Value first, then structure: Lessons from GSK’s data transformation

On structuring to align with business strategy:

Ramakrishnan: I believe that the scale and speed of our transformation requires us to innovate and scale rapidly in each of our business units.  Second, I also think that it requires a very different kind of talent, leadership, and expertise to use data and machine learning algorithms to work with complex biologic or genomic data to accelerate drug discovery, and it requires a different set of skillsets to think about how you can deliver a much more data-driven customer experience in our commercial and our consumer business. 

So, recognizing this, what we’ve done is created multiple chief data officer roles within each of the business units, with leaders who are empowered in each of these areas, and allocate the resources that they need to move at pace in line with their business strategy.  The way we’ve achieved the enterprise lens of this is by creating a federated operating model that identifies what’s core, common, and critical, and then works on those aspects of it together. 

On how GSK started its data transformation:

Ramakrishnan: I think we’ve had a bit of an unconventional starting point to our data transformation.  We did not start with a three-year data strategy and a multimillion-pound investment plan.  We didn’t try to gather and just clean, connect all known data in GSK into a data lake before we started the data transformation, etc.  Instead, what we did was we started with the broadest possible ambition and the smallest possible effort, and then created and confirmed value. 

So, our initial focus was on identifying near-term opportunities that we thought would create massive value, which we could then reinvest in our innovation and growth plan. We set up with an ambition of realizing a billion pounds in value through the use of data.  We didn’t know if we could achieve it; we didn’t know how we would achieve it. 

We then started with a set of big business questions across the enterprise, like, “How can we maximize yield from our vaccines in specific brands?”  It was specific enough, but bold enough and big enough questions. 

We then set up cross-functional data squads to work together and develop new data models and algorithms to answer those questions, using agile approaches that you might have seen used in developing software products in a typical world.  Only this time, we were using it to develop data products.  When these solutions created the intended value, we scaled them rapidly, and we were able to develop and deliver value in sprints like you would with the software products. 

On committing to a data platform strategy:

Ramakrishnan: One of the common issues that organizations run into when they start becoming more data driven, is that they start seeing all kinds of issues with data quality, availability of data, connecting the data.  And soon you realize that wrangling and connecting the data sources across all the silos in your organization just takes up a disproportional amount of time and resources. 

So, by committing to a data platform strategy to power the data backbone behind all the analytics and other data products, you then create this ability to ingest, clean, connect, and democratize data at scale, and then benefit from it on an ongoing basis once you’ve done it once. 

On advice worth sharing:

Ramakrishnan: I would say focus on value first, structure can follow later.  Figure out how to create value first and be agile, that will be my first step. 

And then make sure that you have enough people on your team, in leadership positions and driver’s seats, who really understand what the target state looks like.  I think fumbling your way through incremental changes may get you a bit further, but not far enough for impact. 

And then last, I would say, just really be patient with yourself. And bringing the organization with you is probably the most important part of the journey, not just getting your one project right.