Maximizing Survey Information (An Overview)
Yearly satisfaction, benchmark and voice of the customer (VOC) surveys are expensive undertakings. Most of the budget usually goes to acquiring a representative sample and managing the design of the survey. The remainder frequently goes towards simple analyses involving simple bar charts and tables.
Rarely are statistical analyses, with their ability to analyze huge amounts of information from all parts of a large survey, included. These more advanced analyses can easily double or triple the actionable insights from an important survey while only increasing costs by a fraction.
As the diagram below shows, these large surveys generally contain multiple sections.
By combining information across multiple sections, M Squared Group has helped our clients answer additional questions without the expense of additional surveys.
Possible additional areas of analysis include:
* Using decision trees on demographic variables to identify segments with a propensity for a particular behavior to guide media buys
* Combining factor and correlation analysis to identify structure and representative statements in an attitude or usage study.
* Using cluster analysis to identify unique groups of customers that warrant separate treatment and measurement
* Combining multiple related questions into a scale that is more predictive of future behavior than any given single question
Approaches to Survey Analysis
Simple:
* Intelligent cross tabulations (frequently based on recoded variables)
* Simple (bivariate) correlation analysis
More complex:
* A wide variety of multivariate methods are available depending on the details of the problem and data. The most commonly used methods include: Decision Trees, Multiple Regression, Cluster Analysis, Factor Analysis, and Path Analysis (Structural Equation Modeling).
Common Mistakes
1. Grouping together respondents who have differing needs or experiences. We have frequently seen this with top-two box analysis. There are times when the top box and the second box represent two distinct types of respondents.
2. Taking results too literally. It’s important to consider trade-offs across attributes and discounting when applying survey results to the real world.
3. Looking only at overall averages. Frequently there are insights to be found simply by making a meaningful grouping variable and performing cross-tabulations.
4. Prioritizing actions based largely on a statistical analysis without enough consideration of business needs. Research needs to be actionable and technically sound.
Why it can take time
* Smaller sample sizes require stricter adherence to the assumptions for parametric statistical analyses.
* On a strictly exploratory analysis, the number of possible cross-tabulations goes up exponentially with the number of survey questions.
What is the M Squared seasoning?
* Experience generating additional insights from survey data that already exists.
* Statisticians that can talk to marketers. This ensures that our clients’ critical business questions are answered utilizing robust methodologies.
* We present results as part of a straightforward story for your marketing team. Our combined teams of marketers and statisticians will never complicate a presentation with an equation.