420 likes | 965 Views
Introduction. Satisfaction surveys measure value judgments of customers on products or services. The objective of such studies is twofolds:To deliver an accurate and robust measurement of a Global Satisfaction Index of customers.To prioritize the quality improvements that could be undertaken, in order to get a significant impact on this Global Satisfaction Index. .
E N D
1. PLS Multinomial Logit in Satisfaction Surveys 6th International Conference on Partial Least Square and Related Method Beijing - Sep. 4 – Sep. 7 INTERSTAT I’d like to look with you at an application of PLS Logit Multinomial Regression in Market Research, a special type of Market surveys called “Satisfaction Survey”I’d like to look with you at an application of PLS Logit Multinomial Regression in Market Research, a special type of Market surveys called “Satisfaction Survey”
2. Introduction Satisfaction surveys measure value judgments of customers on products or services.
The objective of such studies is twofolds:
To deliver an accurate and robust measurement of a Global Satisfaction Index of customers.
To prioritize the quality improvements that could be undertaken, in order to get a significant impact on this Global Satisfaction Index. 1 INTERSTAT Essentially …Essentially …
3. Introduction Typically, these studies entail implementing Correlations, Multiple Regression or Logit Multinomial Regression in order to relate Global Satisfaction to more specific components of product quality.
However, a common pitfall of such studies is that the various quality assessement considered as explanatory variables of the Global Satisfaction Index tend to be highly correlated.
2 INTERSTAT ……
4. Introduction
From the research work of M. Tenenhaus, V. Vinzi and P. Bastien, we developed a software solution implementing...
PLS Logit Multinomial Regression
The present communication is intended to bring to light the benefits of this methodology over the classical Logit Multinomial Regression on a sample of ten real surveys selected in various domains.
3 INTERSTAT ……
5. Hypothesis to check in this test The regular Logit Multinomial Regression highly suffers from missing data and multi colinearity between explanatory variables:
Multi colinearity leads to a lack of robustness in the model estimations.
Another outcome could be the so-called suppressive effect that leads parameters to be – erroneously - non significant.
A third effect is that some parameters may display counter-intuitive results (e.g. unexpected sign of a parameter bh)
Missing data in explanatory variables are not uncommon in Market Research surveys; the Logit MNL method deals with missing data “list wise”, i.e. each row of data with at least one missing data is discarded from the analysis, which may turn out to be very costly, as we will see later. 4 INTERSTAT Phonetic: “Ve:riable”
e.g.: for examplePhonetic: “Ve:riable”
e.g.: for example
6. Hypothesis to check in this test 5 INTERSTAT So …So …
7. Summary Real issues of Satisfaction Surveys
Choice of an appropriate MNL Regression model
The PLS Logit Multinomial Regression model
Description of the data sets sample used in the test
Main results of the Test
Conclusion 6 INTERSTAT We will cover successively these six points.
First, we will go back quickly to the real issues of satisfaction surveys in Market Research.
Based on some of those considerations, we will clarify the question of what should be an appropriate Regression model.
Then we will present the PLS Logit MNL algorithm that we implemented.
At this step we will enter the core question of the present test.
In presenting first the data sets used … then the results of the test
We will cover successively these six points.
First, we will go back quickly to the real issues of satisfaction surveys in Market Research.
Based on some of those considerations, we will clarify the question of what should be an appropriate Regression model.
Then we will present the PLS Logit MNL algorithm that we implemented.
At this step we will enter the core question of the present test.
In presenting first the data sets used … then the results of the test
8. Real issues of Satisfaction Surveys 7 INTERSTAT Let’s begin by looking at Satisfaction surveysLet’s begin by looking at Satisfaction surveys
9. 8 Real issues of Satisfaction Surveys INTERSTAT ……
10. 9 Real issues of Satisfaction Surveys INTERSTAT As …As …
11. 10 Real issues of Satisfaction Surveys INTERSTAT At the end
I wont comment on this. You will find all the argument in the literature. Feel free to ask if you are interested.At the end
I wont comment on this. You will find all the argument in the literature. Feel free to ask if you are interested.
12. The "Net Promoter Score" In the following test, we will go through ten surveys based on the "Net Promoter Score" paradigm.
The selected criterion is a zero-to-ten scale, according to the degree of agreement to the selected unique statement:
"Would you recommend Brand X to a friend or a colleague?"
from 0 ="not at all likely"
to 10="extremely likely".
11 INTERSTAT ……
13. Definition of the "Net Promoter Score" 12 INTERSTAT ……
14. From the "Net Promoter Score" to Action Plan Once having this Global NPS score, the key issue in consumer satisfaction management is to specify and implement an appropriate Action Plan.
A first approach, purely managerial, is based on an internal development of what could be called an NPS corporate culture, as for instance a remuneration plan based on the NPS score.
The other, more Market Research oriented, consists in identifying a series of significant satisfaction drivers to prioritize.
13 INTERSTAT At the end
On this point, I’ll come back soon in the next section
At the end
On this point, I’ll come back soon in the next section
15. CHOICE OF AN APPROPRIATE REGRESSION MODEL 14 INTERSTAT Where does that NPS score lead us to?Where does that NPS score lead us to?
16. "Drivers" 15 INTERSTAT … …
17. A two-dimensions scheme 16 The one dimension scheme assumes that each satisfaction drivers operate along a continuum from "dissatisfaction" to "satisfaction". INTERSTAT Left part:
(After the comment on the slide)
A driver is like a lever.
If you handle this lever in one direction or the other, the Global Satisfaction Index will move symmetrically toward more or less satisfaction
Right part:
Driver #2 only impact in dissatisfaction
Driver #3 only impact in satisfaction
Left part:
(After the comment on the slide)
A driver is like a lever.
If you handle this lever in one direction or the other, the Global Satisfaction Index will move symmetrically toward more or less satisfaction
Right part:
Driver #2 only impact in dissatisfaction
Driver #3 only impact in satisfaction
18. "Attractive" and "Must be" drivers 17 INTERSTAT A diagram
Bottom:
A good example of this is …A diagram
Bottom:
A good example of this is …
19. The Category Base Logit Multinomial Model We meet again our dependant variable with the 3 clusters that allows to compute the NPS score.
Detractors
Neutral (Passively Satisfied)
Promoters
18 INTERSTAT ……
20. Meaning of the 2 sets of b parameters This two dimensional arrangement allows to get two sets of b parameters.
The meaning of those 2 sets of b parameters is made explicit in the following logit expressions.
19 INTERSTAT Equation
The expression on the left connects cell #1, the probability of being “Dissatisfied” to the base cell #2. It includes the b1 parameters.
The expression on the left connects cell #3 , the probability of being “Satisfied” to the base cell #2.
It includes the b2 parameters.
CLICK
Bottom
(Before reading)
The category of the driver depends on the respective value of parameters b1 and b2 .
When b2 greater than b1 we are faced with an Attractive driver,
etc.Equation
The expression on the left connects cell #1, the probability of being “Dissatisfied” to the base cell #2. It includes the b1 parameters.
The expression on the left connects cell #3 , the probability of being “Satisfied” to the base cell #2.
It includes the b2 parameters.
CLICK
Bottom
(Before reading)
The category of the driver depends on the respective value of parameters b1 and b2 .
When b2 greater than b1 we are faced with an Attractive driver,
etc.
21. Estimation of the of the b parameters
We first estimated these b parameters using a Logit Multinomial Regression model, namely the Nomreg® procedure in the SPSS package.
In this Logit Multinomial model, the criterion used to estimate the parameters is the Maximum likelihood. The parameters estimates are computed using the Newton-Raphson iterative algorithm.
We recall here that this algorithm deals with missing data listwise, which means that every record with at least one missing data in the explanatory variable is discarded. We will see in the analysis of real surveys data set, the importance of this feature.
20 INTERSTAT
22. The PLS Logit Multinomial Regression 21 INTERSTAT Let me turn now to the PLS Logit Mutlinomial RegressionLet me turn now to the PLS Logit Mutlinomial Regression
23. The PLS Logit Multinomial : an iterative process 22 INTERSTAT The PLS Logit Multinomial Regression follow an iterative processThe PLS Logit Multinomial Regression follow an iterative process
24. Step 1: To compute new weights 23 INTERSTAT We have here another – internal – iterative process.
Starting with the first loop, we need to implement K Simple Logistic Regression, each having only one independent variable xj .
In a current loop – lets call it number “h” - the xj become the residual xh-1,j , and we have to add the PLS component already computed until then.We have here another – internal – iterative process.
Starting with the first loop, we need to implement K Simple Logistic Regression, each having only one independent variable xj .
In a current loop – lets call it number “h” - the xj become the residual xh-1,j , and we have to add the PLS component already computed until then.
25. Step 2: To compute a new PLS Component 24 INTERSTAT Then we compute the PLS component, according to NIPALS principles in order to cope with missing data.Then we compute the PLS component, according to NIPALS principles in order to cope with missing data.
26. Step 3: To compute new X residuals 25 INTERSTAT Then the residuals Xhs are computed in order to go back to the first step of the iteration
CLICK
We have here to test a stop criteria, essentially the maximum number of component reached, which is left to the decision of the modeler.
In this first version of the Logycs model we didn’t implement yet the PRESS statistic.Then the residuals Xhs are computed in order to go back to the first step of the iteration
CLICK
We have here to test a stop criteria, essentially the maximum number of component reached, which is left to the decision of the modeler.
In this first version of the Logycs model we didn’t implement yet the PRESS statistic.
27. Datasets Sample 26 INTERSTAT Let’s now have a look at our sample of data sets.Let’s now have a look at our sample of data sets.
28. Description of the ten datasets First of all, the studies were addressed to different populations, in different fields, which implied different level of satisfaction scores.
They also have different forms, with very different size of samples or different numbers of explanatory variables.
27 INTERSTAT ……
29. Selected Sectors and Target Groups 28 INTERSTAT We have essentially three sectors of Services:
Bank / Finance
Mobile phone
Call center
As well as three target groups:
Consumers
Business
Employees
We have essentially three sectors of Services:
Bank / Finance
Mobile phone
Call center
As well as three target groups:
Consumers
Business
Employees
30. Number of variables 29 INTERSTAT The number of independent variable vary in a wide range of 7 to 66The number of independent variable vary in a wide range of 7 to 66
31. Mising values 30 INTERSTAT The dark grey area correspond to the numbers of cells with missing data.
The light grey area correspond to the percentage of rows of data with at least one cell of missing data, what is called missing data “list wise”.
We’ll be examining this point in more details later on …
The dark grey area correspond to the numbers of cells with missing data.
The light grey area correspond to the percentage of rows of data with at least one cell of missing data, what is called missing data “list wise”.
We’ll be examining this point in more details later on …
32. Dependant variables frequencies 31 INTERSTAT The purple bars correspond to the frequencies of “Detractors”
The orange bars correspond to the frequencies of “Promotors”
… the NPS scores is the differences between the two bars.
The purple bars correspond to the frequencies of “Detractors”
The orange bars correspond to the frequencies of “Promotors”
… the NPS scores is the differences between the two bars.
33. Net Promoters Score 32 INTERSTAT The data sets are sorted in decreasing order along the NPS ScoreThe data sets are sorted in decreasing order along the NPS Score
34. RESULTS OF THE TEST 33 INTERSTAT Turning now to racing results …Turning now to racing results …
35. 4. Results of the test The following results aim at doing comparisons between Logit Multinomial Regression with SPSS ™ and PLS Logit Multinomial Regression. For the latter, we used the Logycs™ software developed conjointly by Interstat and GN Research.
We are interested here in two global criteria:
34 INTERSTAT Here are the rules of the game …Here are the rules of the game …
36. Scores of recognition 35 INTERSTAT
37. Scores of recognition 36 INTERSTAT
38. Data sets features revisited 37 INTERSTAT
39. How much more efficient the PLS MNL model is …. 38 INTERSTAT
40. Scores of conformity to b sign expectations 39 INTERSTAT
41. CONCLUSION 40 INTERSTAT
42. Conclusion The two pitfalls of the regular MNL method considered in the premise of this test was missing data and multi colinearity.
Missing data management appeared to be the main contribution of the PLS method, due to the frequency of missing data in our Satisfaction surveys sample. The “list wise” manner of the regular method to deal with missing data leads to huge reductions of data available. It appears to be totally inappropriate and brings considerable biases in the parameter estimates.
Multi colinearity leads to counter-intuitive results. We tested the frequency of “unexpected signs of bh parameters”. Here again, the PLS method brings a significant contribution. We estimate a 20% decrease in the number of counter intuitive parameters. 41 INTERSTAT If I can just sum up the main points, we considered the two pitfalls of the regular MNL method…If I can just sum up the main points, we considered the two pitfalls of the regular MNL method…
43. 42 gnresearch