In the last post, we discussed how to set up a customer satisfaction survey. Today, we’ll discuss how to analyze the results of that survey to identify the attributes that drive satisfaction. Let’s again assume that you’re a restaurateur and you’ve administered a C-Sat survey.
Gather your dependent and independent variables
Your dependent variable – the variable you want to measure – is the answer to the question: “Overall, how satisfied were you with your dining experience tonight, on a scale from 1 to 5?” Assume 5 is “very satisfied” and 1 is “very dissatisfied.”
Your independent variables are derived from the questions about the actual service delivery: “The server greeted my party as we entered the restaurant.” This can be a point scale ranging from “strongly agree” to “strongly disagree,” or it can be a dichotomous variable: 1=”yes” and 0=”no.” Other independent variables can be derived from questions such as: “The manager was visible on the floor,” “the food was of adequate temperature,” and/or “the server checked back with us frequently.”
Use Regression Analysis
The easiest means of identifying satisfaction drivers would be regression analysis. Regression analysis with a few variables can easily be performed in Microsoft Excel.
When you perform regression analysis, Excel (or any other statistical analysis tool) will examine all the data points and generate an equation. This equation will provide coefficients for each of the independent variables as well as values called a t-statistic. Also, the equation will generate an intercept term.
Interpreting the results
Look at the t-statistic. You want to make sure it is significant at either a 90% or (preferably) 95% confidence interval. If the t-statistic is at least 1.645 (absolute value), it will be significant at the 90% level. If the t-statistic is at least 1.96 (absolute value), it is significant at the 95% level. At these levels, you can be confident 90 to 95% of the time that the independent variable is associated with satisfaction.
Also, look at the R-squared statistic – also known as the coefficient of determination. This is a number from 0 to 1, and indicates the percentage of variation in the dependent variable that is explained by changes in the independent variables. You want this number to be as close to 1 as possible, as it would mean 100% of the variation is explained. If you have significant independent variables, but your R-squared is only 55%, that means that 45% of the variation in satisfaction is being explained by changes in factors you are not measuring.
Use your equation to predict satisfaction
Now that you have your equation, see how well it predicts satisfaction changes. Using some satisfaction surveys that were not used to build the model, plug in the scores these respondents gave into the equation and see if the equation estimates an overall satisfaction score that approximates the one the respondent gave. The equation’s predictive quality will be dependent on how close its estimates come to the actual.
Keep in mind…
- While independent variables that are significant are associated with the changes in the dependent variable, that does not mean they caused the change.
- You can model more than one dependent variable – but separately. In this example, we modeled overall satisfaction. In another case, you might want to use “how likely are you to recommend this restaurant to a friend.”
- A high R-squared doesn’t mean your model will be an accurate predictor of satisfaction. Some models with high R-squared predict poorly while some with low R-squares predict quite well.