R&D Data Science within the Analytics Operating Model

DISCLAIMER: The views expressed herein are mine and do not necessarily reflect the position of my employer or the the organisations with which I am associated.

Many people have been asking me to delve deeper on the topic of how Pharma R&D Data Science function can be setup or "actualized" as I've put it in a previous post. As mentioned in that post, I would assume that this will become particularly relevant in the Nurture or Mature phases of building a Data Science function. In the current article, I will try to go one level deeper into how this Data Science function would look like - from a organisational, personnel, Data and IT infrastructure as well as a cultural standpoint.

Before I begin, I would like to clarify what I mean by the term "Analytics Operating Model". Operating models are common place within various different business units (say R&D, Commercial, HR etc.) as a set of processes which envision how different functions come together, work cross-functionally and deliver business value. They also delineate roles and responsibilities for every project which is executed using that Operating Model. In the example of R&D business unit, this project could mean the discovery, development and approval of a novel drug. As no one function can deliver on this project alone, business value for each function is measured with the help of metrices such as KPIs and individuals within a function are asked to setup their job priorities in line with those KPIs. In a nutshell, it's a plan on how business units are supposed to operate. Analytics is one of those larger business units under which Data Science function might be potentially located. Analytics is more often an enterprise-level business unit, which means that it works across all the business units. Why is this information at all relevant for the Data Science function, you might ask? The short answer is that Data Science by itself, although incredibly valuable in generating its own ROI (more on that in a subsequent post), only stands to gain when it works together with other "data" functions.

As Data Science derives most of its business value by working closely with Domain and Computational Scientists in the first place (see this post), it makes Data Science practitioners perfectly suited to play a similar central role when working within the Analytics unit. There could be multiple functions in Analytics which would come together when working on specific use-cases. These might be -

• Engineering function which could include people who are Data Engineering, Software Engineering and ML Engineering experts. The function might also include Operational experts who would have skill sets such as DevOps, DataOps and MLOps.
• Business Intelligence and Visualisation function with experts in Dash-boarding and reporting apps who provide descriptive analytics.
• Cloud IT and Infrastructure function which is responsible for provisioning and maintaining Cloud Computing and Data Warehousing resources along with SaaS analytics applications built on top of the cloud infrastructure.
• Data Stewards and Data platform managers along with AI/ML outsourcing and external strategic partnership managers.

I have written about some of these functions already in this previous post from the point of view of making R&D decision making process more data-driven (at least in the first phase). As I envision Data Science function to drive new and high ROI use-cases within R&D, it can create new ways of working jointly within the Analytics unit. Data Science being one of the new functions (along with the MLOps, ML Engineering etc in other functions) might also mean that there are significant differences in the cultures of these different Analytics functions. While Data Science is fully embedded in the scientific decision making process in R&D, the other functions might have different ways to approach and measure business value. For example, exploration and experimentation of novel use-cases which is fundamental to Data science use-case discovery and development might not be fully appreciated by the other functions.

I love the idea of the hub and spoke model of working together - the Data Science function which brings in its unique value proposition could serve as the hub of interactions between the other Analytics functions. This would be crucial if we are to aspire to get from just a "data-driven" organisation to a "data-centric" organisation where Data Science has a say in the R&D decision making process.

In my future posts, I will try to explain on how Data Science creativity, productivity and output can be quantified in terms of ROI. Can the initial process of ideating and coming up with potent use-cases be something that already adds value to the R&D decision making? All this would mean further expanding on the operating model of Data Science function itself. Stay tuned!