All Analytics community members generally agree that the NPS, a much-used customer preference metric, as a measure unto itself is of increasingly limited value. What most seem to be unsure of, however, is what to do about that. One commenter, for example, wrote that the NPS has indeed "gone the way of the dodo." However, he continued by saying he was not sure what the best replacement might be. "Things happen quickly in social media, and many are influenced by what happens to their friends and what they have to say. I think these factors make it really hard to predict behavior. I think overall loyalty is decreasing across the board."
Perhaps, then, the discussion ought not to be about retiring the NPS, but of supplementing it, as well as survey results, with more advanced analytics. Randy Collica, a senior solutions architect supporting the communications, entertainment/media, and travel industries at SAS joined the conversation with this suggestion: "… whatever metric is being used, understand the limitations, but use the analytics to the best and use this along with domain expertise to make the best possible decisions for the organization."
In a follow-on email with All Analytics, he stressed the importance of moving beyond just the NPS. "There really is so much that organizations can do than just trend the past, and we're attempting to move them in the right direction."
Painting the big customer picture is best done by combining survey results with a promoter score and text analytics, along with predictive modeling. In doing so, companies can gain a much better understanding of what their customers think about their products and services -- and of the monetary impact of those attitudes. This is the crux of the matter, after all. As one community member commented, "The real question is how does the score translate into sales?"
For more on how this would work, Collica pointed to the SAS whitepaper, "From Customer Risk to Corporate Strategy: Using Text Analysis and Predictive Modeling to Improve Promoter Scores." The whitepaper suggests:
To quantify the impacts to the entire customer population, organizations can extend their survey results by using the promoter scores to predict the behavior of customers who were not surveyed. One method of doing this is to gather all of the customer contact history and combine this unstructured text with the survey promoter score responses -- as well as with the structured data the organization already has. Unstructured text can include verbatim comments from the promoter survey, call center notes, and comments from sales or technical support representatives.
The analysis can actually predict promoter scores, which when combined with customer revenues and other financial metrics, the whitepaper says, can show the relationship of the likelihood to be a promoter (or detractor) with potential revenues. "Then you can use this relationship to develop strategic plans centered on the voice-of-the-customer (VoC). For example: If X customers are moved from passive to promoter status, what is the effect on overall revenue?"
The idea is to perform simulations on data using econometric modeling methods to glean insight into how much value the company can achieve by moving a customer up the promoter score curve. Below you'll see an outline of the methodology. The key, the whitepaper explains, is to turn the information contained in customer call notes, transcripts, and support staff logs or emails, those types of unstructured data sources typically helpful in predicting customer sentiment and likelihood to purchase or recommend, "into something that is structured so that predictive models can be developed and applied to the entire customer population."
I won't go until full detail here, but I do recommend that anybody looking into how to boost the effectiveness of decision-making around NPS dig into the whitepaper. What you'll learn is how to better use the promoter score and survey results to figure out the value in migrating customers up the promoter score curve. And that, in turn, will lead you to a state of "quantified confidence."
Can you see the value in mining text and using predictive modeling to gain more confidence in your customer-related decisions?