In February, Forrester Research released its report titled “The Forrester Wave™: Insights Service Providers, Q1 2017 | Leaders Emerge in a Nascent Insights Services Market” written by Jennifer Belissent, PhD and Elizabeth Cullen. Forrester included Market Force Information® as a domain-specific insights provider in their consolidated vendor landscape. A graphic from this report recognises Market Force in an ecosystem of industry-specific, domain-specific, and broad insights service provider. We are very proud to be recognised for our innovations in linking customer experience metrics to financial ROI, and a joint video with Forrester presents case studies giving examples of how we do that.

I am one of the data scientists on the Market Force analytics team driving insights innovations in the CX space. We’ve been addressing a thorny modelling issue related to collinearity. Collinearity is a statistical phenomenon in which two (or more) explanatory variables are highly correlated, meaning that one can predict the other. For example, in the retail space, the predictors—associate helpfulness and associate friendliness—can predict customer satisfaction, but they can also predict each other. This creates a situation where the behaviors share explanatory power of customer satisfaction, creating redundancy and causing significant issues with predictors, including their perceived impact on customer satisfaction and the deterioration of integrity within the modeling process.

The presence of collinearity in customer experience data can negatively affect the quality of predictive models and may lead to incorrect or incomplete insights. That means that when those insights are converted to business initiatives, there is potential to focus on the wrong things. The hard work put in by managers and operators will not have the predicted impact. This sounds dismal for multi-location businesses, but innovations in statistics and machine learning have brought predictive modelling techniques to the market that mitigate the impact of collinearity.

Enter LASSO (least absolute shrinkage and selection operator) and ridge regressions. These techniques are machine learning linear regression models that use sophisticated computation analysis to control collinearity and produce the most predictive models possible. The two modelling techniques, each presenting its own unique solution to collinearity, work by computing tens of thousands of algorithmic operations to determine the most predictive combination of coefficients, or weighting, to be applied to each predictor. Both modelling techniques remove collinearity through shrinking the weighting of variables and eliminating the redundant explanatory power of collinear predictors.

In preliminary tests using these machine-learning techniques, Market Force Information has discovered that model predictiveness increases by as much as 10%. The integration of these sophisticated models has increased the ability to generalise findings and predict future outcomes. This results in clear and productive direction to our clients on where they should focus to improve the customer experience, increase revenue, and reduce costs.

If you would like to discuss how we can help you leverage your CX data to link to a financial ROI, please schedule a briefing or call us at +44 1908 328 008 and we will be glad to discuss!

Schedule a Briefing

Aaron Zelmanow is Senior Data Scientist at Market Force Information with a Master's in Statistics and a background in business operations and finance. Specializing in the retail industry, Aaron provides analytics and actionable insights for clients to maximize customer satisfaction and increase their bottom-line revenue.