Calculating the CSI (Customer Satisfaction Index)

[ISOMan200]

hi,

The attached table shows a method to calculate the CSI (Customer Satisfaction Index). The original survey was based on 3-point Likert scale: satisfied, neutral and dissatisfied. The results were coded as 1, 0 and -1 respectively.
I've assigned weights for the factors. Speed = 40%, Quality = 40% and Courtesy = 20%. Is this method accurate?
Is there other way to assign weight based on statistical procedure? The survey did not include overall score question nor a questions for importance rating.

Thanks for your help

 

[Ajit Basrur]

hi,

The attached table shows a method to calculate the CSI (Customer Satisfaction Index). The original survey was based on 3-point Likert scale: satisfied, neutral and dissatisfied. The results were coded as 1, 0 and -1 respectively.
I've assigned weights for the factors. Speed = 40%, Quality = 40% and Courtesy = 20%. Is this method accurate?
Is there other way to assign weight based on statistical procedure? The survey did not include overall score question nor a questions for importance rating.

Thanks for your help

Welcome to the Cove

Your calculation is correct. You may also check one of my earlier post - Customer Satisfaction - How to calculate CSI (Customer Satisfaction Index) for additional information.

 

[harry]

hi,

...............I've assigned weights for the factors. Speed = 40%, Quality = 40% and Courtesy = 20%. Is this method accurate? .......................

1. The fact that you had unilaterally assigned weight age means the results of the survey will be skewed towards your preference or decision.

2. Technically, you should carry out an initial survey to find out which dimensions (or survey parameter) are important to the customer and their preference - from which you should get a less biased weight age.

 

[ISOMan200]

Dear Ajit Basrur and harry,

Thank you very much for your responses.

It is true that the result may be skewed based on the weight given. for example with weights of 35%, 35% and 30% a different CSI will be obtained. There was a debate on relying on the weight given by the customers and some prefer derived importance approach. I’ve reviewed some literature discussions on derived importance but – to my knowledge- computing derived importance will required results on a question for overall satisfaction which is lacking in this survey. I’ve tried to find a statistical way to give weighs to the measured factors to eliminate any bias in the final results, but did not succeed. I proposed the weight here in a tradeoff with using un-weighted average…anyone know something about standardized weight? Or if it can help here?

Thanks,
Khalid

 

[Ajit Basrur]

Khalid,

Fixing ratios for each of the categories also depends on type of your industry and your overall Organization objectives.

To start with, do not get bogged down with statistically correct ratios

 

[Steve Prevette]

When analyzing surveys, I use the technique that was taught by Glenn Lindsay at the Naval Postgraduate School. It has several advantages over assuming numerical quantities for each response category - the analysis determines the score values for the categories. It also derives standard deviations for each response to each question.

http: //www. efcog .org /wg/esh_es/Statistical_Process_Control/docs/Survey_Analysis.pdf - DEAD 404 LINK UNLINKED

 

[Brunetta]

When analyzing surveys, I use the technique that was taught by Glenn Lindsay at the Naval Postgraduate School. It has several advantages over assuming numerical quantities for each response category - the analysis determines the score values for the categories. It also derives standard deviations for each response to each question.

In the article it references an excel spreadsheet that accomplishes the operations described, would it be possible to provide it?

 

[Steve Prevette]

In the article it references an excel spreadsheet that accomplishes the operations described, would it be possible to provide it?

See attached.

 

[ISOMan200]

Thanks for the useful contribution. But the statistics looks scary . I’ll try to implement it. However, a colleague calculated the CSI by simply averaging the scores. In my attached sheet this will provide CSI of 64.5% while the weighted average shoots up to 68.8% ....which one is most accurate?

 

[Steve Prevette]

Thanks for the useful contribution. But the statistics looks scary . I’ll try to implement it. However, a colleague calculated the CSI by simply averaging the scores. In my attached sheet this will provide CSI of 64.5% while the weighted average shoots up to 68.8% ....which one is most accurate?

They are both perfectly "accurate" against the operational definitions used to calculate each. The question is - which operational definition is better? One check would be to calculate the standard deviation of each score and see if they are significantly different or not. I suspect they will not be significantly different.