Digging a little Deeper into NPS metrics

Previously I discussed how the NPS metric is flawed by including satisfaction and dissatisfaction within the same metric. I pointed out that the satisfaction scale and the dissatisfaction scale do not have the same metrics/components and hence cannot be combined into a single metric such as the NPS metric. Of course, there may be other metrics that may be able to combine satisfaction and dissatisfaction based on the assumption that they will be represented accurately on the same scale. In order to understand the differences between these two scales, it is necessary to fully describe the parameters of each of the scales (satisfaction and dissatisfaction).

Although both satisfaction and dissatisfaction are often described using a Likert scale, there is no reason to believe that the units on each scale have the same value or meaning. For example, a measure on the satisfaction scale of "satisfied" (second-highest to "very satisfied") may not have the same intensity or value with a value of "dissatisfied" (second-highest to "very dissatisfied") on the dissatisfaction scale.  From another perspective, one might want to believe that the amount of customer involvement of moving from "satisfied" to "very satisfied" is equivalent to the amount of customer intention of moving from "dissatisfied" to "very dissatisfied".  Of course, this movement can be expected when we are using ordinal scales but that does not transfer to equality.  Remember that ordinal scales only indicate order but does not include a metric of distance along the scale.  Thus, the movement from one value either up or down on the scale does not require that the distance be the same.  Likewise, movement on a Likert scale for satisfaction does not require the same value for movement on the dissatisfaction scale.  Although the scales are numbered sequentially, there is no requirement for the linear relationship of the ordinal scale to accurately reflect equivalent linear values for both the satisfaction and the dissatisfaction scales. 

If in fact, decreasing satisfaction does not generally lead to dissatisfaction, then a better terminal value for least satisfied would more likely be the customer is "indifferent".  Hence a more accurate satisfaction scale would follow the sequence very satisfied, satisfied, slightly satisfied, and indifferent.  A similar scale for dissatisfaction would follow the same sequence of very dissatisfied, dissatisfied, slightly dissatisfied, and indifferent.  I am using the term "indifference" to describe in non-psychological terms that point in the customer's attitude where the customer stops caring. In this zone of "indifference".  At this point, the customer is no longer connected with the evaluation process and has become indifferent. 

As an example, I once had a consulting relationship with the company that had very high satisfaction scores and believed that their customer relationships were all strong. However, once we analyzed all the customer relationships we found that a relationship that we thought had no meaning was in fact very important. It turns out that one of the company’s largest customers had been put on credit hold because of some delay in the paperwork. Even though the customer represented the largest account, personnel in accounts receivable were trained only to collect accounts. There’s no customer training included for personnel in the Accounts Receivable department. Although the large customer did not change the supplier, it did introduce some significant stress to the company management.  The satisfiers were well taken care of and the areas of dissatisfaction that were measured were small. Hence, the NPS score was very high.  The simple answer might be that the company did not include the Accounts Receivable department as part of its metric. In that sense, a better NPS score may have been computed. However, the values of satisfaction and dissatisfaction on the metric scale would not have given had a significant cause for concern since it would not have been included in the calculated NPS metric. In this case, it almost led to the loss or at least a diminished relationship with a very large customer.

The bottom line is that using a single Likert scale for satisfaction and dissatisfaction is attempting to mix two very different parameters. Previously I have indicated that lower levels of satisfaction do not generally lead to values of dissatisfaction. Similarly, using the same logic, reducing levels of dissatisfaction does not generally lead to satisfaction. It is time to create complementary scales for satisfaction and dissatisfaction so that more accurate assessment of the satisfiers and the satisfiers of the customer relationship can be presented and understood.  It is usually reasonable to believe there is no perfect customer relationship. All customer relationships have strengths and weaknesses.   Not only is a time to create complementary scales for satisfaction and dissatisfaction, it is also time to incorporate into the metrics of the customer relationship, multiple factors that more accurately describe the company – customer relationship. Simplicity is no longer the answer.