Chewy

Over the past five years, I have been asked a lot about unlocking the value of data science and big data programs for a variety of applications in finance and supply chains. This question became even more important as economic growth slowed down (with COVID in the middle as an unpredictable, idiosyncratic event).

For companies wishing to make the most of their data science programs, there are two critical aspects you need to keep in mind related to your data science programs:

(1) The functional component divided between operational and strategic activities

(2) The technical component is divided between infrastructure and science.

Operational vs. Strategic'

Where and when should I invest?

Functional areas for data science programs are often categorized into pure operational activities or strategic activities. Though they are not necessarily mutually exclusive, they often confront each other when it comes time to assign and fund initiatives for the year to come.

Data science activities revolving around operations are necessary to conduct day-to-day business. They involve building automated decision systems along with the proper calibration of inputs for these systems. In supply chains, for instance, such systems can be forecasting systems relying on machine learning algorithms or Inventory management systems relying on operations research techniques.

Data science activities revolving around strategy are used either to identify new opportunities from the data or potential risks that need to be mitigated. A few examples of applications include pricing strategy, long-term scenario planning, or product lifecycle strategy.

Chasing Growth Under Economic Uncertainty With Data Science As A Driver

When deciding upon funding your data science programs, look where you are in the economic cycle. What I define as an economic cycle is not solely based on real GDP growth but the state of consumer spending, labor market, inflation, and how easy it is to access money through debt/credit for businesses to fund innovation and for households to fund consumption.

 Keep in mind that your ROI profile on infrastructure and science capabilities will vary as a function of time/ sophistication and how mature your organization is 

When at the trough of the economic cycle, leadership traditionally emphasizes operational considerations at the expense of strategic considerations. This necessity is mainly driven by the maximization of shareholders' short-term value ‘one year ahead.’ I want to argue that this strategy will yield suboptimal results in

Keep in mind that your ROI profile on infrastructure and science capabilities will vary as a function of time/ sophistication and how mature your organization is9

the sense that it will hurt your long-term growth prospects; let me explain. Most important decisions are often made in a procyclical manner, which means they are inherently made in reaction to the current state of the economy to optimize for short-term benefits. Optimally you want to identify new opportunities and risks ahead of time in a counter-cyclical manner (preferably at the peak of the economic cycle or slightly on the downside at the latest) in order to give you a chance to implement relevant action plans on time. You also want to understand who your customers are and be able to anticipate their reactions at different stages of the economic cycle. This is so you can take full advantage of the upside of the economic cycle and not miss out on the upside momentum.

Science vs. Infrastructure

Where is my ROI?

I will start by showing you my view of the ROI profile as a function of time/sophistication (I assume that as time passes, they become more sophisticated) from the time you make an investment.

When I speak about infrastructure, I am speaking of the degree of sophistication of your data science pipeline, which includes your data warehousing solution, your ETL tools, and your software engineering capabilities to facilitate the development and deployment of your science at scale (whether it is MLOps capabilities, simulation capabilities or real-time optimization). I will leave the debate on cloud vs. on-premise aside for a future article. When you invest in data science infrastructure, it is sometimes hard to see the benefits immediately. You start with your data infrastructure; then you move to improve your modeling capabilities, and finally, your model orchestration along with the performance monitoring process. The value of your infrastructure is more obvious towards the end when you can translate it into savings on your SG&A due to end-to-end automation of complex modeling tasks. If done right from the beginning, you can minimize technical debt and create actionable value for the future as your infrastructure can scale to other business applications.

When I speak about science, it is about the degree of sophistication of your analytics capabilities. It starts with simple data visualization or statistics and goes all the way to custom algorithms developed specifically for your use cases. Science is also about your people. There are multiple flavors of data scientists, including (but not limited to): machine learning scientists, experts in predictive modeling, operation research scientists, an expert in optimization, and quantitative economists who usually sit at the junction of these fields. Quantitative economists can be used to tackle both operational and strategic problems because they have a combination of skills in causal inference, experimental, and statistics. They can estimate causal parameters used as inputs in your decision systems or support the identification of risks and opportunities using custom econometric modeling adapted to your specific problems and data. Results from science are almost immediately quantifiable, particularly if your organization starts at a low level of data science maturity. However, as the complexity of your science increases, the ROI on your science might not be linear as a function of time/sophistication. It might plateau as you keep exploring alternative science methods but will definitely bounce back as you invest in custom capabilities specific to your business and geared towards understanding your customers.

Conclusion

Chasing growth in uncertain economic conditions requires an understanding of the economic cycle and who your customers are so you can anticipate how they will react at each stage of the economic cycle. My recommendation is to always keep some funding for your strategic data science capabilities to keep the innovation going so you can make the most of the upward phase of an economic cycle.

Keep in mind that your ROI profile on infrastructure and science capabilities will vary as a function of time/sophistication and how mature your organization is. The more mature your organization, the more challenging it will be to generate the incremental gain you are looking for, but it is completely worth it in the end as long as you focus on understanding your customers’ behavior and adapt.