HCLS Symposium 2021: Eli Lilly

0:00 Foreign.

0:19 Welcome to our session today. My name is Kent Saponczyk and with me today is Chris Blessing. We're both a part of the advanced analytics and data sciences team at Eli Lilly and Company. We're excited to tell you about our journey to democratizing data across the enterprise at Eli Lilly. We're going to go through several different things in our conversation today. First is vision and values, followed by our approach, our capabilities, our lessons learned and then we'll end on our continual improvements or opportunities for improvement.

0:47 We're going to start now with the vision. Our vision at eli Lilly Co. In terms of the enterprise data program is to transform Lilly into a data and insights driven powerhouse by developing a robust data and analytics ecosystem which fully harnesses the power of data to accelerate our purpose to unite carrying with discovery. As we do this, we're going to focus on four major themes. Find, access, trust and deliver. From a find perspective, we're looking at how we can take our data assets and make it easier and quicker to find the information that people need.

1:17 From an access perspective, we want to make sure that as we share those data assets that they have the proper permissions and and security controls. Looking to streamline how we do things today. From a trust perspective, we want to make sure we trust the data both in terms of what's in our environment, but also the insights that people find and how they use that information downstream. And finally we want to be able to deliver in a modernized way and that's in terms of the capabilities and the processes that we're leveraging.

1:42 And as we do that, we want to make sure we're delivering on our mission as a company to make lives better for patients around the world. As we do that though, what's key is to make sure of enabling that vision that we focus on the value that we're going to be creating. And we looked at value through a variety of different lenses, kind of top line and bottom line types of approaches. We did sell our program initially on operational efficiencies.

2:05 So by bringing together the redundant infrastructure that we have, the hardware and software licenses that we have in place, reducing the duplicate investments that took place in each one of the functions we reducing the data maintenance costs, the support cost, the acquisition and ingestion costs, as well as the cost of redundant data were all things that we took into account in our business case. And in fact, as you can see in the slide after year one, we've been able to accelerate our return on investment projection by over 50%, meaning we're gonna Cut that period of time in half, that we believe we can get to those operational efficiencies by leveraging the AWS set of capabilities.

2:39 On top of that, what we're working towards is a stronger use case enablement as well. And we're looking at a couple of different mechanisms where we can begin to see that play out. First is in a research environment we have actually established where we have 900 plus clinical trials across 17 therapeutic areas that can now be accessed in a matter of minutes. On top of that, we have in clinical, we streamline the digital biomarker data collection for analysis while also enabling reusable IoT capabilities.

3:07 Then finally, from a real world evidence perspective, we've been able to reduce our data loads by over 40% while also increasing the productivity of those working in that area by 25%. Additionally, one of the areas that Chris is going to talk about a little bit later is around enhanced collaboration and decision making and the work that we're doing through our marketplace. Marketplace being that that front end, the end user front end that will allow individuals to be able to collaborate, talk with the data owners, be able to exchange insights and ideas and pose new questions of the data that we have at Eli and Company.

3:39 As we do that, we think about how we're going to achieve that value. One of the things that you need to keep in mind is that, you know, we see data as fueling, fueling and generating the insights and solutions that drive value across our flywheel. But there's a variety of different challenges that come into play. And as you can see listed on the screen here, many of those, many of those are there. I'm going to guess that for many of you, you have some of those same exact challenges, but those challenges actually impact the business value and we refer to them as latencies.

4:08 I'm going to talk about this on the next slide. So if you look at the graph that's on this slide, on the Y axis you have business value and on the X axis you have time. And on the upper part of the Y axis, when a business event occurs, there's data that's captured around that business event. And then, and then on the x axis over there to the right, there's action taken. The gap between those two or the latencies that exist actually minimizes business value.

4:35 You want to collapse as much as possible the time between when an action is taken and when that event occurs. So there's a decision latency and analytic latency. And then in our case, most of the things you saw in the prior slide are related to data latencies. That's really the effort that we're trying to address within our enterprise data program is how do we minimize those data latencies to maximize the value with our whole effort here around accelerating how we're trying to get value.

5:00 In fact, our CEO in the summer of 2019, when he told him the story actually said, hey, this is all great, I totally get why we would do this and the value that we can achieve. But to the IT organization said, how do we do this faster? And that's really the journey that we've been on now, as we're now in the second year of the program, is how do we do this faster? So let's talk a little bit about the impact and the approach that we're going to take here in the next slide.

5:26 So when you think about this, you can kind of think about this as a continuum as you see there in the slide around laying the foundation, moving to productization, enabling then data democratization, and eventually being able to leverage data as a strategic asset. And it's along that path is the journey that we've been pursuing. And the capabilities that you see at the bottom are all things that you need to take into account as you're thinking about things on that journey.

5:49 From the platforms, the processes, the data products, the access are all things that we took into account as we went through this journey. But I'm going to take you back a little bit and tell you a little bit about how we progressed down here. We started with laying the foundation in mid-2019 during a phase that we call the MVP phase. We focused in that phase around three things. One, we focused around establishing capabilities. Our initial focus being the enterprise data backbone.

6:14 We focus on data delivery, getting data into that environment, and then we focused on prioritized use cases. We focused on or commercial or our sales and marketing area. Our real world evidence, our clinical and our research areas are where we wanted to focus. So it was efforts around the patient and around the healthcare practitioner as we felt were the most important areas for us to focus. As we did that, we applied a federated approach. So we, we put everything in the center and built a common enterprise solution.

6:39 But then we work with the teams at each one of our business functions in that federated manner to deliver the data and the use cases. So our intent being is to deliver in an agile fashion, incrementally delivering value and then building the excitement around the use of the platform going forward. We then shifted into the productization phase where after we started to get the data into the backbone. So we started to introduce our concept around the enterprise data marketplace.

7:04 We started to then introduce data products. So we're positioning data not just for the primary use cases of that data, but also the secondary uses of that data. And then that put us in a better position for the enabling the data democratization, not having that data available as a product in marketplace. And then we started to layer on top of that the analytic workbench which is more the advanced set of capabilities. And together the EDB Marketplace and the analytic workbench work as one data analytics ecosystem to streamline how we deliver data in a different, more modernized way.

7:35 At this point we're in what we would call the institutionalized phase of 2020 was our scale phase. And now we're in the institutionalized or adoption phase, the program and Chris is going to talk to you more about the capabilities in the architecture. So I'm going to turn it over to Chris now to tell you more about those areas.

7:51 Thanks Ken. My pleasure. To explain more about the entire ecosystem. Our ecosystem is made up of five key areas. First, enterprise data backbone where storage, data pipelines and many other services and features come together to deliver data products for sharing and analytics. It's a very exciting environment. It's a single environment built in Amazon Web Services. We're really excited about this because we've been able to harvest and harden many capabilities from all of our teams within functions and from the center and create best practices for reuse, efficiency and cost savings.

8:31 Next is important to avoid the pitfalls of data lakes. We've learned lessons from the past. We catalog all of our business and technical metadata in one single repository, both to deliver information to the users and for us to manage our data appropriately. Next, it's really important for us to focus in on how to deliver the experience desired that is done through the enterprise wide data marketplace. In this experience, users can find, manage, access, use and collaborate data analytics and insights.

9:09 It's very important to create the seamless experience for both our features, our services and our reusable capabilities. Also, in addition, this is where others can access information for training as well as register as a new user. Now really going forward, we really become Eli Lilly and company, becoming a data and analytics powerhouse. The analytic workbench is very key for us now. We've integrated this into the marketplace experience as well as the enterprise data backbone. This is where our advanced data scientists and data engineers can use the best of tools, services and technologies to find insights, build models, train models and monitor.

9:49 It's Very important though as a pharma company that we don't forget. We have very important information and we have a data governance group that data governance across the entire ecosystem, across the entire program is very important. Now these are our five key areas. But none of this is as important as our OCM strategy. Being able to communicate, to drive change, to create a modern ecosystem, a modern way of sharing data and additional capabilities to really uplift the entire business and IT organization to really take advantage of all these services, all these capabilities and really bring the power of data to the forefront.

10:32 Now to explain a little bit more how this is done, we really focused in on six key areas as we design features and harvest and harden again those features throughout our industry, throughout our company, from areas across the functions in our key partnerships. We've taken the best of all things to create multiple methods to ingest data. Now not only are we storing data one time at Lilly in its raw form, but we're integrating that in with our catalogs, our ability to monitor the success of receiving data, but we're creating reusable patterns.

11:07 These patterns and these capabilities allow us to configure fast leverage CICD to produce and grab new data types. That is from our traditional data types of structured batch processing to real time IoT data for analytics and streaming. Now that data is there and it's great to take the next step and build data pipelines and processes in multiple ways, leveraging key services like Blue, ETL and Lambda functions as well as other self service tools and advanced capabilities for different types of users that build data products.

11:43 Now we have information management principles and practices for data quality enrichment and the use of creating data products for multiple purposes. In his space we have trusted data products for broad use as well as trusted data products for single analytics. Now we drive that to delivery, we drive that to delivery in applications through the consumption layers and APIs, our API gateway from Amazon Web Services as well as visualization tools. And we we provide a standard set of visualization tools taking full advantage of quicksight plus some of our traditional tools integrating that a secure high performance way.

12:21 Again at the bottom we have the analytic workbench. This goes across the entire ecosystem. We're able to already institutionalize how we do natural language processing for unstructured data sets as well as beginning the journey on image processing and videos. This area also includes model management and the ability to monitor the model and look for biases, then drive that into visualization. On top of that we still have the marketplace bring that collaborative experience to be able to find data, request access, gain authorization to use it appropriately and share each with each other about how to use the adult data and what is possible within our limitations.

13:06 Now all of that is great, but we have our foundational core elements and info security is one of the key elements and our ability to monitor performance within the ecosystem and throughout its use in and out. Now, to move us a little forward, there were really key perspectives in partnering with aws. First, we were introduced and leveraged many of the product owners and product teams as well as other key partnerships.

13:40 Whenever we had something new or a new use case or a new product we want to create, we went to them first and allowed our they enabled our partnerships to partner as well with them. This was instrumental from going fast again, the vision is easy to understand. We all have similar problems. But to go fast you need to reduce the barriers, go straight to those who are the experts and collaborate again. Harvesting and hardening from all over industry and all over our ecosystem at Lilly, but we didn't stop there.

14:14 We were going towards a product mindset. The ecosystem isn't just an infrastructure that we leverage. It's taking the Amazon web services, bringing them together to drive key features throughout. These features not only save us time, they save us money and they create repeatable processes to drive the innovation is needed around the data. In addition to this, each product that is promoted and then published also has key characteristics.

14:46 All right now, it's designed for reuse, it is also understood, but most importantly it's secure. So we don't forget these things. A product mindset is we have these items, but you have to have the right rights to obtain and procure these items. Now those are logical concepts, but it's really that experience, it's the excitement, it's the understanding that's really going to drive innovation to the forefront of the data and analytics environment at Lilly.

15:17 Now, to drive success, what worked well first is creating a key partnership with aws. In addition, AWS enables have other key partnerships through our professional services and other key vendors. This holistic partnership has really helped us drive innovation quickly. We bring our questions to this group, they answer them as best as possible. What they do great about it though is they answer the questions they provide. How do we use the right services together to deliver value?

15:50 In addition, we'll never stop improving and we'll learn to avoid cost, create a cost reduction and leverage the best services for the right use cases. Now we do this through the well architected review. We do this every six months. From that I take and My technical teams take the key elements learned and add those to our backlog. We now have the continuous improvement built in to provide the best platform and use of all of the services.

16:23 Last but not least, what's been great about working with AWS is we have access to the innovative product owners and product teams who are not only sharing with us what's been built, but also what's coming. If you've been at Re Invent, it's been an exciting experience to have these sort of conversations. In addition, we've been able to connect to our peers in industry and learning from each other, much like we're doing today. It's been a very exciting journey.

16:49 Now, you know, there are a few things that didn't work well. One is there are preferred tools that users have today. Sometimes familiarity and key features are missing from what we exist on our endpoints from Amazon Web Services. But we do realize that having those integrated into an ecosystem in a smart architectural way has really helped us. So we believe putting those tools next to the data really creates the performance and the security to leverage the best of both breeds.

17:20 In addition, we design it to be able to pivot when the market changes. Secondly, the ability to really key in on advanced services provided by aws. In the analytics space, there's a lot of great materials for the technical audience. Those same materials need to improve for advanced data scientists and data engineers in their context. So having the same information provided in multiple ways will be a benefit of the future.

17:54 So how has AWS really helped us bring it together? Well, first, we have a great account team. The account team is on point to get us the right connections in the right areas when we need it. Again, they've really accelerated us. We're not talking weeks and months, we're talking days to gain access to the product teams, the experience of others, and some of their support. One other item has been really fabulous is their professional services.

18:26 We've made their professional services available to each of our teams. Having Amazon Web Service professionals on site with us, even in a virtual environment has created the ease of access, plus given us the confidence to move fast, make decisions really quickly and drive change. In addition, there's a lot of training resources. There is a lot of other access to individuals that really have brought new ideas in addition to really helping us understand the product and the release cycles, be able to pivot and really look at an entire universe.

19:10 Again, the thing we've delivered in the past year that we scaled is a single data and analytics environment. Last but not least, I really Advise everyone to attend. Re invent. You create the excitement not only from learning like you do for many conferences, but connecting with others who are really leaping into the modern ecosystem, the modern data architecture, and really taking advanced intellects to the next level. Insights will be the foundation for an ever changing experience in healthcare.

19:44 All right, so now we're going to talk about the learning both applied and acquired as we went along our journey. So first of all, we cannot emphasize enough, as you see at the top of the slide there, the right people, right culture and right partners. And we're going to talk more about the right culture here at the end of the presentation. But from a people perspective, it was extremely important to, as we look to put together the team at Lilly and with our partners that we look for those who are passionate about data, we're able to influence the organization.

20:10 Strong leadership skills, ability to take on a problem on, take on the risk and flip those negatives at times into positives. We also wanted people that are comfortable with ambiguity, not knowing exactly where the program would go. We look for individuals that could play into a variety of different roles. And we've been very proud to say that we've got an outstanding team in place that is able to exhibit that across the different capabilities that we have.

20:32 Also important is to define the scope very clearly. So we focus on centralized ownership. So we build the solution in the center and then work again, as I mentioned earlier, with our partners in a federated manner to deliver the data and the use cases. And that's actually proven through that governance model and then their use of Agile and kind of our Scrum of Scrums type of approach where we bring together on a weekly basis and we ensure alignment.

20:56 In addition, we use a quarterly PI planning process and in there we get alignment across the different business functions, both around the business priorities, but also the capabilities that we need to have in place and the data that needs to be there to enable them. So that Agile approach and that PI or that program increment planning has been fundamental in being able to have that centralized and federated approach work for us as we went. We also started from the very beginning planning with an enterprise mindset and then implementing incrementally from a governance and change management perspective.

21:25 Chris touched on this a little bit already, but OCM from day zero is super important. In fact, we started from the beginning. I almost wish we would have started earlier when we're actually doing the planning for this from an OCM perspective, getting that senior sponsorship is also important. As I mentioned earlier, our CEO early on asked about how we could accelerate, but we also had the support of our cfo, our Chief Information Digital Officer, our CDAO as well as our cto.

21:51 We're all on board with the program and provided that senior level sponsorship for what we were trying to do. We also recognize the importance of collaboration and partnership and at times that will be challenging, particularly as we expected some level of resistance around the data sharing and then creating a level of transparency and accountability. We had clear metrics around the data that we had onboarded the data products that were created, the use cases and then the outcomes again, both the operational efficiencies as well as the business outcomes that we were enabling.

22:18 And that was very important in order to continue to maintain momentum throughout. And then the final element of governance was very strong consistency from an architecture perspective. Chris and his team worked to establish guidelines and consistency early on with our enterprise Information Platforms team. And then we made sure that the different teams that came on board understood those and then worked with us to maintain those as we implemented the program throughout the first year and a half.

22:43 And now I'm going to hand it over to Chris to talk about the data and technology and the adoption learning.

22:48 Thanks Kent. Like Kent has said, we really focused in on cloud native first and with AWS we were able to do that. What's amazing is we have a variety of data, we have a variety of data pipelines. We use over 60 services and technologies in this ecosystem. It's important for us to standard the use of those acts. Standardizing is important both from a data perspective from information management principles plus modernizing the ecosystem, both from a platform and a modern data architecture.

23:20 Now we've been able to do this pretty simply with Amazon's web services help. But like Kent said, we're doing quarterly releases. We're always mirroring up what innovation is happening in AWS as well as the innovation throughout our functions to bring out new ideas of improving features we have or enhancements to really meet their use cases. This is done on a quarterly basis. That's why release management is so important. So the fourth area, adoption, very important to us.

23:51 First we prioritize use cases and features to match those use cases. We did both at the same time, increasing the platform and delivering use cases. This is very important to show both value in the platform as well as business value. In addition, we leveraged a user centered design. So what is the user experience for the variety of different types of users, whether that be those who are producing data products in the environment and the consumers. This Meant we don't just focus in on the IT audience, we focus on our business as well as other areas.

24:23 The last piece is we focus in on centralized funding. Now, how do you drive quick change when you start looking for those priority use cases, those teams throughout the functions, bring them in and help invest in them to deliver faster. Both features within the platform as well as the use cases. This was very pivotal to our model. This removed the barriers of everyone trying to ask for separate dollars. This was instrumental.

24:51 All right, so we're going to close things out with opportunities for continual improvement. So some of the learning that we've had along the way is one, recognizing there's tensions when you try to do something at the enterprise level of trying horizontally across the enterprise with the functional priorities that exist and having early and unambiguous support throughout and particularly in that middle layer has been very important because that's really where we're talking about the resources and the funding and the priorities that need to be managed.

25:16 Second, realizing data access just isn't easy and you know, we've worked very hard. In some cases we've pushed more aggressively than the business was ready for and we've got to recognize kind of that middle ground to be able to achieve that. And then the final thing again, so I won't hit on all these points will be on the, on the right side there. The shift in the right to left thinking, particularly as we go more into the adoption phases now and the institutionalizes, how do we begin with what the end outcome is that we're trying to achieve and work backwards from the right to the left in terms of then the data that's needed and the capabilities that need to be delivered.

25:46 And that's one of the things that we're now shifting as we go forward. So we want to end on the team Lily side and talk about the right culture that I alluded to earlier. Team Lilly refers to four key behaviors include innovate, accelerate and deliver. And those behaviors when exhibited by the entire team. Both Lilly and our partners have been the differentiating factor for us. We'd like to take this moment to thank all the 270 plus people that have been a part of this program, both in the business and within the IT function, as well as across different geographies and different partners such as AWS, Deloitte, PwC and TCS who have all worked closely with us in this program.

26:25 We'd like to thank you today for joining our session and hope that we've shared something with you today that you can learn around your journey to data democratization. Thank you.