Analytics & Big Data
Mar 11, 2016

Business analytics can learn a lot from other science

March 11, 2016 - Companies are in the business of supply and demand, a topic at the heart of the field of Economics. Businesses create media and marketing campaigns to drive consumer response, a topic at the heart of Behavioral Psychology. Businesses attempt to optimize production, supply chain, and shelf space, all of which are at the heart of Operations Research. Businesses engage in a variety of tasks, which are measured, tracked, analyzed, and optimized using a variety of tools. However, the analytic tools businesses use are not always as well established as the corresponding sciences to which they relate, such as Economics, Behavioral Psychology, Operations Research, or other fields. Instead, these analytics are often simplified and adapted versions from these fields, custom created around a business problem, and thus do not necessarily benefit from the entire field of these other sciences.

All businesses, for example, struggle with optimizing product pricing. They use tools such as marketing mix models from Econometrics to estimate price elasticity, test and control studies from Statistics to find which price point resonates, or price gap analysis from Management Science. The key here is that these analytics do not fully benefit from insights from the wider field to which they relate, yet there is learning these other sciences offer. The field of Physics, for instance, offers a solution to this price optimization question. There is an important law in Physics called the “Inverse Square Law” which says the intensity of the force between two objects is proportional to the inverse of the square of their distance. For instance, the gravitational force between two planets, the attraction or repulsion force between two magnets, or the decibel volume heard by a person a distance away from a speaker all adhere to this law wherein an increase in their distances from one another decreases the intensity in proportion to the square of their distance.

The ubiquitous nature of that law in the world of Physics can be generalized to the price optimization in business. Consider the price difference between two nearby gas stations. Given the closeness to one another, a very small price advantage is sufficient for consumers to switch given how easy it is for customers to travel from one gas station to the other. However, for customers to travel to a far away gas station requires the price advantage be considerably larger. If we examine the gap between the market share of products as a function of the distances to their competitors, we can find the differences in shares are a function of the same inverse square law. As such, we can determine the optimal price of products based on distances to competitors with similar products by using the square law models from Physics.

Time series analytics in business are often built with statistical procedures like regression. Despite all the innovations in regression modeling, such as lag functions, diminishing returns functions, and synergy functions, regression assumes that the coefficient of the forces are constant over the entire time period of the data. But is the effect of TV (the TV coefficient) really constant over three years? Is the elasticity of price (the Price coefficient) really constant over the three years? Fields like Econometrics and Biophysics have established time varying parameter based models. Business analysts could benefit by testing these fleshed out time varying methodologies from these other fields.

Customer lifetime value (LTV) is an important metric companies estimate and track to prioritize customers, create customized messages and campaigns, and plan accordingly. The traditional approach to estimating LTV in the business world considers the revenue cycles from customers, estimating the lengths of time customers will be engaged, customer switching propensities, and various estimates of responses to messaging. In other fields, like Chemistry, various longevity estimations, models, and tools, such as radiocarbon dating, provide frameworks for the addressing the same question in a completely different way. Business analysts could benefit tremendously by investigating the portability of these other well-studied solutions and methodologies in these other fields to customer LTV.

Media planners regularly discuss the optimal number of Gross Rating Points, which is a combination of reach (how many people see a set of advertisements) and frequency (how many times they see the set of advertisements). Keeping the Gross Rating Points (GRPs) as low as possible is key given the high cost of media. However if the GRPs are too low, then there may be an insufficient volume of advertising to drive sales. GRP optimization is often addressed with test and control studies, media mix modeling, and primary research. However, analyses of this kind in the fields of Behavioral Psychology and Psychometrics have led to approaches to estimating aided and unaided recall as a function of the nature of exposures, length of the exposure, complexity of the messaging, and other attributes. Media analysts could benefit by investigating these other kinds of models from these other fields.

Improvements and advanced in business analyses are not limited to advances in big data, faster technologies, newer platforms, and improvements in visualization. Advances are likely to also come when analysts look at problems holistically, through the lenses of all scientific disciplines, and by being open to entirely new ways to look at problems. One is unlikely to find an entirely new, innovative approach to improving analytics without widening the search space, which includes the vast array of other scientific disciplines involved in the same search.

About the author

David Hauser, PhD

David Hauser, PhD

Chief Science Officer, CPG and Specialty Analytics

Follow David Hauser, PhD on LinkedIn