1

Operations Management Techniques for Ecommerce Industry

Quality Function Deployment(QFD) and Forecasting

SUBMITTED BY:
Group -7 Section D

Dhananjay Shukla (210102018)

Kunal Pareek (210102031)
Lovneesh Vij (210103086)
Paresh Prabhudesai (210102039)
Rhea Nehulkar (210103126)
Sajithsharan Sureshkumar (210101091)
Soumik Bhattacharya (210101110)

A REPORT SUBMITTED IN FULFILMENT OF THE REQUIREMENTS

OF THE COURSE OM-1005 OPERATIONS MANAGEMENT
21 SEPTEMBER 2021
 2

Contents
1. Executive Summary…………………………………………………………………………………………………...3
2. Ecommerce Industry Overview ………………………………………………………………………………….4
3. Amazon Overview ……………………………………………………………………………………………………..7
a. ABC Analysis (applied to business model) ……………………………………………………….8
4. Quality Function Deployment …………………………………………………………………………………….9
a. House of Quality …………………………………………………………………………………………...11
5. Sales/Demand Forecasting ………………………………………………………………………………………13
a. Simple Moving Average ………………………………………………………………………………...14
b. Weighted Moving Average……………………………………………………………………………..15
c. Exponential Smoothing………………………………………………………………………………….16
d. Linear Regression Forecasting……………………………………………………………………….17
6. Recommendations … ……………………………………………………………………………………………….20
7. Appendix…………………………………………………………………………………………………………………21

Executive Summary
The purpose of this project is to analyse current issues prevailing in electronic commerce
industry, mainly with respect to operations management and try to address the challenges of
order management, demand fluctuation and overall services of e-commerce by providing
recommendations using quantitative and qualitative techniques such as Demand Forecasting
and Quality Function Deployment.

The report commences with the overview of the E-commerce industry and Amazon and the
operational business challenges faced by them. It then delves into the technique of Quality
 3

Function Deployment (QFD) to understand the priorities for Amazon to allocate right financial
or people resources to the processes.

Next section of the report concerns with Sales and Demand Forecasting technique using annual
sales data of Amazon for last 17 years and forecasting the sales for the next year or for near
future depending upon the accuracy of the forecast. Comparison of various forecasting
techniques such as Simple Moving Average, Linear Regression Forecasting is done and the best
method is recommended. This forecast can be used by Amazon to determine business
strategies for various departments within the company.

E-commerce Industry Overview

The electronic commerce in India (& overall, in the world) is gaining momentum and has a lot
of potential to grow primarily due to retail logistics being a key enabler for the industry.
Increase in internet penetration and mobile usage has led to people choosing e-commerce
platforms over brick-and-mortar stores and placing orders from the comfort of their homes.

As can be seen in the image 1 below, the ecommerce retail logistics market is dominated by e-
tailer captives1 like Amazon. These players have their own in-house logistics infrastructure to
control the supply chain and to give better delivery service. The conventional and long-
established retail players have also started their delivery services due to heavy competition in
1
E-commerce retail logistics in India, Driving the change, KPMG, May 2018
 4

the market. They have decades of experience and networks of suppliers; however, they have
not been able to capture large share of the market as e-retail logistics requires lot of technical
capabilities. As logistics enables the e-tailers the speed and competitive advantage (although it
may have impact on delivery costs), lot of investments in India from abroad are for the purpose
of strengthening logistics.

Image 1

Market Share of ecommerce logistics players in India in 2018 by type

23%
28%

Sources:
KPMG: 49%
Spire Research and Consulting:2018

Traditional Players Captive Logistics Arm E-commerce focussed players

The image 2 below shows the mechanism that is followed by e-commerce players while
delivery the good to the customer destination. Initially, the package is collected from the
original seller and then depending upon the business model, it is then sent to the fulfilment
centre or warehouses. Whenever the customer places the order, it will be picked from the FC
and forwarded to the centre where they are categorized as per the address or location of the
customer. Generally, the next process would be to use air or surface transportation to dispatch
the parcel to the main hub and lastly, to the delivery station near customer’s location.

Image 2

Reverse Logistics or returns have become a major concern for the ecommerce sector. The
reason for this is because it increases supply chain costs significantly as the operations are
predominantly designed for the forward delivery and not the reverse. This is the reason many
players keep the option open for customers to cancel the order without charging any fee before
the shipment is processed (First, the order is received, then prepared and lastly processed).
 5

Image 3 below shows the major costs that the e-tailers incur 2. As the costs of logistics are
increasing, many companies are facing the pressures to improve their operations and reduce
their costs by focusing on land transport instead of air transport. Also, there are changes in the
taxation laws in India, such as introduction of good and services tax, which has taken away
some tax benefits for companies and made them increase the distribution of their warehouse
facilities to multiple locations. This may also reduce the returns from customers as closer these
hubs are to the customer, faster would be the delivery and higher chances of customer getting
the delivery in promised time.

The next stage of growth for e-commerce retailers is most likely going to come from tier 2 and
3 cities. Also, customers who initially used to prefer to purchase bulky items like furniture etc
by going to the store now prefer purchasing online which has encouraged e-tailers to make the
delivery for these items as well.

Image 3

E-commerce retail transportation cost split

2% 8%

Processing First Mile

40%

(40-45%)
Last Mile Line Haul

(45-50%)
Sources: 50%
KPMG: Spire Research
and Consulting:2018

Business Challenges facing the industry:

Image 4

2
E-commerce retail logistics in India, Driving the change, KPMG, May 2018
 6

Amazon Overview
Amazon is currently the world’s largest ecommerce company started by Jeff Bezos in 1994. It
started initially as a book retailer and then increased its presence in many segments and
 7

markets to sell a very large category of products now3. Amazon mainly operates on a kind of
hybrid model i.e., fulfilled by Amazon model which means that its seller partners actually
purchase any kind of inventory and then store it in Amazon FCs. So, all the further process of
checking for quality and shipping etc. is taken care of by Amazon. Operations is at the core of
Amazon’s business model. It has global fulfilment centres as well as warehouses to manage
inventory and customer demand expectations.

Following are the different types of Fulfilment centres4 Amazon has:

 Fulfilment centre (sorted):

o These centres generally have very large space of almost a million square feet
and employs large number of employees. Sortable items such as books etc. are
stored at these centres that are then packed and shipped. Also, these centres
have robots to increase the operational efficiency.

 Fulfilment centre for bulky items (unsorted):

o These centres have a slightly lesser space and lesser number of workers than
the sortable centres and are used to keep, pack and ship the bulky items like
furniture or heavy electronic items etc.

 Centres for sortation:

o At these centres, the inventory is categorized according to the ultimate location
or the place of the delivery and it is also packed on to the trucks for fast
delivery. This is why Amazon has been smoothly able to make deliveries on
Sunday as well.

 Centres for keeping high demand goods:

o These centres are used to keep the large stock of the items that characterise
high demand and are anticipated to sell quickly.

 Special Fulfilment centres:

o These centres may house the type of items that need special type of storage or
the items that are mainly sold seasonally or on special occasions.

 Dispatch Centres:
o Orders are prepared for final delivery at these buildings and allow for faster
fulfilment of customer orders.

Under this business model of third-party online platform, Amazon takes orders, ships products,
handles sales returns & customer queries in addition to advertising and product promotion.

Amazon also uses various methods of transportation such as a cargo airline, personal vehicles
of independent delivery personnel, its own vans. Now it is also exploring the delivery service
through drones.

3
https://en.wikipedia.org/wiki/Amazon_(company)
4
https://www.aboutamazon.com/news/operations
 8

ABC Analysis for Amazon5:

ABC Analysis (Applied to Business Model)

Revenue Breakdown in 2017

Business Segment % Segment % Total % Total Revenue

Revenue Category segments
Online stores 60.95% A 17% 60.95%
Third Party Sellers 17.90% B 33% 27.72%
AWS 9.82%
Physical Stores 3.26% C 50% 11.34%
Subscription Services 5.46%
Advertising and other 2.62%
services

This Analysis clearly shows the most important business segments for Amazon and the
management should focus on then them accordingly

Quality Function Deployment

Quality function deployment (QFD) is a process to take into account the voice of customers and
use their feedback to improve technical characteristics and operations required to
build/design that product or a service6. The ultimate objective of this method is to prioritize

5
“Operations and Supply Chain Management; Chase, Shankar and Jacobs”
6
“Operations and Supply Chain Management; Chase, Shankar and Jacobs”
 9

the design characteristics of a product or a service that would meet the requirements of the
customer. House of quality is a matrix that forms the major part of this model.
The stages of the method for improving the quality of Amazon services are as mentioned
below:
1) Identify the issues and needs of a customer for services provided by Amazon

Customer service
centre responds slowly Theproduct delivered
Order stock became
to queries. Amazon is not as per the
unavailable after
representative call the desciption or image on
placing the order
customer after few the website
days.

The packaging of the The product is not in a

returns process takes a
product is damaged good condition and is
lot of time and not
when received by the damged duting
convenient
customer transporation

The information of the

product on website is
not detailed (e.g. the
size of the apparel is
missing)

2) Build the list of important factors from the issues mentioned above and also list the
technical specifications that they may relate to. Technical specification would be:

Order Monitoring or
tracking mechanism for
checking goods packaging of
team(fulfilment recoding order
before delivery goods
centres play a major times
role)
up to date
complaints
description call centre inventory
reolution
details of product operations information/man
mechanism
agement
3) Take customer survey and identify the importance7 of each factor to the customer.
Primary research was conducted to know the voice of the customer regarding the
quality of service and issues in electronic commerce industry. Responses of total 108
customers (convenience sampling) were recorded through google form (refer
Appendix). For further calculations, mode or most occurring value is considered.

7
https://quality-one.com/qfd/
 10

Image 5

Customer Importance Rating

70

60

50

40

30

20

10

0
] ] ] ] d] d] ] s] s]
e ss nt ili
ty on ne ne er nt nt
sl ie iti m ai ai
i ve
n l ab nd ti o ti o
s to pl pl
e
on ai co en en cu m m
tim sc av od m m g co co
er
y si em go tio
n
e rly
m
in er er
l iv es it in ir p op or om om
e oc re
d d sc pr nf st st
rd pr
de re de re ut
i c u c u
rd
e rn r l vi e e ta ho to of
[o tu [O de th c i t s e on
r[ e e r du w on uti
ar s pe pro l ed sp ol
s a l r e s
oo
d e he ce re
ar tt an ick ick
[G d s ou t c
[Q
u u
oo ab no [Q
[G i ls re
et
a sa
[D der
r
[o

Rating 2 Rating 3 Rating 4 Rating 5

4) Build the interrelationship and correlation matrix of the house of quality.

Interrelation matrix deals with identifying which factor from customer voice survey
relates to which technical or service specification so that it can be improved. This matrix
also involves the customers’ ratings for the competitor product for the same factor.
Although this may not help the company identify internal priorities, however, it
provides additional dimension to the analysis and help understand where the company
stands as compared to that of its competitor.

House of Quality
Image 6
 11

Type of Correlation
-+ + = Supportive
+ -
- + - - = Impeding
- + - -
+ - +- - -
+ - + - - +
+ - + + + + -

recording order time(order management software)

information on stocks/inventory is up to date

checking goods before delivery

complaint resolution mechanism

descrption details of product
order monitoring(tracking team)

call center operations

packaging of goods

Improvement Ratio
Competitor Ratings
Amazon Ratings
Importance
Customer

Weight
order delivery time is less 9 9 3 3 5 4 4 1.25 6.25
return process is convenient 9 9 4 4 3 1.00 4.00
goods are delivered in good condition 9 9 9 5 4 4 1.25 6.25
goods are as per the description mentioned 9 9 9 5 3 3 1.67 8.33
ordered item availability 9 9 4 3 4 1.33 5.33
orders are not cancelled without informing customer
9 3 5 4 4 1.25 6.25
details about the product are properly mentioned 9 5 4 4 1.25 6.25
quick response to customer complaints 3 1 9 5 3 3 1.67 8.33
quick resolution to complaints 3 1 9 9 5 3 3 1.67 8.33
Total Weighting 342 56.3 167 75 131 186 111 66.8
Relative Weighting 30% 5% 15% 7% 12% 16% 10% 6%
Priority Rank 1 8 3 6 4 2 5 7

9 - (Stong Relationship)
3 - (Medium Relationship)
1 - (Weak Relationship)

Formulae used in the above matrix are:

Improvement Ratio=Customer s ' current rating ¿ Amazon ¿ that factor ¿ '

Customer s importance rating ¿
'
Weight=Improvement Ratio∗Customer s importancerating
Total Weighting=∑ ( Interrelationship coef ∗Weight ) for all parameters

5) Calculate and identify the technical or service priorities to increase customer

satisfaction.
 12

The priorities that resulted from the QFD process are as below from highest to lowest:
Image 7

1. order monitoring or tracking

2. call centre operations

3. checking goods before delivery

4. product description on site

5. complaint resolution

6. packaging of goods

7. inventory information

8. recording order times

Conclusion:
Based on the Quality function Deployment analysis done using the voice of customer on
service quality parameters, quality parameters like monitoring and tracking of orders,
quality check before the final delivery, call centre services have to be improved so that
they meet the expectations of customers. Amazon senior management can utilize this
analysis to allocate the optimal resources for different processes within the company so
that improvement ratio is achieved.

Sales Forecasting
 13

Forecasting of demand or sales basically deals with using predictive analytics to anticipate the
demand for products and services of the company in future to optimize the supply chains and
take precise business decisions.

Problem Statement
The goal is to forecast annual sales of Amazon in 2021 and near future from the historical
annual sales figures available. Sales are highly correlated with the demand or the orders that
Amazon receives. Hence, this forecast would help the management to plan operations along
with marketing, sales and hiring etc. in a structured manner.

Data Description
Annual Sales data of Amazon is available for the year 2004 to 2020. There are no missing
values or any outliers and the data can be utilized in a readily available format. The data is
basically a time series with 17 data points. However, as the quarterly or monthly data was not
available, it does not have seasonality or cyclicity characteristics. It mainly displays a trend
component as shown in the diagram below.
Image 8

Simple Moving Average

 14

This technique uses the average over a small subset (n periods) of data from the whole data set
to analyse the trends and patterns that occur in the data points. It is fairly straightforward
technique and easy to implement as well.

Image9

5-year MA 3-year MA
MAD 67.19 45.14
MAPE 49.53% 38.26%
 15

As can be seen from the figure and MAPE error, as the number of periods is increased, the error
of the forecast increases. This is because the series is growing exponentially and the more
weightage is given to the past, lesser would be the accuracy of the technique. Overall, 3-year
MA technique slightly performs better than 5-year MA.
Next technique of weighted moving average tries to somewhat resolve this issue by giving the
user the flexibility to choose weightage to be given to the past values.
Weighted Moving Average
Image 10

MAD 34.25
MAPE 28.91%
 16

From the above analysis, weighted moving average slightly takes the edge over 3-year MA
technique as the MAPE has lowered. However, still the absolute value of error ~29% is very
high.

Exponential Smoothing

Image 11

ES alpha ES alpha ES alpha ES alpha ES alpha

  =0.1 =0.5 =0.6 =0.8 =0.9
MAD 75.86 37.95 33.78 27.78 25.56
MAPE 63.04% 34.58% 31.06% 25.82% 23.82%

As we increase the alpha value closer to 1, the algorithm learns the sales trend faster and MAPE
error of the forecast also decreases. However, as can be seen from figure still the actual sales
curve and exponential curve with alpha = 0.9 value deviates by a significant amount and ~24%
error is very high in absolute terms.

Linear Regression Forecasting:

 17

Instead of taking actual year values, we take dummy variables for creating a more convenient
correlation estimate with the sales figures in this technique.

Image 12 Image 13

As can be seen in the figure, the time series is exponential and linear regression forecasting
applied directly fails to capture this relation and results is a huge error in the forecast. But this
can be taken care of by transforming the sales values to the logarithm of the sales values to
make the relationship between dummy year variable and log(sales) to be linear.
Image 14
 18

Image 15
 19

SUMMARY OUTPUT

Regression Statistics  
Multiple R 0.998
R Square 0.997
Adjusted R Square 0.996
Standard Error 0.077
Observations 17

Best techniques comparison

3 Y e a r M o v in g A v e r a g e
M A P E : 3 8 .2 6 %
2 0 2 1 S a le s F o r e c a s t: $ 2 9 9 .8 2 B N

W e ig h te d M o v in g A v e r a g e
M A P E : 2 8 .9 1 %
2 0 2 1 S a le s F o r e c a s t: $ 3 4 9 .6 4 B N

E x p o n e n tia l S m o o th in g ( A lp h a = 0 .9 )
M A P E : 2 3 .8 2 %
2 0 2 1 S a le s F o r e c a s t: $ 2 8 6 .2 5 B N

L in e a r R e g r e s s io n fo r e c a s tin g ( tr a n s fo r m e d Y )
M A P E : 5 .6 4 %
2 0 2 1 S a le s F o r e c a s t: $ 5 0 2 .4 1 B N
 20

Recommendations

Based on the above analysis, linear regression forecating technique with transformed
(logarithm) sales variable performs the best and should be utilized by Amazon for future
demand forecasting, until and unless the performance of the model deteriorates.
Sales & Demand forecasting can be used as a basis to develop annual operating plans using
chase and level strategies. It can be also be utilized to analyse the inventory level needed to
meet the demand of the customers.
It was very clear from the Quality Function Deployment Analysis that the order management
and call centre service is at the core of delivering value and experience to the customers.
Demand forecasting can also be referrenced to know the optimal number of employees and
resources allocated to call centres to meet the demand along with optimizing the other
operations of Amazon.
 21

APPENDIX
22
23
 24

R Code

A part of the R (for exponential smoothing) code used for the project is as below for
illustration:

### Exponential Smoothing Curve

df_temp <- read.csv("~/Exp_Smooth.csv")

p4 <- ggplot(df_temp, aes(x=Year, y=Sales)) + geom_line(aes(color="#69b3a2"), size=1.5) +

geom_point()

p5 <- p4 + geom_line(data = df_temp, aes(x = Year, y = alpha1, color="orange"), size=1.5) +

geom_point(data = df_temp, aes(x = Year, y = alpha1))
 25

p6 <- p5+ geom_line(data = df_temp, aes(x = Year, y = alpha2, color="green"), size=1.5) +

geom_point(data = df_temp, aes(x = Year, y = alpha2))

p7 <- p6+ geom_line(data = df_temp, aes(x = Year, y = alpha3, color="yellow"), size=1.5) +

geom_point(data = df_temp, aes(x = Year, y = alpha3))

p8 <- p7+ geom_line(data = df_temp, aes(x = Year, y = alpha4, color="red"), size=1.5) +

geom_point(data = df_temp, aes(x = Year, y = alpha4))

p9 < -p8 + geom_line(data = df_temp, aes(x = Year, y = alpha5, color="light grey"), size=1.5) +
ggtitle("Exponential Smoothing") +
xlab("Year") +
ylab("Sales(in billion dollars)") +
geom_point(data = df_temp, aes(x = Year, y = alpha5))

p10 <- p9 + theme (legend.position = c(0.1, 0.95), legend.justification = c("left", "top"))

p10

Excel File for forecasting Methods and QFD

Responses of customer survey