Correlation Coefficients, the Net Promoter Score and Other Simplistic 'One Number' Approaches
I recall a study that found that C-level management spend on average just 10 minutes on a decision. Not surprisingly, bulky marketing research reports are considered for this audience as the antithesis of the expected, few key indicators.
The danger seems to come when those managers reporting to the executive level demand similar levels of brevity. Surely those charged with the implementation of the strategy require sufficient levels of granularity to identify what must be done? While executives necessarily crave simplicity, alas management must continue to reside in a level of detail sufficient to actually direct change.
Take the Net Promoter Score (Reichheld, F, December, 2003, The One Number You Need to Grow, Harvard Business Review) for example. Assume for a moment, as Reichheld contends, that it is the ‘one number’ – what do you do with it? How do you fix it? Where is the best place to start? What are its drivers? The Net Promoter Score (NPS) might be fine for executives to demand but for managers, as envisaged by Reichheld, it fails to provide sufficient diagnostics to be meaningful.
There is no ‘one number’
The faddish adoption of the NPS has confirmed management’s unyielding desire for its information to come free from complexity. That desire is not confined to just the NPS. It seems increasingly management is replacing derived models with simple correlation coefficients. Indeed, in some quarters of management the univariate correlation coefficient seems to be becoming as widely used now as it was at the height of its popularity, prior to being largely replaced by multivariate analysis.
It was decades ago that marketing scientists understood that virtually nothing is driven by only one explanatory variable, yet this is what is assumed when using a simple correlation coefficient. This is, after all, the equivalent of the regression coefficient in a simple linear model.
What then has led management to being prepared to trade off so much valuable information for the sake of handling a single number? With technology improvements businesses have been inundated with massive amounts of captured data. In the rush to find new and innovative ways of analysing the data, it seems that there has failed to be a commensurate focus on developing the corresponding explanation of the findings of these techniques. The more sophisticated the technique, often the more demanding it is to explain the findings to management in a way that is meaningful to their context.
This simplistic approach, where only one explanatory variable is considered, takes marketing research back 20 years or more. In business where complexity is the norm, examining how a single factor influences another, in complete isolation, ignoring all interaction influences, is breathtakingly naïve. Correlation coefficients, after all, only consider pair-wise combinations, not any interrelationships that may exist among many potential drivers.
Translating findings
The primary objective of the researcher should be to synthesise and reduce the research findings to their simplest, most salient components. These can then be presented to management with the ability to readily drill down to greater detail wherever and whenever required.
A failing to translate the findings in this way to ‘management speak’ has been touted as the primary reason that something like the Net Promoter Score or indeed, a correlation coefficient is gaining popularity amongst management. But again, these simply provide another overly simplistic ‘one number’ approach to marketing research and do not address the need for actionable, holistic conclusions that are communicated to management with clarity and simplicity.
It’s understandable that C-level executives demand brevity but nevertheless it remains a necessity for the management reporting to those executives to continue to reside in a level of detail sufficient to support decisions that actually bring about change. In this context, little things like a correlation coefficient or the NPS do not mean a lot when it comes to implementation.
