3. PROJECT MAPPING
D
Define
M
Measure
A
Analyse
I
Improve
C
Control
Customer Sample Comments
Key Output Characteristics
Important to Business
VP Operations & Quality Assurance
Need to initiate the project on service quality as from last 6
months i.e. Aug-17 to Jan’18 we are unable to meet the desired
target of 90%. The objective is to increase the internal quality
score without compromise on the process productivity.
Service Quality
Client Satisfaction
R&P
Customer Relationship Manager
From last 3 months the average quality scores are stagnant at
87%, Need to work with all the team members and understand
where is the gap so that the new target 90%(monthly) can be
achieved and sustained.
Quality Target
Customer Service
Customer Experience Management
Manager (OPS)
Urgent attention required on quality as we have not been able to
achieve the target 90% from last 6 months despite the incentive
program rolled out for quality.
Weekly Quality Scores
Monthly incentives on quality
Number of agents meeting the target
Productivity
4. Dec-17toFeb-18QualitySummary
0.980.840.700.560.420.280.140.00
Median
Mean
0.8900.8850.8800.8750.870
1st Q uartile 0.85420
Median 0.88540
3rd Q uartile 0.91670
Maximum 1.00000
0.86917 0.87706
0.88540 0.89000
0.10866 0.11424
A -Squared 352.84
P-V alue < 0.005
Mean 0.87311
StDev 0.11138
V ariance 0.01241
Skewness -6.2990
Kurtosis 46.5711
N 3064
Minimum 0.00000
A nderson-Darling Normality Test
95% C onfidence Interv al for Mean
95% C onfidence Interv al for Median
95% C onfidence Interv al for StDev
95% Confidence Intervals
Summary for AQS
3 MONTHS DATA GRAPHICAL SUMMARY
1.41.21.00.80.60.40.20.0
99.99
99
95
80
50
20
5
1
0.01
AQS
Percent
Mean 0.8731
StDev 0.1114
N 3064
AD 352.844
P-Value <0.005
Probability Plot of AQS
Normal - 95% CI
• Since, P-Value is less than 0.05, hence data distribution is Non - Normal
• Since data is Normal so, central tendency should be taken as Median
5. PROJECT CHARTER
D
Define
M
Measure
A
Analyse
I
Improve
C
Control
Six Sigma Project Charter
Belt Name: Vijay Kumar Project Name: Quality improvement in PHD process
Division: PHD Process Champion (PO): Nitin Sachdeva
Start Date: 1-Feb
Master Black
Belt:
Nitin Sachdeva
End Date: Team Members: Syed Ali
Gaurav Kumar
Vinay Sharma
Shivangi Tripathi
Kannu Bhardwaj
Problem Statement: Objective & Scope
Internal call quality, an important
performance metric at Pizza Hut contact
centre operations in inTarvo Technologies
Pvt. Ltd, was dipping from 87.9% to 86.7%
during Dec-17 to Feb-18. Analyzing the data
from last three months the average quality
was observed to be 86% which is 4% less
than the desired target of 90%.
The objective of the project is to improve the quality stats through
the identification of the variables that influence the quality. The
goal of the project is to sustain an incremental improvement in
quality so as to reach to desired target of 90%, gain reward on its
billing and enable maximum of its operational employees to
achieve incentives. The scope of the project is centralized order
taking centre at inTarvo Technologies Pvt Ltd connected by a call
centre number 011-39883988.
Primary Metric: Internal Quality Secondary Metrics: Average Call Handling Time
Consequences
Metrics:
6. PROJECT CHARTER: Primary & Secondary Metric
D
Define
M
Measure
A
Analyse
I
Improve
C
Control
90% 90% 90%
88%
87%
87%
85%
86%
87%
88%
89%
90%
91%
Dec-17 Jan-18 Feb-18
AvgQTYscores
Primary Metric
Target
Actual
165 165 165
179 178
179
150
155
160
165
170
175
180
185
190
195
Dec-17 Jan-18 Feb-18
AverageCallHandlingTime
Secondary Metric
Target
Actual
7. TERMS & ACRONYMS USED
D
Define
M
Measure
A
Analyse
I
Improve
C
Control
Indicators Definition
AQS Average Quality Score
QA Quality Analyst
IVR Interactive Voice Response
FCR First Call Resolution
SL Service Level
AHT Average Handling Time
QTY Quality
8. STAKEHOLDER ANALYSIS: ARMI
When Populating the Stakeholder, consider the ARMI:
• A= Approver of team decisions
• R= Resource or subject matter expert (ad hoc)
• M= Member of team
• I= Interested parties who are to be informed
Key Stakeholders Define Measure Analyze Improve Control
VP Operations & Quality Assurance - Intarvo
CRM Head – DIL
GM - Ops & Quality
Project Mentor - IMT
Customer Relationship Manager – DIL
Operations Manager
Quality Manager I/R I/R I/R I/R I/R
Team Members M M M M M
I/A I/A I/A I/A I/A
I/A I/A I/A I/A I/A
I/A I/A I/A I/A I/A
I I I I I
I/A I/A I/A I/A I/A
I I I I I
9. RASIC
D
Define
M
Measure
A
Analyse
I
Improve
C
Control
• Responsible (R) : Solely and directly responsible for the activity (Owner) - Includes approving authority (A)
• Approve (A) : Reviews and assures that the activity is being done as per expectations
• Support (S) : Provides the necessary help and support in the project
• Inform (I) : Is to be kept informed of the status/progress being made
• Consult (C) : Is to be consulted for this activity for inputs
RASIC Chart for Define & Measure
Activities
VP
Opera
tions
&
Qualit
y
Assur
ance-
Intarv
oCRM
Head-
DIL
Projec
t
Ment
or-
IMTCusto
mer
Relati
onshi
p
Mana
ger-
DIL
Opera
tions
Mana
ger
Qualit
y
Mana
ger
Team
Memb
ers
Memb
ers
(Assoc
iates)
Pre- DMAIC
Collection of VOB from all stakeholders I I I I/A S/C/R R - -
Data Collection for last 3 months I I I I/C R/S R S -
Analysis of data I I I C I R S -
Report out on the Pre DMAIC Analysis I I I I/C I/C R - -
DEFINE
Project Charter Creation I I A I/S/C I R - -
Charter Approval from Executives/Project Guide A I A S/R - - - -
Build SIPOC I I I I I R S -
Build Process Map - - - S/C S/C R - -
MEASURE
Build the data collection plan I I I S/C S R - -
Get the DCP Approval - - C A - R - -
Approve DCP I I A - - R - -
Collect Data - - I/C I S R - -
Validate data - - C S/C S R - -
Publish next steps to stakeholder I I I I I R - -
12. THE SIPOC PROCESS MAP
D
Define
M
Measure
A
Analyse
I
Improve
C
Control
Suppliers Inputs Process Output Customer Requirements
• Phones
• Office depot
• System
• CRM
• Customer
• Store
• Accounting
team
• Complete call <
3mins
• Correct address
• Correct price
• Accuracy of
punching
• Customer’s query
• Customer details
• Address
• Name
• Time, date
• Volume
• CRM navigation
• Training, Coaching &
feedback
• Shifts
• Hold
Call received
by CC
Associate
Customer’s
concern
identified
Customer details
verified in CRM
Information
provided to
customer
Call is
documented in
CRM
Call closed by cc
Associate
Quality Audit
done next
day/same day
• Call completion
• Case tagged
• Data on cycle
time
• Process metrics
data saved
14. DATA COLLECTION PLAN
D
Define
M
Measure
A
Analyse
I
Improve
C
Control
Y
Operational
Definition
Defect Definition
Required Performance
Standard
Specification Limit Opportunity
Improvement in internal Call
Quality
Points Achieved/Points
Available
<85% 90%
USL- NA
LSL- 85%
Weekly
Y Data Type Unit of Measurement Decimal Places
Database
Container
Existing/New
Database
If new, when
would the
database be
ready
Planned Start
date for Data
Collection
Improvement in internal
Call Quality
Discrete % Yes Excel Existing NA 15h
Apr
Data Items
Needed
Formula to be
Used
Equipment Used for
Measurement
Equipment
Calibration Info
Responsibility Training Need
Operator
Information
Quality data for last 3
months
Points
Achieved/Points
Available
Audit Legend Quality Lead NA Quality Team
MODE OF COLLECTING DATA
15. MSA: MEASUREMENT SYSTEM ANALYSIS
VinaySyedShivangi
95
90
85
80
75
70
Appraiser
Percent
95.0% C I
Percent
VinaySyedShivangi
95
90
85
80
75
70
Appraiser
Percent
95.0% C I
Percent
Date of study: 20-March-2018
Reported by: Vijay Kumar
Name of product:
Misc:
Assessment Agreement
Within Appraisers Appraiser vs Standard
ATTRIBUTE AGREEMENT ANALYSIS
• The sample of 60 units has been evaluated by 3 different evaluators and the audit variance
of 3 appraisers is ranging between 85% to 87%
• The appraiser vs standard agreement has also been observed between 80% to
81%
• Above results are enough to conclude that the evaluation scores are accurate and
precise, however appraisers need some calibration with master calibrator
16. MSA: MEASUREMENT SYSTEM ANALYSIS
D
Define
M
Measure
A
Analyse
I
Improve
C
Control
Process Sigma Level Calculator - Discrete Data
Sample Data (user inputs):
Number of units n 3,064
Total number of defects observed d 1,889
Number of defect opportunities per unit o 1
Sigma Shift (typically +1.5 for long term data) 1.5
Results:
Defects per Unit dpu 0.61651436
Defects per Million Opportunities dpmo 616,514.4
Defects per Opportunity dpo% 61.65%
Yield yield% 38.35%
Process Sigma Level sigma 1.204
18. D
Define
M
Measure
A
Analyse
I
Improve
C
Control
Hypothesis Testing – Overall Summary
Key Input Variables Input Type Data Collection possible Data Type - X Data Type - Y Data Normality (Y) Test Used Outcome
Call Type N Yes Discrete Continuous No Kruskal-Wallis test Reject Null Hypothesis
Duration P Yes Continuous Continuous No Spearman Correlation Reject Null Hypothesis
Trainers C Yes Discrete Continuous Yes ANOVA Fail to reject the null hypothesis
Feedback Delivery C Yes Discrete Continuous No Wilcoxon rank sum test Reject Null Hypothesis
Team leader wise QTY scores C Yes Discrete Continuous No Kruskal-Wallis test Reject Null Hypothesis
Shift Stretch/Overtime P Yes Discrete Continuous No Wilcoxon rank sum test Reject Null Hypothesis
FCR C Yes Discrete Continuous No Wilcoxon rank sum test Reject Null Hypothesis
Gender C Yes Discrete Continuous Yes Two Sample t-test Reject Null Hypothesis
Peak non peak N Yes Discrete Continuous No Wilcoxon rank sum test Reject Null Hypothesis
Age C Yes Continuous Continuous Yes Pearson Correlation Reject Null Hypothesis
Different Shift C Yes Discrete Continuous No Kruskal-Wallis test Reject Null Hypothesis
Part time/full time C Yes Discrete Continuous No Wilcoxon rank sum test Reject Null Hypothesis
Associate's Tenurity P Yes Continuous Continuous Yes Pearson Correlation Reject Null Hypothesis
Experienced vs Fresher Associate C Yes Discrete Continuous Yes Two Sample t-test Reject Null Hypothesis
High hold usage/avg hold C Yes Continuous Continuous Yes Pearson Correlation Fail to reject the null hypothesis
Day Hours N Yes Discrete Continuous No Kruskal-Wallis test Reject Null Hypothesis
Week Day N Yes Discrete Continuous No Kruskal-Wallis test Reject Null Hypothesis
Education background C Yes Discrete Continuous Yes Two Sample t-test Reject Null Hypothesis
Hiring Source C Yes Discrete Continuous Yes ANOVA Reject Null Hypothesis
Rushing on call C Yes Discrete Continuous No Wilcoxon rank sum test Reject Null Hypothesis
Training, feedback & coaching C Yes Continuous Continuous Yes Pearson Correlation Fail to reject the null hypothesis
Call Length C Yes Discrete Continuous No Wilcoxon rank sum test Reject Null Hypothesis
Attrition C Yes Discrete Continuous Yes Two Sample t-test Fail to reject the null hypothesis
Pressure of Other Metrics N Yes Continuous Continuous No Pearson Correlation Fail to reject the null hypothesis
Top Call Drivers C Yes Discrete Discrete No Pareto Analysis NA
Tagging % N No Discrete Discrete NA NA NA
Technical issues N No Discrete Continuous NA NA NA
Process Complexity N No Discrete Continuous NA NA NA
Floor Environment C No Discrete Continuous NA NA NA
Training Material N No Discrete Continuous NA NA NA
19. D
Define
M
Measure
A
Analyse
I
Improve
C
Control
Paramatric Testing – Data Normality
Shapiro-Wilk normality test data:
Agentwise_data_Monthly$QTY_Score W =
0.98135, p-value = 0.07097
Anderson-Darling normality test data:
Agentwise_data_Monthly$QTY_Score A =
0.48113, p-value = 0.2284
Null Hypothesis: The agent wise quality score follows a normal
distribution
Alternate Hypothesis: The agent wise quality score does not follow a
normal distribution
Result: As the p-value in both the test (Shapiro-Wilk normality test &
Anderson-Darling normality test) is greater than 0.05, we fail to reject
the null hypothesis of data follows a normal distribution.
20. D
Define
M
Measure
A
Analyse
I
Improve
C
Control
2 Sample T-Test – Gender wise QTY scores
Null Hypothesis: Average quality score are equal in male & female
both categories
Alternate Hypothesis: Average quality score for females is higher
than that of males
Result: As the p-value for t-test is less than 0.05, we reject the null
hypothesis of mean quality scores are equal and conclude at 95%
confidence level that we have enough evidence of assuming that
mean quality scores of female employees are higher than the male
employees.
21. D
Define
M
Measure
A
Analyse
I
Improve
C
Control
2 Sample T-Test – Experience wise QTY scores
Null Hypothesis: Average quality score for fresher & experienced
employees are equal
Alternate Hypothesis: Average quality score of experienced employees
is higher than that of fresher employees
Test result: As the p-value for t-test is less than 0.05, we reject the null
hypothesis of mean quality scores are equal and conclude at 95%
confidence level that we have enough evidence of assuming that mean
quality scores of experienced employees are higher than the fresher
employees.
22. D
Define
M
Measure
A
Analyse
I
Improve
C
Control
2 Sample T-Test – Attrition
Null Hypothesis: Average quality scores are equal for attired and non
attired employees
Alternate Hypothesis: Average quality scores are not equal for attired
and non attired employees
Test result: As the p-value for t-test is greater than 0.05, we fail to
reject the null hypothesis of mean quality scores are equal and
conclude at 95% confidence level that we have enough evidence of
assuming that mean quality scores of both category of employees are
same.
23. D
Define
M
Measure
A
Analyse
I
Improve
C
Control
2 Sample T-Test – Educational Background
Null Hypothesis: Average quality scores are equal for graduate and
under graduate employees
Alternate Hypothesis: Average quality scores of graduate employees
are greater than of under graduate employee
Test result: As the p-value for t-test above is less than 0.05, we reject
the null hypothesis of mean quality scores are equal for both level of
education background and conclude at 95% confidence level that we
have enough evidence of assuming that mean quality scores of
graduate employees are greater than under graduate employees.
24. D
Define
M
Measure
A
Analyse
I
Improve
C
Control
ANOVA - Trainers
Null Hypothesis: There is no significant difference in trainer wise (3
trainers) average quality scores
Alternate Hypothesis: At least one trainer trained trainees’ quality
scores are different
Test result: As the p-value for t-test above is greater than 0.05, we fail to reject
the null hypothesis of trainer wise mean quality scores are equal and conclude
at 95% confidence level that we have enough evidence of assuming that mean
quality scores are same irrespective of the batch trained by any trainer.
26. D
Define
M
Measure
A
Analyse
I
Improve
C
Control
ANOVA - Hiring Source
Null Hypothesis: There is no significant difference in average
quality scores of employees hired from different sources
Alternate Hypothesis: At least one group of employees’ quality
scores are different
Test result: As the p-value for t-test above is less than 0.05, we reject the null hypothesis of mean quality scores are
equal for associates hired from different hiring sources and conclude at 95% confidence level that we have enough
evidence of assuming that mean quality scores of at least one hiring source employees is different.
27. D
Define
M
Measure
A
Analyse
I
Improve
C
Control
Hiring Source Pairwise Probabilities Matrix
Pairwise
Probabilities
Campus
Drive
Job
Portal
Referral Walk in
Campus Drive 0.0000 0.0002 0.0058
Job Portal 0.1605 0.0039
Referral 0.2086
Walk in
As the P-Values in pairwise probabilities matrix for campus drive
are less than 0.05 against the other sources of hirings, we
concluded that scores are poor for employees those who have
been hired from campus drive. Similarly walk in.
28. D
Define
M
Measure
A
Analyse
I
Improve
C
Control
Correlation Test – Hold and Feedback
Null Hypothesis: There is no correlation between
average quality scores and average hold time
Alternate Hypothesis: There is a correlation between
average quality scores and average hold time
data: Avg_Feedback_Break and QTY_Score
t = 1.0884, df = 128, p-value = 0.2785
alternative hypothesis: true correlation is not equal to 0
95 percent confidence interval:
-0.07770563 0.26360135
sample estimates:
cor
0.09576174
data: Avg_Hold_Time and QTY_Score
t = -1.3789, df = 128, p-value = 0.1703
alternative hypothesis: true correlation is not equal to 0
95 percent confidence interval:
-0.28718967 0.05228902
sample estimates:
cor
-0.1209864
Test result: As the p-value for t-test above is greater than 0.05, we fail to reject
the null hypothesis and conclude at 95% confidence level that we have enough
evidence of assuming that there is no correlation between average quality scores
and average hold time.
Test result: As the p-value for t-test above is greater than 0.05, we fail to reject
the null hypothesis and conclude at 95% confidence level that we have enough
evidence of assuming that There is no correlation between average quality scores
and average feedback breaks.
Null Hypothesis: There is no correlation between
average quality scores and average feedback breaks
Alternate Hypothesis: There is a correlation between
average quality scores and average feedback breaks
29. D
Define
M
Measure
A
Analyse
I
Improve
C
Control
Correlation Test – Age and Tenurity
Null Hypothesis: There is no correlation between average
quality scores and age of the associates
Alternate Hypothesis: There is a correlation between
average quality scores and age of the associates
Null Hypothesis: There is no correlation between average
quality scores and tenurity of the associates
Alternate Hypothesis: There is a correlation between
average quality scores and tenurity of the associates
Test result: As the p-value for t-test above is less than 0.05, we reject the null
hypothesis that there is correlation between quality scores and age of the
associates and conclude at 95% confidence level that we have evidence of
assuming that there is a moderate correlation between average quality scores
and age of the associates.
Test result: As the p-value for t-test above is less than 0.05, we reject the null
hypothesis that there is correlation between quality scores and tenurity of the
associates and conclude at 95% confidence level that we have enough evidence
of assuming that there is a correlation between average quality scores and
tenurity of the associates.
30. D
Define
M
Measure
A
Analyse
I
Improve
C
Control
Linear Regression - Age
Null Hypothesis: There is no significant prediction of quality
scores through age of the associates
Alternate Hypothesis: There is a significant prediction of quality
scores through age of the associates
Result: Since the R sq value is 13%, associates age predict only
13% variation in quality scores. We can not consider the model
as R sq value is below 50% and conclude at 95% confidence
interval that there can not be a significant prediction of quality
scores through age of the associates.
RegressionModel:QTY_Score = (0.739839) + (0.005364)*Age
31. D
Define
M
Measure
A
Analyse
I
Improve
C
Control
Regression
Null Hypothesis: There is no significant prediction of quality
scores through tenurity of the associates
Alternate Hypothesis: There is a significant prediction of quality
scores through tenurity of the associates
Regression Model: QTY_Score = (0.822761) + (0.000262) * AssociatesTenurity
Result: Since the R sq value is 51%, associates tenurity predict
51% variation in quality scores. We can consider the model as R sq
value is above 50% and conclude at 95% confidence interval that
there can be a significant prediction of quality scores through
tenurity of the associates.
33. D
Define
M
Measure
A
Analyse
I
Improve
C
Control
Pareto Analysis – Top Call Drivers
1. As shown in above Pareto graph 80% of the defects on the calls are due to 20% of the parameters available
in quality sheet.
2. Majorly these defects are on soft skills parts only
3. No major process knowledge gap identified
34. Non-Parametric Testing
0.980.840.700.560.420.280.140.00
Median
Mean
0.8900.8850.8800.8750.870
1st Q uartile 0.85420
Median 0.88540
3rd Q uartile 0.91670
Maximum 1.00000
0.86917 0.87706
0.88540 0.89000
0.10866 0.11424
A -Squared 352.84
P-V alue < 0.005
Mean 0.87311
StDev 0.11138
V ariance 0.01241
Skewness -6.2990
Kurtosis 46.5711
N 3064
Minimum 0.00000
A nderson-Darling Normality Test
95% C onfidence Interv al for Mean
95% C onfidence Interv al for Median
95% C onfidence Interv al for StDev
95% Confidence Intervals
Summary for AQS
1.41.21.00.80.60.40.20.0
99.99
99
95
80
50
20
5
1
0.01
AQS
Percent
Mean 0.8731
StDev 0.1114
N 3064
AD 352.844
P-Value <0.005
Probability Plot of AQS
Normal - 95% CI
Null Hypothesis: The agent wise quality score follows a normal distribution
Alternate Hypothesis: The agent wise quality score does not follow a normal distribution
Result: As the p-value in both the test (Shapiro-Wilk normality test & Anderson-Darling
normality test) is negligible & less than 0.05, we reject the null hypothesis of data
follows a normal distribution and conclude with 95% confidence level the we have
enough evidence of assuming non normality of the data.
35. D
Define
M
Measure
A
Analyse
I
Improve
C
Control
Kruskal-Wallis rank sum test – Team Leader wise QTY scores
Pairwise comparisons using Wilcoxon rank sum test
Null hypothesis: There is no difference in team leader wise quality
scores’ distributions
Alternate hypothesis: At least one team leader wise quality scores’
distribution is different
Result: reject the null hypothesis
36. D
Define
M
Measure
A
Analyse
I
Improve
C
Control
Wilcoxon rank sum test – PT/FT
Null hypothesis: Median of quality scores for part timers and full timers
are equal
Alternate hypothesis: Median of quality score of part timers is lesser
than that of full timers
Result: As the p-value is less than 0.05, we reject the null hypothesis
that median of quality scores for part timers and full timers are equal
and conclude with 95% confidence level that we have enough evidence
of assuming median of quality score of part timers is lesser than that of
full timers.
37. D
Define
M
Measure
A
Analyse
I
Improve
C
Control
Kruskal-Wallis rank sum test – PT/FT
Null hypothesis: There is no difference in quality scores’ distributions of
different categories of calls
Alternate hypothesis: At least one category of calls has different quality
scores’ distribution
Result: As the p-value is negligible & less than 0.05, we reject the null
hypothesis of different categories call quality scores’ distributions are
indifferent and conclude with 95% confidence level that we have
enough evidence of assuming at least one call category quality scores’
distribution is different. As shown in box plot the QTY scores for
complaint category of calls are pretty poor
38. D
Define
M
Measure
A
Analyse
I
Improve
C
Control
Kruskal-Wallis rank sum test – Different Shifts
Null hypothesis: There is no difference in quality scores’
distributions of different shifts
Alternate hypothesis: At least one calls audited for one shift
has different quality scores’ distribution
Result: As the p-value for Kruskal-Wallis test is negligible & less
than 0.05, we reject the null hypothesis of call quality scores’
distributions in all shifts are indifferent and conclude with 95%
confidence level that we have enough evidence of assuming at
least one shifts’ quality score distribution is different. Distribution
of 19:30-23:30 shift quality scores is different than others.
39. D
Define
M
Measure
A
Analyse
I
Improve
C
Control
Wilcoxon rank sum test – Peak vs Non-Peak
Null Hypothesis: Median of quality scores are equal in peak and non
peak hours
Alternate Hypothesis: Median of quality score in peak hours is lesser
than in non peak hours
Result: As the p-value is negligible & less than 0.05, we reject the null
hypothesis that median of quality scores are equal in peak and non
peak hours and conclude with 95% confidence level that we have
enough evidence of assuming peak hours quality scores are less than
non peak hours quality scores.
40. D
Define
M
Measure
A
Analyse
I
Improve
C
Control
Wilcoxon rank sum test – Call Length
Null hypothesis: Median of quality score for long calls is less than or
equal of median of short calls
Alternate hypothesis: Median of quality score for short calls is greater
than of long calls
Result: As the p-value is 1, we fail to reject the null hypothesis that
Median of quality score for long calls is less than or equal of median of
short calls and conclude with 95% confidence level that we have
enough evidence of assuming that long calls quality scores are less than
short calls.
41. D
Define
M
Measure
A
Analyse
I
Improve
C
Control
Wilcoxon rank sum test – Overtime
Null Hypothesis: Median of quality scores in overtime hours is less than
or equal to non-overtime hours
Alternate Hypothesis: Median of quality scores in overtime hours is
greater than to non-overtime hours
Result: As the p-value is 1, we fail to reject the null hypothesis that
Median of quality score for overtime hours is less than or equal of
median of non-overtime hours and conclude with 95% confidence level
that we have enough evidence of assuming that overtime hours quality
scores are less than non-overtime hours quality scores.
42. D
Define
M
Measure
A
Analyse
I
Improve
C
Control
1. Quality scores of part timers are lesser than that of full timers and they are contributing more on lower side of quality
scores
2. Complaint category quality scores are extremely poor and demands immediate attention
3. As mentioned in point no 2 part timer who majorly come in 19:30-23:30 shift are impacting the quality scores severely
4. Quality scores improves for next call if feedback is given for previous call within 24 hours of call audit
5. Peak hours which are generally observed between 06:00 PM to 10:30 PM are having greater impact on lesser side of
quality scores
6. During the peak hours 08-09 PM & 09-10 PM are the major hit area and an strategy must be formulated for these hours
7. Wednesday quality scores distribution is different than others.
8. Associates are unable to score good on long calls and these calls are accounting for lower quality scores
9. Associate are not active, energetic or serious during their overtime hours and scoring poor in quality scores during
overtime hours
10. FCR calls quality score are higher than non FCR, which clearly shares a message that if the query of a customer is not
resolved in first go it becomes an unpleasant experience for guest and tough for agents to handle such calls
11. TL Sidhartha Kapoor’s team has been scoring poor in quality scores
12. Quality scores of female employees are higher than the male employees
13. Quality scores of experienced employees are higher than the fresher employees
14. 80% of the call defects are due to lack of the soft skills of associates
15. Quality scores of graduate employees are greater than under graduate employees
16. Associate tenurity can be modeled to predict the quality scores, higher the tenurity, higher the quality scores
CONCLUSIONS
43. D
Define
M
Measure
A
Analyse
I
Improve
C
Control
1. Employees who are having experience from same
industry are scoring good in terms of quality
1. Refresher training for fresher employees every week
2. Fresher Employees are scoring poor quality scores 2. Stretching training days for fresher employees
3. Fresher Employees are adding more fatals to the
calls
3. Conducting briefings on top call drivers
4. Sharing feedback on aily basis so that the employees can work
on defaulting areas
1.Female employees are scoring good quality scores 1. Conducting refresher training for male associates
2. Male employees are scoring poor quality scores
2. One-o-one session with the employess so that they can be
motivated toward their work
3. Call listening sessions on floor so that best practice can be
followed across.
1. Quality Scores vary team leader wise
2. Atleast one team leader has low quality scores
CallType
1. Poor quality scores in complaint calls
1. Team Leaders will share feedback on real time basis
2. Daily pre/post briefings will be conducted by Team leaders
2. Hiring less number of part timers
1.Agents would be coming to office 15 mins prior to their shifts starts time so that they can be given briefing on the basis of Day-1's
observations
2. Refreshers and complaint handling session to be conducted for defaulter
1. Part timers are having low quality scores in
comparison to full timers
PT/FT
1. It is observed that part timers attend less number of refresher and other sessions so the number of sessions to be increased
Gender
ExperiencedEmployeesvs
FresherEmployees
1.More no. of female employees to be employed
1. Hiring Experienced candidates so that the good quality can
be ensured
Defaulting Areas Factors Corrective Preventive
TeamLeaders
RECOMMENDATION
44. D
Define
M
Measure
A
Analyse
I
Improve
C
Control
1. Short calls are having good quality scores
2. Long calls are having poor quality scores
Overtime
1. Employees who are doing overtime are having poor
quality scores
HiringSource
1. Associate hired from campus drive are scoring poor
in quality
1. Associates who are scoring poor consistently are to be
indentified and must be issued a perforance improvement plan
1. More number of employees are to be hired from Job portal
and employee referral.
Tenurity
1. As the agent is tenured in he system his QTY scores
improves
1. Management to work on 0-90 days tenurity bucket in term of attrition
WeekDay
1. Wednesday having special offers which increases
the demand and stores lack on scalability which turns
customers unhappy and contribute in lower quality
scores
1. To push client to align more resources to store
2. To align 100% manpower at call centre
Call
Length
1. Team of knowledge holders will be prepared for handling calls efficiently
2. Arranging call back for the customers
1. Effcetive manpower/batch palnning to be done
DifferentShifts
1. Associates in shift 19:30-22:30 are scoring poor
quality score
1. Refresher training for employees
2. Conducting call play sessions for the associates
3. Conducting pre shift briefings to motivate the employees
Defaulting Areas Factors Corrective Preventive
RECOMMENDATION
Editor's Notes
This data includes the critical errors/defects which are rated 0%
This data does not include the critical errors/defects which are rated 0%