The success of any Business Intelligence project is contingent upon people, not technology. Analysts and end users must work in concert to ask a concise question, identify the data available to answer that question and, validate interpretation of analytic outputs in context of the business environment. From there, the subject-matter experts (statisticians, data analysts, data miners, etc.) must be allowed the freedom to draw upon their breadth of knowledge and experience to select the best methodology for the job.
I cannot tell you how many times a business unit manager has come up to me and with all of the confidence of a just-learned-to-stand toddler and declared “I need a model!” “Really?” I respond. What type of model? Logistic? Linear? What kind of data do you have for me to work with?, and a plethora of other rather technical questions. My point is that predictive models have been used quite successfully in marketing for many years. In a business environment where “half of the organizations surveyed do not take advantage of analytics to help them target, service, or interact with customers” according to Accenture’s Customer Analytics survey, predictive models have gained the esteem and notoriety akin to Steve Jobs.
Convincing a call center manager, or their manager, that they do not need a model may be challenging. Beyond that, below are two runner-ups to predictive modeling that professionals charged with improving the customer experience tend to have missing from their toolkit:
Discriminant Analysis – Discriminant analysis is used to determine which variables discriminate or separate naturally occurring groups. In a call center environment, we may want to understand what variables discriminate between customers who a) become promoters of the brand, b) become detractors of the brand, or c) remain neutral about the brand. Alternatively, we may want to understand the variables that discriminate between customers who will perceive issue resolution at the end of their call vs. those who do not. Was wait time a driver of outcome? Perhaps the number of times they were transferred or placed on hold during the call?
Cluster Analysis – Customer analysis is used to divide populations (in marketing, our populations are typically customers) into groups or clusters, so all customers in a cluster are similar in some (meaningful) way and more importantly, dissimilar from the customers in other clusters. This methodology can be used to help call center organizations better understand the customer experience. A cluster analysis based on call outcome could reveal (prove) that after three transfers, customers lose faith in the interaction and the organization, defecting to the nearest competitor. A different cluster analysis based on product, could indicate an issue with the level of knowledge of company-approved servicers / repair professionals.
So, as you can see this not a reporting activity, it’s much more than that. But you don’t need Steve Jobs. To be successful, you do want to work with a business intelligence analyst to help guide the model building and application of that information. The exciting part about data can be fully realized when the analytics are strong (and correct!). Just because you can get a model out of an Excel spreadsheet, doesn’t mean that you have accomplished what you need to gain a competitive advantage and do more with less.
- Stop the Freaking Customer Feedback - April 27, 2016
- What is your Contact Center Top Priority? - April 11, 2016
- Nine words to stop using to describe your quality assurance program - March 10, 2016
- What NOT TO DO with your contact center budget - March 9, 2016
- What to aim for with your Contact Center Budget - February 15, 2016
- 3 Strategies for Handling Peak Season Call Volumes - February 15, 2016
- 5 Ways to Show Contact Center Agents Love - February 9, 2016
- Notes on Thriving in Contact Center Performance Webinar - November 16, 2015
- How to thrive in contact center performance - October 29, 2015
- What measurement is best? Net Promoter Score, CSAT, Customer Effort - October 12, 2015