Correlation analysis allows us to measure the degree of linear relationship between two variables. As you may recall from STAT 101, the correlation coefficient may take on any value between plus and minus one. The sign of the correlation coefficient (+ or -) defines the direction of the relationship, either positive or negative. A positive correlation means that as the value of one variable increases, the value of the other increases; as one decreases the other decreases. A negative correlation coefficient indicates that as one variable increases, the other decreases; as one decreases the other increases. Alternatively, the absolute value of the correlation coefficient measures the strength of the relationship. A correlation coefficient of r =.8 indicates a strong positive relationship between two measures, whereas a correlation coefficient of r =-.6 indicates a less powerful, albeit strong negative relationship between the two measures.
Figure 1: Web Metrics vs. Financials Correlation Matrix
In our e-commerce case study (see Figure 1), the correlation coefficient measures both the level of magnitude and directional impact of Web metrics (e.g., stickiness, frequency) on the bottom-line financials (e.g., revenue, operating income). Stickiness thus appears to have a significant relationship to revenue with a correlation coefficient of .9 (the value in the cell where stickiness and revenue cross). This makes sense because stickiness is a term used to characterize the attractiveness of a site, section or page, and is measured by the average number of page views, session duration and page depth. Thus stickiness today could be an excellent predictor of revenues tomorrow. It is also important to note that stickiness and page views are highly correlated at a .8 level. This implies that both are measuring similar characteristics, and that stickiness can be used as the KPI (and as a surrogate for page views). Keep in mind that the values above the diagonal mirror those values below the diagonal, and only those values of .5 or higher are shown. Although this is a simplistic example, this approach can be used with all "out-of-box" metrics to determine which metrics are the true KPI metrics. As a final note, it is important to remember that correlation does not always indicate causation. Before causation is validated, the relationship must pass an important criteria - the causal variable must always temporally precede the variable it causes.
Kent Bauer is the managing director, Performance Management Practice at GRT Corporation in Stamford, CT. He has more than 20 years of experience in managing and developing CRM, database marketing, data mining and data warehousing solutions for the financial, information services, healthcare and CPG industries. Bauer has an MBA in Statistics and an APC in Finance from the Stern Graduate School of Business, New York University. A published author and industry speaker, his recent articles and workshops have focused on KPI development, BI visioning and predictive analytics. Please contact Bauer at kent.bauer@grtcorp.com.
Be the first to comment on this post using the section below.