2. What is Random Walk Theory
2
It is one of the simplest but important time series
forecasting model
It assumes that for every time period, a variable takes a
random step away from its previous position and these steps
are independently and identically distributed in size
Fx currency market and US stock forecasting are two
examples of application of this model
3. Can Customer Satisfaction be explained by
Random Walk?
3
Customer Satisfaction as measured on industry
benchmark such as JD Power is based on perception of
customers of US utilities( As stock values represents
general public perception of a company)
It is not easy to predict how different parameters such as
rate cases, Utility’s actual performance and market
conditions are going to affect customer satisfaction
4. CSAT Random Walk: Mathematical proof
4
We looked at JD Power quarterly CSAT data from 2010 to date
and ran the quarterly CSAT values for large utilityTop Quartile
and 6 major utilities over this period
We ran the various tests to verify seasonality, autocorrelation and
random walk thru’ various standard tests
As we can see the results, CSAT data supports random walk
theory
JDP Top Q Utl 1 Utl 2 Utl 3 Utl 4 Utl 5 Utl 6 Overall
Original Series Plot Y Y Y Y Y Y Y Y
First Different Plot Y Y Y Y Y Y Y Y
Standard Deviation Check Y Y Y Y Y Y N Y
Random Walk Statistic Test (SAS Default) Y Y Y Y Y Y Y Y
Seasonality There is NO seasonality in the series Regress CSAT scores on season dummy variablesY Y Y Y Y Y Y Y
AutocorrelationThere is NO autocorrelation in the seriesSAS Autoreg procedure Y N Y Y Y N N Y
Conclusion: The data series follows a random walk pattern in general, so we can use random walk model to predict the CSAT scores
Test name Test Methods
Section
Y=Support hypothesis
N= Not Support hypothesis
Hypothesis
There is random walk in the seriesRandom Walk
5. Comparing fundamentals and predict CSAT behavior
with very different probable results
5
Year 2 Year 3
Growth Rate Std Dev Min Average Max Min Average Max
Top Q 0.58% 1.52% 734 742 751 748 760 772
Utl 1 0.49% 2.86% 705 720 736 714 735 756
Utl 2 0.51% 2.30% 715 728 741 726 743 760
Utl 3 0.85% 3.48% 703 722 741 721 747 773
Utl 4 0.49% 2.88% 726 742 758 734 756 778
Ult 5 0.56% 4.50% 711 736 761 719 753 787
Utl 6 0.12% 6.75% 647 682 716 639 685 731
Utl 1 and Utl 3 have similar growth rate and Std deviation but produce
different results
6. Probabilistic Range of Utl 3 and Utl 1relative
to the Top Q range
6
734
748751
772
660
680
700
720
740
760
780
800
2017 2018
Top Q
Utl3
Utl1
RandomWalk Theory helps to predict customer satisfaction and its
relative position toTop Quartile
7. Random Walk helps to calculate the opportunity Cost
of the Delayed Actions to improve CSAT
7
Utl1 Utl2 Utl3 Utl4 Utl5 Utl6
Positive Drift # 14 12 16 16 13 10
Negative Drift# 10 12 8 8 11 14
Positive Drift Avg 2.28% 2.47% 2.83% 2.03% 4.03% 6.59%
Negative Drift Avg -2.03% -1.45% -3.11% -2.60% -3.53% -4.50%
EconomicValue 0.49% 0.51% 0.85% 0.49% 0.56% 0.12%
Positive Drift
Probability
0.58 0.50 0.67 0.67 0.54 0.42
Negative Drift
Probability
0.42 0.50 0.33 0.33 0.46 0.58
Cost of Delayed Action
Min (CSAT Rate)
1.80% 1.96% 1.98% 1.54% 3.47% 6.47%
Cost of Delayed Action
Max (CSAT Rate)
4.31% 3.92% 5.93% 4.62% 7.56% 11.09%
Cost of Delayed Action
Min (CSAT Points)
13 14 15 12 26 44
Cost of Delayed Action
Max (CSAT Points
32 29 44 35 57 76