Let’s sketch some broad strokes to delineate these roles. Though exceptions abound – and every organization is different – this framework might help clarify some important overall distinctions between these roles. 

Analyst: the traditional Analyst role primarily engages in standard reporting and ad-hoc analysis to probe deeper into business performance. They certainly support decision-making processes (or what we used to think of a decision-support) and they are usually directly responsible for monitoring important success metrics and KPIs, though the analytical toolset available to them may be somewhat limited.  In my experience, Analysts typically specialize in certain product areas/divisions or specialized domains like Web Analytics, Content Measurement, Paid Media, or CRM, for example, using business intelligence (BI) tools in combination with spreadsheets and pivot tables to help better contextualize data and provide drill-down into summary statistics, trends, etc. They tackle various questions as they emerge from the business but there is no expectation to develop sophisticated data models or “look around corners” with machine learning and predictive insights. But, that said, many attributes of an excellent Analyst — such as strong business acumen, analytical thinking, etc. — are very similar to that of a Decision Scientist.  And all Decision Scientists must be good Analysts!

Data Scientist: a Data Scientist usually adopts a much more specialized role than the Analyst. For instance, they might develop a machine learning model tailored for a very specific purpose or application within a company, such as predicting customer churn or the next-best product to sell to a customer.  In these scenarios, use case requirement documents, which tend to be very technical and prescriptive, describe the functional specifications of the end user to the Data Scientist. These applications of data science are crucial for supporting strategic decision-making in certain contexts, yet the critical analysis often falls to others. And, almost always, these sophisticated insights fall short of presenting essential “what-if” scenarios to decision-makers, including quantifying the risk/reward trade-offs associated with different choices. This is because decision-support is ultimately corollary to the predictive insights themselves, not at the core of their outputs. More effort is required.

Decision Scientist: is a blending of the Analyst and the Data Scientist roles because BI is too limited in scope and predictive analytics by itself doesn’t generate the necessary value. And, unlike the other two roles, the Decision Scientist is fully dedicated to supporting strategic decision-making. They possess a deep technical skillset in the areas of data mining, statistical modelling, and predictive analytics — just like the Data Scientist — but typically don’t maintain models in production (such as those predicting customer churn or next-best product recommendations). Rather, for the Decision Scientist, predictions serve as inputs to expected-value frameworks and scenario analyses that ultimately drive the decision-making process. And, importantly, Decision Scientists are never just answering questions with data; they are obsessively focused on understanding and optimizing incrementality and lift to drive return on marketing investment (ROMI). They need to understand goals, anticipate how decisions are made, and be able to take conflicting evidence into account when providing recommendations. 

Finally, in terms of tools, Data Scientists and Decision Scientist alike both tend to rely heavily on “data lakes”, or similar platforms, to integrate data from multiple sources for model building. However, Decision Scientists may not need to be directly supported with robust MLOps processes in the same way because, generally speaking, they are not operationalizing models into productive systems as part of their typical day-to-day.