first, because I know your audience is diverse. So, in the US, healthcare is not required. It is a choice. And so, there are different ways folks can get health insurance. A lot of people get it through their employer. There's also the government who has Medicare and Medicaid. They cover that.

‍

And then there's the ACA, which is relatively new. So that's the Affordable Care Act, about 10 years and counting. It was intended for folks who didn't have insurance through the other means. And so it brought this whole segment of folks in who typically didn't have a lot of choice. And so very excited to be able to serve these folks who really now I think we provide a really excellent customer experience too. So the market was roughly 20 million plus folks who didn't have insurance in so.

‍

Oscar serves 1.6 million folks, members across 18 states. And so it is, I think, a really nice choice product. So they can go out, you know, figure out the solution that's best for them. We offer things based on conditions with different language preferences, so really a nice product. So Oscar Health is a health insurance company. We are a payer, and so we are, the member would go out just like they choose other health insurers, and they would choose Oscar largely in the ACA market.

‍

Oscar being a relatively new company, I think one of the very exciting things is we don't have legacy systems to deal with about 10 years old. And so we built to be very data forward, very tech forward. And so our systems collect a lot of data. We're always trying to figure out how to use that data to make our members lives better, to make it easier for our providers, our brokers, really change the U S healthcare system to make it more transparent. So little story about US healthcare.

‍

me away when I joined Oscar. So I'm relatively new to Oscar, three years here. And so one of the things that really surprised me is, well, this part didn't surprise me, but healthcare in the US is not very transparent. So when you go to the doctor, especially when you go to like the emergency room in the hospital, you often don't know what you're going to pay. And so it's not like when you go to the market to get bread, it's like you go in here and you, it's just, it's a shock. And so the more that you can prepare people, the more that you can make it transparent. And that's really where data comes in. How can we make it something that people understand?

‍

and don't get surprised and make sure that they make very informed decisions. So really using data on everything that we do.

‍

Tarush Aggarwal (03:12)

That's awesome. How do you describe the company culture? Because, know, the gist of it, seems like data is just a very integral part of the culture. You need this data in order to go figure out how to price your products. you know, so it's been there since day one. So how does that really affect the company culture? And then more specifically, what's the data team culture like?

‍

Theivanai (03:32)

So I think that the company growing up, everyone is to an extent across the company, very data interested. And so it is a part of almost every decision we make. It's really fun for my team because we get to be part of the business making decisions. And so it's things like, like you said, how do we make the right plan for the member? And so we look at what has worked in the past, what is not, what are our customers conditions look like?

‍

Is there a particular geography that we see a particular chronic condition happening more often? And so can we create something for that geography to really vary data forward? So the business looks to us, the people who create the plans look to data, data really partners with them to create it. And so it's a very collaborative, like data is part of the business, the way we really run things at Oscar.

‍

Tarush Aggarwal (04:18)

What's one of the things about the culture which surprised you when you joined? I love the example of the US healthcare systems and your insights into that. What's something at the cultural level? Why Oscar for you?

‍

Theivanai (04:32)

I think because of the mission, the affordable and accessible for all, that to me was a real draw that it's not just a mission statement. It's really something that the company lives and breathes. We're one of the few, maybe the only one with a high net promoter score in the US. Because like I said, again, healthcare is not an experience people generally like and look forward to. But because we have made it very digital, very transparent, very we'll give you as much information as we can. I think we do a nice job with our members.

‍

And it really is front and center to what everyone does. It's how do we use our data, our technology, our processes, to best serve our members. And so that's just part of the culture here, which has been a really exciting place to be because it is really everyone's working in that same mission.

‍

Tarush Aggarwal (05:15)

Yeah, know, very typically private companies are, you know, sort of offer their own employees healthcare and things like this. Do you do employees at Oscar ever get to, you know, try out and use their own product? Like do all of you have because I know it's a different segment. So we'd love to hear any thoughts on that.

‍

Theivanai (05:34)

Yeah. So it was there before I joined. So before I joined, it was largely a New York employer with New York employees. And so they did have a local insurance that you could use in New York. With COVID, the companies become very distributed. We have employees all over the country. And so to offer that very specific geography care is not really feasible for everyone. And then there were also some privacy implications as we became a larger and public company.

‍

Tarush Aggarwal (05:41)

Yeah.

‍

Theivanai (06:01)

that we didn't offer it. I will say I have a lot of employees on my team who have been here when we had, it was called Oscar for Oscar, and they absolutely love it. And people are clamoring to bring it back because it was a great experience. And I know it is on the list of things to consider. Can we work through the technology and bring it back? So it's out there. It was there before and very well loved.

‍

Tarush Aggarwal (06:01)

Yeah.

‍

That is such an incredible feeling to have your team actually use, of, food your own product on a daily basis. And I love that it's on the roadmap again. And that totally makes sense of why it must be so difficult just given privacy and all sorts of things which become exposed. Talking about using your own product, what does power the...

‍

you know, this sort of data stack at Oscar. And we'd love to know more about the data team.

‍

Theivanai (06:56)

Yeah, so our technology, the data team in particular across Oscar Engineering uses a whole variety of things. Our team in general uses BigQuery, so Google products. We get a lot of our data from our engineering systems. We get a lot from vendors, external sources, put a lot into making sure that data is really of a form that we can use for analysis. So we have a lot of governance process in place, a crazy number of integrity checks in place just to make sure that

‍

everything is set up correctly. We also try to put our metrics into our data warehouses itself, because we found if we let them be calculated outside, then we have different definitions. And so we put it all into our really well-governed BigQuery source, which is what most of the enterprise uses. We have a reporting layer. We use, again, Google product. We have Looker. And so that is our front end. Again, to the...

‍

Tarush Aggarwal (07:37)

Amazing.

‍

Theivanai (07:46)

Everyone at Oscar loves data. We have tried to create a whole suite of reports that folks can get to and slice and dice every which way. And so they're using sort of a very governed process, but they can get the insights that they need. We're constantly sort of evolving and building that.

‍

Tarush Aggarwal (08:01)

What do you use for data catalog or data governance? And then you mentioned there's a lot of checks on the modeling layer. Do you use a specific modeling tool?

‍

Theivanai (08:10)

It's very homegrown right now. So Oscar, like I said, is a relatively new company. And so the focus was on getting the data, being able to answer the questions. And so the public company is about three years old. And so that's when we started putting a lot of these into place. So we're still early. We don't have a lot of external tools that we're using. is largely homegrown. And maybe another thing to point out is, sorry, my group is really embedded. And so we understand what the business is going to do.

‍

Tarush Aggarwal (08:31)

That's so exciting.

‍

Theivanai (08:37)

And so our checks are based on, know how the business is going to use it. And it's not like a, we have the, the field shouldn't be blank, but we also have a, we know that this should never be greater than this because we've worked with the business enough. And so we have a whole bunch of very specific roles.

‍

Tarush Aggarwal (08:51)

How big is the data team and how big is the company?

‍

Theivanai (08:54)

we keep growing. When I started, we were probably around 30 or 40, we were probably 60, 70 right now. The company is around 3000 employees. Like I said, we use data everywhere. And so the way that we're organized is really across the data science teams, we're organized by business area. And so I have like a set of folks who really understand networks and contracts, another set of folks who really understand claims payments, so very embedded in the business areas. And so that's how we're organized across our teams.

‍

Tarush Aggarwal (09:19)

Yeah.

‍

And how big is the entire data team, the entire data group?

‍

Theivanai (09:24)

So I lead, it's largely centralized at Oscar. And so the 70 people on my team are a lot of it. We have some analysts sort of embedded in other groups, but the data science itself system is centralized 60 to 70 % team.

‍

Tarush Aggarwal (09:38)

Very awesome. in the data science world, obviously everyone today is thinking about AI and a lot of people have AI anxiety. What's Oscar Hell's take on AI and any of the technology stack around the AI world which you feel comfortable sharing?

‍

Theivanai (09:55)

So we have been working with OpenAI for a while, from the beginning. So we certainly are using a lot of their product. I would like to say most people, if not all people on my team, access it relatively often. So we do use theirs. We have deployed a number of use cases. A lot of it is for internal, like, can we figure out the intent of a call? Like, how do we really understand what our members are telling us or our providers need?

‍

We have a couple of external facing AI use cases right now. They are largely with an internal Oscar person in the loop. they would be things like you're reading, you know, there's a lot of press on this. A lot of companies are coming out specifically for things like this, but can we help the doctors with their notes? So, Oscar, one of the cool things is we have our own sort of virtual care practice. And I think many insurers have that, but we have this set of data that's very rich and we have our own doctors. And so can we help them code, you know, do their

‍

their notes quicker. Can we help them with their follow-up? And so we wouldn't do anything member-facing, but we would do that draft. And that would save a lot of time and inefficiency. So sort of the use cases that you read about in the media that company niche companies are coming in, I think many of them we do, especially given the data as ours and we're good to go because we built our tech stack to be able to do these sort of things.

‍

Tarush Aggarwal (11:10)

Very exciting. Talking about tech stack and deploying data science use cases in production, what's one achievement which you've had the opportunity to spearhead, which you're proud of, which has had a significant impact on the business?

‍

Theivanai (11:25)

So I would probably hit two, I know you asked for one. I would start with data governance. I think that was one that, again, we moved really quickly and people were moving fast and very well intentioned, but that meant they would go where they could get the data quickly. And it didn't necessarily come with the governance. might've been in someone's playground. And so someone may have put their own filters on. So you didn't really know what you were getting. And so to move away from that, everyone use whatever playground you're used to.

‍

Tarush Aggarwal (11:29)

Yeah.

‍

Theivanai (11:51)

to using these centralized data sources across the company has been a really big change, which I think has added a lot of value. So that's one I would call out. The second one I think has been there, and it just continues to get deeper and deeper, is that across the business, people really value the insights that data can bring. And so it is part of the, I'm not just going to do something or I know I need to measure it. is really like data is integral to everything we do. And I think that's really unique.

‍

to want people, to have people want those insights before they make decisions. The business thing, I wanna be able to measure this, help me. So I think that's been really rewarding and it's nice to see sort of the impact.

‍

Tarush Aggarwal (12:28)

Yeah.

‍

I love these two, you know, everyone, you know, we speak to so many different companies and one of the challenges they face is, you know, data can be exported and, you know, shape shifted and, know, you can always zoom in and zoom out and have to go, you know, have data support, any, any set of hypothesis you have.

‍

How were you able to move away from what I call the sort of playground of data where anyone can go, know, store in the local computer or their favorite BI tool? How did you move away from that into a more governed world where people are accessing it, you as you said, from your front-end, which is Looker.

‍

Theivanai (13:08)

I think some of the advantages that we have is this centralized team. And so it all really sits in one shop. So if the mandate is everyone will go to this centralized source, they do go. And so that's been part of it because all the data scientists sit in one place. The other is we decommission those other sources, those playgrounds, a lot of them are gone. That doesn't mean we're a hundred percent governed. People are still going to be doing their own things, but we are trying to push them to the centralized looker platform.

‍

Tarush Aggarwal (13:30)

Yeah.

‍

Yeah.

‍

Theivanai (13:34)

where

‍

everything comes out of these data sources. And so don't go create your own thing. If you need something, come tell us and we'll figure out how to put it in a Looker BI so that you can then go pull the insights. So really try to make it so easy that they don't need to go do their playground work.

‍

Tarush Aggarwal (13:48)

Yeah.

‍

Yeah. And you still have the export button from Looker enabled where they can, if they wanted to, they could go do it because you're starting to see all sorts of tools built on top of GPT. We are about to list that AI, which is an incredible job and pretty much giving you a PhD level analyst to go analyze your data. it is tempting to export.

‍

into a CSV and load that into any tool and that's got all sorts of privacy and security concerns. So what are your thoughts on this?

‍

Theivanai (14:24)

Yeah, so our access controls some of it. So it's very tightly controlling, especially like PHI. There's only certain people who can access it. And so you can't download that by mistake in most cases because of our access controls. That being said, you can always download Excel. You can always download into a Google Sheet and manipulate it and do something. I think what I like to think is a control is that when it gets to leadership,

‍

very well aware of our centralized data sources. And we often get the question from leadership, did that come out of a centralized data source? Did it come out of one of your sources? And therefore we're comfortable with it. And so it's a check. It's not a hundred percent. People will still do what they need to do, but hopefully we're making it less necessary for them to go do these things.

‍

Tarush Aggarwal (15:13)

I love that, like on a scale of 1 to 10, how data-driven would you see EOP?

‍

Theivanai (15:16)

he's data driven. would put him very high up there. Nine, 10.

‍

Tarush Aggarwal (15:20)

And on the other side of this question, what are one the challenges which you are facing or have faced which you are currently focused on solving?

‍

Theivanai (15:30)

Yeah. So one of the things that, that I'll say has always been there, but it's sort of hitting me even harder now is we have lots of data. We're really a data forward company, lots and lots of it. have lots of metrics. think sometimes we have too many and that's where I, I spend time worrying. Do we have so many metrics out there that people may miss the really important ones? Are the metrics the ones that are really going to drive action? And so really trying to think through how do we make it not

‍

only a data-driven company, really an action based on the data. So it is, I think every company is on a journey in different places. I think we definitely hit that first. have the data, how we move to the insights and the actions as much as I would like. I don't think we're there yet.

‍

Tarush Aggarwal (16:15)

And how do you solve this?

‍

Theivanai (16:16)

I have been slowly trying to change the conversation from instead of give me data and give you metrics to tell me what you need to do. And then let's collectively think about what information would help you make that decision. And then measure later, did it really get you what you wanted to do? And it's, it's a hard focus because it's easy for people to say, tell me the payment accuracy rate or tell me this, but not if it crosses X threshold, then I'm going to worry and need to do this. Cause I would love to know.

‍

If it crossed this number, then it's a problem because I don't want them to have to go look for a report. I'd love to have something just spit out to them, give them an alert. Your payment accuracy rate crossed X percent today. You need to go look today. And so I'm trying to change the discussion, but it's, it's, it's hard.

‍

Tarush Aggarwal (17:00)

Yeah, and if someone wanted to go define and alert, right? If someone wanted to be proactive about this and you want to move to alert-based, do x, y, when something happens, how do you do that today? You can't go define and alert a looker or maybe you can. How do you?

‍

Is this something you want to enable for the team to do start service so they have to work with the data team? What are your views on this?

‍

Theivanai (17:28)

So actually Looker can do it today. We are not using it to its capability. And my team can certainly set it up for folks, but they can really set it up for themselves too. So there's a combination that we can set up, send out reports, we can send out alerts. So the tool does allow for it today. I don't think we're utilizing it fully.

‍

Tarush Aggarwal (17:30)

Enough.

‍

Yeah.

‍

Awesome. if there's one, that's a really great answer. If there are folks listening to the show who are interested in Oscar, what is one piece of advice you would give for anyone who's looking at the data team at Oscar?

‍

Theivanai (18:03)

So I would say a passion for making a difference for the member. We really are all about the member and curiosity. So one of the very exciting things for me and probably even my folks have been here a really long time is everyday healthcare will throw a new challenge at you. There will be just something new that you did not think of. Like the fact that you can have two, there's a recent Medicare article in it. This is not ACA, but Medicare, where people, two people can have the exact same insurance, go to the exact same facility.

‍

but pay completely different costs. Like how does this happen? They came in with the same insurance. so just every day we learn something new. And I think that's what keeps it really fun. And so curious, willing to ask questions because a lot of this, even though there's a lot of data, it hasn't been mined. Healthcare feels very new, very green field, even so many years in.

‍

Tarush Aggarwal (18:39)

Yeah.

‍

Tehwani, thank you so much for being on People of Data