Usual tools and techniques like cross-tabulations, bar charts and finding mean differences between groups are too simplistic to enable us to derive the most valuable information from collected survey data. It is very important to retrieve essential and insightful information from collected data to take right decisions.
We can use following four types of advanced analysis to gain important insights that we might miss using usual basic methods.
By using these methods we can delve deeper into data to increase our understanding of survey responses and respondents, create better measures of important concepts, and make more accurate predictions about behaviors and attitudes of consumer or respondents.
Here is a brief overview of the four techniques:
1) Cluster Analysis: It is used to discover similar groups, or segments, of respondents. Segmentation enables us to focus sales and marketing efforts on defined groups. We can also use subgroups in analyses, to be more sensitive to find differences between respondents.
2) Factor and Reliability Analysis : It enable us to combine several questions into a more valid and reliable measure of an important concept. It also helps us in isolating survey questions that may be redundant or unnecessary.
3) Regression Analysis: It is used to create predictive behavior models that include many predictor variables simultaneously. Regression analysis enables us to identify the best predictors, so we can focus on them in future actions.
These methods helps us in analyzing many survey questions simultaneously, in order to cluster respondents, group questions, and make predictions with greater accuracy.
Segmenting survey respondents with cluster analysis:
Cluster analysis helps in group respondents with similar preferences, behaviors, or characteristics into clusters, or segments. Through segmentation, we get greater understanding of important similarities and differences between respondents.
Clustering works on logic of creating groups based on their proximity to, or distance from, each other. Respondents within a cluster, therefore, are relatively homogenous.
Cluster analysis requires researcher to:
- Check the number of respondents in each cluster, as clusters of only a few respondents are not very useful
- Assess whether the clusters make sense, and whether their characteristics are easy to understand and describe
- Validate the clusters by analyzing how they relate to other variables
For example we take a example of cluster analysis which uses the Two-Step Cluster procedure in SPSS, which incorporates statistical criteria to determine the optimal number of clusters.
We can create clusters on the basis of our requirement or parameters answered by the respondents.
The first output from the Two-Step method tells us how many clusters were found, and how many respondents are in each cluster.
Additional information from the Two-Step method enables us to understand what type of customer each cluster represents. From below table for example, customers in Cluster 1 travel more for work than customers in the other two clusters, and that they frequently stay at hotels. Measured by length of customer relationship, how- ever, customers in Cluster 1 are not the most loyal.
Now that we have identified the clusters of respondents, we can use them in analyses and reports. For example, to see how overall customer satisfaction relates to cluster membership, we use a clustered bar chart to show the results.
We can infer from the Bar chart that customers belonging to cluster 2 are most neutral towards satisfaction.