Brian's 3rd point is spot on. From a purely best-practice perspective, an 11-point scale is simply not the way to go. It skews responses and adds no utility over 5 or 7-point alternatives. This is particularly true when the scale ends up being artificially trichotomized on the backend; whether intentionally as NPS dictates or unintentionally as another poster noted.
That being said, NPS is something that the CX community has embraced and doesn't seem to be going anywhere. Because of this, should you choose to use an 11-point scale for the entirety of your survey, there are benefits in terms of reducing cognitive load on your respondents (i.e.- not switching between scales). On the flip side, using a 5 or 7-point scale for all non-NPS questions would insure you are taking steps to adhere to best practice.
When analysis is concerned, it really depends on what you are looking to do. Methods like SEM are robust to scale differences whereas less sophisticated analysis may require you to standardize scores (e.g., Z-score transformation).
------------------------------
Kyle Groff, Ph.D.
Principal Consultant, CX
Qualtrics
Salt Lake City UT
kyleg@qualtrics.com
Original Message:
Sent: 03-03-2016 07:22
From: Karl Sharicz
Subject: Applying the 0-10 NPS scale consistently on surveys?
Fred,
When I started out developing surveys it was all done using a 5-choice scale. Notice I didn't say 5-point because we used words describing each of the five levels of satisfaction and not numbers to define our scale and every satisfaction question we asked in the survey used that scale.
Even when it came to the "recommend" question we used the same scale but words were built around "likelihood" to recommend. Later on when we all got enamored with NPS and wanted to compare ours to that of other companies (which is problematic to impossible unless you are conducting a benchmark study) we switched over the Reichheld scale, not only for the "recommend" question but for all of the survey questions. The anchor words were only placed at the extreme ends of the scale--Extremely Likely and Extremely Unlikely. During the transition year, we used both the 5-choice and the Reichheld scale in measuring the NPS and both scales produced the same NPS within a point so we felt we validated the switch. We never could realistically compare our NPS to other companies for the obvious reasons.
Using the Reichheld scale for all survey questions was done for consistency, as was pointed out. After some years of using that scale, I've come around to believe that the 5-choice model was and is the better option. It's easier as words are used rather than numbers and it reduces the number of choices the customer has to face and, as I said, you still can calculate an NPS if that excites you and honesely it doesn't excite me much these days.
The bottom line here I feel is to use whatever scale you like for your organization but just be consistent and by all means make things easy for your customer by keeping surveys short. I have personally abandoned more surveys coming at me these days because of length. More data isn't better than better data.
Karl
------------------------------
Karl Sharicz, CCXP
Founder | Principal
CX Partners, LLC
Quincy, MA 02169
508 989-7379