What is the problem with customer rating?
Dr.Peter Granasy, MBA Program Manager
I lead product development/ process improvement projects on-time, in-budget with right quality in Automotive and Medical devices
Similarity with Employee Satisfaction Survey
Understanding customer behaviour is one of the most significant tasks for artificial intelligence applications. If we know the customer, we recommend products that they will like, which generates extra revenue. This is similar to the?Employee Satisfaction Survey, as both try to measure something subjective. We want to conclude. Should we?
But first, let′s dig deep into the?data analysis perspective.
1.????? In all measurements, we must define the scale with a starting and endpoints. As for temperature, zero is?where the ice melts/ water freezes, and 100 is?where water boils (same pre-defined pressure for both values). When we rate something, we must do the same: set a proper definition of 1 to 10. We can imagine at least three different types of persons. On the Graph, you see the Ideal, the Type 1 never rates anything zero, and the Type 2 has higher expectations/ baseline.
2.????? Additionally, we expect linearity in the whole region. For a twice-higher performance, the score must be double.
领英推荐
3.????? Repeatability. When we ask someone to answer a question before and after lunch or on the weekend, etc., we should get the same answer. It may happen, but we must prove it.
4.????? Bias. The management or other people should not influence the rating. Plus, the reviewer/employee must do the rating sincerely. There is always a risk that the person just added a random number.
5.????? Usually, the rating is between 1 and 5. It is five categories, but if someone has learnt about Gage R&R studies, know that the number of distinct categories (NDC) should be over 10. If it is under 5, we rate them as a go/no-go gauge. So, any such measurement is more like/don′t like.
When I look at an Employee Satisfaction Survey or a Movie Rating (IMDb passes at least one criterion), they are not meeting these standards. We should not trust the results when we do not follow the abovementioned principles.
The situation can be far more complex. I made only a sketch of the problem.
#ai #deeplearning #projectmanagement #interim #sixsigma #projectleadership #medicaldevices #noaigeneratedcontent