Diagnosing the Problem with Centralized Data with Omar Khawaja, Head of Business Intelligence at Roche

Hashmap on Tap Ep. 111

--

Photo by NeONBRAND on Unsplash

Omar Khawaja is Head of Business Intelligence at Roche where he and the team are building data products utilizing modern data principles and a modern data stack. He recently joined Hashmap on Tap host, Kelly Kohlleffel on the show to discuss data ideas and principles from the healthcare industry. Roche has been around since 1896 and provides innovative medicines and diagnostic tests to millions of patients globally. As the world’s largest biotech company and leader in diagnostics, Omar Khawaja has plenty of fascinating and critical use cases.

Omar started his career in tech over two decades ago. After graduating with a degree in computer science, Omar started out as a programmer and loved it. This led to his time at Unilever while he was still in Pakistan. After doing some ERP-related work, Omar jumped into the amazing world of data and analytics. His time at Unilever provided a lot of great experience on the business side which provided Omar with powerful insights on how data can be used. This experience allowed Omar to truly visualize what it’s like to be in an executive’s shoes and see just how valuable data and analytics can be. He gained a better understanding of how one sales number can change people’s day-to-day life. So now, Omar truly understands just why the data needs to be accurate as well as the importance of a great data pipeline and why it’s important to the users. After working in several IT roles, Omar looked to expand his horizons past the pharmaceutical world and into the diagnostic area. This brought him to Roche, where he is the Head of Business Intelligence.

“What attracted me is that (Roche) is a great company with a great purpose. I really love their ambition of doing now what the patient needs next.”

-Omar Khawaja

In this podcast episode, Kelly and Omar discuss just how Omar and his team manage massive amounts of business-critical information and deliver it in a way that is valuable and consumable to its users. Below are some key insights into their success.

Are there some data trends that you’re seeing on your end of the Healthcare industry?

Omar has found that the importance of diagnostics has been quite accelerated because of the pandemic. Worldwide, people now have first-hand experience on the importance of diagnostics.

“As we move forward, I hope we can remember that diagnostics play a key role in the life sciences world.”

— Omar Khawaja

The diagnostic world has been accelerating at a rapid pace because of the amount of data we now have access to. Almost everyone has some sort of handheld device that has the capability of tracking key diagnostic metrics, and the life sciences world is continuing to improve upon personal and self-diagnostic applications. So now the question is, “How are we going to utilize this massive amount of data we are generating for the benefit of the patient?”

What’s been going on at Roche from a Cloud/Data perspective?

It’s an exciting time at Roche. Like many others, Roche has many different angles and paths from different technologies and different warehouses. They’re in between on-prem, cloud data warehouses, data lakes, and looking at new concepts like data mesh. Technology plays a key role at Roche as it is a key principle. The way they are implementing now is different than how they have implemented in the past, where more things were done centrally from the central data platform teams. Now things are shifting to how these teams are enabling other data teams to utilize the data platform and its capabilities. This shift has allowed the company to be more scalable and truly data-driven because they are implementing these data teams everywhere instead of confining the data to just one group.

Of course, this comes with its own challenges like data governance and keeping these de-centralized teams connected. Although there are some boundaries and parameters for each team, there is a lot more freedom and flexibility so that each team is empowered to get the most value from the data.

How do YOU define Data Mesh?

“So that’s a million-dollar question and we can do a separate 2-hour webinar on that.”

— Omar Khawaja

Omar’s first reaction, like many, was that data mesh was just another fad data concept that should not be given too much thought or attention. The way technology is expanding and growing from different companies and vendors is incredible, but it doesn’t make picking and focusing on specific technologies easy. The concept of data mesh felt like it addressed all the problems Omar has witnessed over the last 2 decades. For Omar, data mesh made him think that it’s time for all of us to change our approach. While there have been a lot of great things that have come from data warehouses and data lakes, there’s not just one element that’s important. It is never just technology, it’s technology, people, process, and data. The paradigm shift that is data mesh really acknowledges and captures all these factors. It has been defined as a social, technical framework or an architecture pattern.

Data mesh has shifted from a domain-centric approach to an end-to-end lifecycle managed data product approach. This approach is a huge shift from traditional data approaches. While the concept sounds incredible, data mesh still presents its own challenges. One challenge is shifting the mindset that data isn’t just something to protect, but a product that you share and empower others to utilize. Data mesh allows data to be a living and valuable product as opposed to an object that sits in a data warehouse waiting to be imposed on a dashboard that may or may not be useful. This concept gives a purpose behind data. This restructuring allows data teams to be enablers instead of a roadblock, which completely changes the dynamics of a company. Omar believes that data mesh could break the cycle of repeating the same challenges we have had with data for the last few decades.

Are there specific outcomes that you’re looking to drive right now?

“Absolutely, so many. And instead of thinking about these outcomes at the end of the process, this is one of the only things we talk about at the beginning.”

— Omar Khawaja

Omar leads his team to forget the data, forget the output, and approach, and when presented with a challenge, they focus on what business outcomes the team wants to establish. This changes the dynamics of who’s working on what and how they approach a challenge.

When you consider key design principles, what’s most important to you as you’re building that next data product?

“It needs to be focused on delivering that outcome for the business.”

— Omar Khawaja

Omar finds that data ingestion, transformation, and integration change every day because the data is growing from the expedited digital transformation. You need a variety of capabilities to handle all of the various data ingestion patterns. There is no “one-tool solution” for your pipeline, you have to find the right mix that is best suited for your business, points of ingestion, and database. Today, Omar feels that he might have the mix identified, but that could easily change tomorrow based on new data sources or new types of data coming in. So then the process becomes evaluating those data ingestion tools to see what needs to be updated, removed, or added to the pipeline.

What’s your experience been like with Data Vault 2.0? How are you using it?

At the core of Roche’s data product approach is the idea that Data Vault 2.0 should be applied within each of their data products instead of having a central data warehouse team apply it. While Data Vault 2.0 generally takes a centralized approach, Omar challenges his team to apply it to each of their data products within their domain practices.

“The reason for choosing Data Vault was the length and flexibility of cost it provides over the short-comings of the previous practices in the area of analytics. The flexibility it adds when it comes to adding both the data lake experience and the data warehouse experience in the same practice.”

-Omar Khawaja

At Roche, the entire data pipeline is owned by the data product team that is applying these practices. They can apply these practices in a truly agile way rather than waiting for an ideal data warehouse design or taking a data lake approach and dumping the data until they can figure out what to do with it. They achieved this approach by having a team that has a good grasp on this domain. They can look at the data sources and define the data vault for their own domain and for their data product. Even if the data sources are changing, they can adapt easily due to the flexibility that this model provides. In a data mesh way, they can create multiple consumer-oriented data products on top of the raw data walls that the data product teams own. Omar thinks it is a great approach as long as you have a team with the experience and knowledge. People play such an important role in this model from the setup of the product team to their personal skills and capabilities. As long as you have these things, this can be quite a valuable approach.

“To me, it was a match made in heaven between the two approaches we were thinking about.”

-Omar Khawaja

What does a typical development cycle look like at Roche?

For Omar, this goes back to truly operating as a data products team. In his ideal world, the product team follows the development cycle steps below:

  • Dig into data discovery through exercises like opportunity trees
  • Look at different hypotheses
  • Testing those hypotheses
  • Finding the gaps in which they need to focus on
  • Develop a product that addresses those gaps with the features they plan to deliver
  • Deliver Outcomes

Unfortunately, not everyone uses these practices because like everything it takes practice. So it has been a learning journey for Omar and his team. Fortunately, they were approached with ideas from intelligent business and IT people, where they planned to apply their traditional approaches. They then built upon those success stories by asking, “What are the business objectives we are trying to solve?” This is what got the ball rolling. The next questions they asks are “Is this a problem that may be a good space to be solved by a data-driven use case?” and “Who is the end-user for which we are solving?” The end-user could be anyone from an executive to a person on the ground, so this provides a better idea of what kind of product needs to be delivered. Traditionally, Omar and his team have found themselves asking these questions too late in the game. Now, they are learning to ask these types of questions at the very beginning of the process. At this stage of the process, there has been no talk about the data. The conversations have been centered around the problem to be solved and what metrics and key indicators might be helpful to generate insight to solve the problem. After they have fully discussed who will be looking at those metrics and using these tools, they can bring data into the conversation to discuss where it comes from, how to integrate the source system, and how to build out the data product.

There’s a lot more great information and insights that Omar has to offer. Make sure you check out the full episode and keep up with Omar and the team at Roche.

Listen to the episode here:

🚨Subscribe to Hashmap on Tap on Spotify, Apple Podcasts, & Google Podcasts.🚨

Related Resources

Ready to Accelerate Your Digital Transformation?

At Hashmap, an NTT DATA Company, we work with our clients to build better, together. We are partnering with companies across a diverse range of industries to solve the toughest data challenges — we can help you shorten time to value!

We offer a range of enablement workshops and assessment services, data modernization and migration services, and consulting service packages for building new data products as part of our service offerings. We would be glad to work through your specific requirements. Connect with us here.

--

--