In most of my conversations with business leaders who consume output from analytics teams, there is one common theme that emerges – the need for analytics to address business issues. And when I talk to analysts and data scientists, they often mention the business’ inability to apply analytical outputs that can create significant impact to business issues.
There is no doubt that analytics teams build powerful solutions, but then they look for a problem to which it can be applied or, even worse, force the solution onto unrelated business problems. To borrow an analogy from “Apollo13,” the analytics team basically is trying to fit a square peg into a round hole.
The disconnect is because analytics teams (in most cases) see the model outputs, machine algorithms, and data structures as the end objective. But the business wants the analytics teams to provide them answers to business questions, not the output of a mathematical equation.
Here are three simple guiding principles that analytics professionals can apply, and that can act as initial steps toward bridging this gap and help analytics teams deliver actionable insights to drive better business decisions.
The 5 Whys. One of the best consulting professionals I have worked with and who was respected by CXOs of multiple Fortune 500 companies once told me that the first thing he asked when a client came to him with a problem was “Why?” In response to whatever the client said, he would again ask, “Why?”
The 5 Whys is a standard approach that helps contextualize any problem. The solution provider asks the client, “Why?” successively until they reach the root of the problem. Knowing the context can help an analytics team choose the right technique, develop the optimal model construct and, most importantly, develop scenarios and recommendations that tackle the core problem behind the analytics effort.
For example, the analytics team may be asked to perform a price elasticity study.
- Why? The business is planning to increase prices and wants to know where it can affect an increase.
- Why? The business wants to increase profit and, hence, the price increase.
- Why? Because the cost of materials is rising.
Having asked these questions, the analytics team can develop and modify the solution to identify where to increase prices, whether that will result in a profit increase and, if not, determine the right way to improve profit. The analysis could determine that, due to high price sensitivity, a price increase can’t help drive profit. So the analytics team should do a study focused on cost reduction to solve the problem, thus finding the elusive round peg that will fit the round hole.
Blackbox the technique, not the approach. More often than not, I find that analytics teams are blackboxing their approaches against specific solutions – that is, using pre-defined scripts for running specific solutions, rigid workflows, laid-in-stone process steps, etc. This goes against the principle of solving business problems.
Let’s say that you automated a particular model with a certain combination/selection of variables. If the business realities require a different transformation of variables or a different set of variables, then the existing models will fail to take that into account. While the technique – GBM, Bayesian networks, linear regression, etc. – should be married to a solution and can be blackboxed, how the technique would be applied or how the model would be constructed should be customized based on the business problem at hand.
Uncover the biases. We all have multiple cognitive biases. Business teams have biases and beliefs regarding their products, channels, competitors, packaging and shelf spacing, to name a few. Addressing these biases adequately can make the difference between a report that will collect dust and one that is used to make decisions. Therefore, it is important that the analytics team always ask the business what results they expect, what assumptions or biases they hold regarding the analysis, and what hypothesis, if any, they would like to validate. This acts as a great guiding tool, once the results are collated, for assessing which messages will face resistance and which will be readily accepted.
The analytics team then should ensure that it addresses any biases by providing additional information and data points where the message does not align with what the business thinks and highlight the areas where there is alignment. Of course, this does not guarantee that the analytics output will change business beliefs, but it does ensure that the business will look at the report objectively.
There are many more tools from areas as diverse as consulting, semiotics, communication sciences, and psychology that can help analytics professionals increase the impact of their outputs, but the ones listed above are, in my belief, effective and easy to integrate and implement.
Anuj Kumar is Vice President of Analytics Delivery at Fractal Analytics, where he helps consumer goods companies solve marketing and pricing problems through data, analytics and visualization. He is passionate about working with business leaders to identify the “right problem” and then designing the optimal solution to achieve the desired outcome.
Anuj brings over a decade of experience in assisting CXOs in the CPG, Manufacturing and Financial Services industry on decisions regarding strategy, pricing, marketing and business process re-engineering by providing actionable insights. Prior to joining Fractal, Kumar served as Manager – Strategy and Business Development at KPMG. He graduated from Panjab University with a degree in Chemical Engineering and holds an MBA in Strategy and Finance from the Indian Institute of Management in Calcutta.
Subscribe to Data Informed for the latest information and news on big data and analytics for the enterprise.