Desktop productivity for business analysts and programmers

Correlation analysis between customer survey data and attrition

Reply
Learner
Posts: 1

Correlation analysis between customer survey data and attrition

I have customer feedback survey data and attrition data. I'm trying to understand if low customer service scores are correlated with higher rates of attrition. It's been a while since I've had the opporunity to do this kind of analysis, and I need some help getting started. Any recommendations on SAS procedures to look into? 

 

A couple of notes about the data: 

- Customer scores are discrete from 1-10 (1=not at all satisfied to 10=completely satisfied, with no qualitative "value" being assigned to ratings 2-9)

- The data I have gives the average monthly customer satisfaction score by retail location; due to changes in the survey methods, I'm looking at Mar 2016 - Nov 2016. Each month's score is reported as YTD, so April's score is the sample that was collected from both March & April; May's score is March & April & May, etc. Initially I was looking at doing a timeseries analysis, but because the scores "build" on each other each month, I think it makes the most sense to use just the most recent month... but feel free to disagree. 

- The attrition value is net number of customers gained/lost, so it can be positive or negative

- I'm interpreting this data at the store level

Frequent Contributor
Posts: 135

Re: Correlation analysis between customer survey data and attrition

I would recommend proc glmselect, which is at the heart of the predictive regression models task in SAS Studio. The distribution of your dependent, response or target variable is a convolution of several processes that is likely to have some strange distribution that would not meet the usual gaussian/normal residual assumption required for ordinary least squares regression provided by proc reg. If you have enough (aggregated) observations to be able to afford a sampling split into training and validation data that would yield the most reliable inferences, otherwise use AICC as the selection criteria, and allow the store hierarchy to be split up for model selection.

Ask a Question
Discussion stats
  • 1 reply
  • 138 views
  • 0 likes
  • 2 in conversation