banner image

CASE STUDIES Survey Analysis for a Market Research Consultancy

  • Market research firm required statistical "muscle" to help it improve a client VOC ("Voice of Customer") survey
  • BSA implement technically advanced regression and correlation techniques, in order to consolidate questions, and help the client focus on the questions that were most predictive of NPS scores

The Situation

A prominent market research organization specializes in VOC (voice of customer) analysis, by administering focus groups and surveys. A recent client project arose, which required them to make recommendations on how to refine an existing VOC survey.

The Problem

The research firm's client was in the dental industry, and used an annual survey to determine their overall NPS score. In other words, the client was most interested in their customer's “Likelihood to Recommend” the firm, and “Likelihood to Continue” using the product.

The purpose of the research firm’s project was to improve the survey questionnaire, by doing the following:

  • Establish questions that can be consolidated
  • Establish predictors of likelihood measures
  • Establish recommendations to optimize client's customer survey

Although the market research firm had their own analytical resources, they required specialized statistical analysis, and knew that BSA was one of the few organizations capable of providing such high-end analytical abilities.

The Solution

BSA had access to prior survey results, the questions, and supporting data sources. Based on the requirements of the project, we decided to apply the following steps to the problem:

Consolidate Questions: Explore the relationship between survey responses, in general and for different segments. This helps the market research firm determine which questions can be consolidated, in order to shorten the survey.

Determine NPS Predicters: Find which questions are the ideal predictors of NPS Likelihood measures. This ensures that the marketing research firm emphasizes the most important questions.

While many organizations are knowledgeable in regression, BSA knew that "linear regression" - the method most commonly taught and used in research projects - would be inappropriate, for the following reasons:

  • Linear regression is only used when the underlying variables are continuous. Survey responses are not is not continuous, since on a scale of (0 to 9) there is no possibility of answering "3.5" or "7.8" to a question.
  • Since most respondents do not really understand the difference between a 2 and 3, or a 8 and 9, using simple linear regression will cause problems

Given these challenges, BSA settled on Ordered Logit Regression, in order to understand what distance the ordinal variables were from one category to the other.

The Outcome

BSA recommended several questions that could be consolidated, thus improving the client's survey response rate. Additionally, we were able to recommend questions that should be expounded on, based on their relevance to NPS.

The results of this was a much higher accuracy rate, and a marked increase in survey responsiveness, and usefulness in application for other VOC analysis studies.