In the following report, it will be taken
in examination the importance of data gathering for Amazon, an established
company leader in the online commerce. Through the use of complex algorithms
and collecting data from every user’s visit, Amazon is able to empower its
prediction and recommendation algorithm, becoming aware of customers
preferences and possible needs, with the purpose of increasing sales and
profits. Due to the fact of creating personalised recommendations for each
user, using a new version of collaborative filtering algorithm (called
Item-to-item), the Seattle based company has been able to rise its revenue. It needs
to be mentioned that almost the 35% of its earnings is generated by the
customized suggestions. Lastly will be taken under consideration the privacy
protection when dealing with this big amount of sensitive data.

 

 

Known as the biggest market place and
online retailer, Amazon (Amazon.com) was originally founded by Jeff Bezos in
1994 as an online bookstore. After a continuous and gradual sales increase, he
started to widen the market horizons expanding the wholesale business by enlarging
the products offered. Browsing on Amazon.com website, without even creating an
account, is possible to seek millions of different merchandises such as books,
food, electronic devices, home furniture and a cloud service is provided as
well.

Once visited the website and comparing
Amazon prices with other retailers, the need of opening an account begin to be
felt. Even if you did not yield to the temptation of opening the account,
Amazon already started to collect Data about you and your researches. In fact,
from the very first moment you access to the webpage from your browser through
computer or mobile device, cookies start to be collected acquiring all your
info following your Internet Provider address creating a unique identifying
number for your devices. In this way, every time a potential customer is coming
back on the website, Amazon will create a “personalized” home page based on the
previous researches, tempting you to open the account for the first purchase.

After being convinced of opening the account, customers provide voluntary the
basic nominal data such as First name, last name, email address and password in
order to create a unique profile.
This profile will be automatically connected to the data collected before
through cookies (if using the same device/browser) and on the home page will be
shown some of the previous products researched and personalized recommendations.
Once customers are ready to go ahead with the purchase, it will be asked to
insert more details like credit card number, telephone number and address.
Right before proceeding with the payment, Amazon will introduce several
different Items called “frequently bought together” in which the customer could
be potentially interested. In this way, Amazon uses the power of suggestion to
increase the customer shopping satisfaction encouraging buying, leveraging on
the impulse (Krawiec,
2017)gb1 .
All of this process has been possible because Amazon is collecting data through
some software tools such as Java Script since the very first visit on the
website, “collecting session information, including page response times,
download errors, length of visits to certain pages, page interaction
information (scrolling, clicks, and mouse-overs), and methods used to browse
away from the page” (Amazon, 2001gb2 ).
Furthermore, the Item-to-item collaborative filtering algorithm developed by
Greg Linden differs from the most common traditional collaborative filtering
and Cluster models due to the fact that it deals with linking similar items
rather than matching the user to similar customers. Combining the previous
information collected and the Item-to-item collaborative filtering, which is a
recommendation algorithm, Amazon.com is able to personalize the entire website,
recommendation lists and items for each customer using the items viewed,
demographic data and subjects of interest as well.
Collecting all this data from the purchases made by previous customers, the
algorithm assemble a similar-products table by selecting items that consumers
choose to buy together, increasing the recommended list targeting (Linden, Smith
& York, 2003gb3 ).
This algorithm contribute to an effective form of targeted marketing by designing
a customized shopping experience for each client, in fact it can be said that
almost the 35 percent of Amazon’s revenue is generated by its personalised
recommendations (McKinsey,
Meyer & Nobel, 2013gb4 ).

If the sale has not been concluded on site yet, Amazon recommendation continue
through their follow up marketing campaign off site. The first email will be a
general one with similar items that other customers have been flipping through.
Continuing tracking the user movement (time spent on the newsletter and opened
links monitoring the clicks), Amazon collect enough info to produce a more
detailed and targeted email marketing campaign for that specific user. The
following emails will start to introduce the possible items in which the user
could be interested to, combined to some related products frequently purchased in
a bundle, starting a cross selling operations.

After purchasing the product, amazon require more info about the shopping
experience and would be pleased to receive a review about the item previously
purchased. On the basis of the review left and the level of satisfaction achieved
(based on a grade of 1 up to 5 stars), Amazon collect more data about what the
user expected with the purpose to increase and improve the his next experience.
In this way, the algorithm working aside of an AI machine learning called DSSTNE (Deep Scalable
Sparse Tensor Neural Engine) gb5 is
able to predict what could be the best possible next purchase to propose to
that user because it would be more likely to click on it and buy the product
(Chung, 2016). Furthermore, this info will be used to increase performances and
recommendation for other customers, which bought the similar item. Throughout
this system, Amazon is perfecting its recommendation engine after any purchase
made by any customer in their database.

Talking about how much is important gathering
data for Amazon, in order to improve its service, it needs to be mentioned the
latest home brand products released such as Amazon Echo and Amazon Echo Look;
with these devices, it has been able to enter in our home and daily life as a
personal assistant for anything needed. Querying the devices about weather
forecast, recipes, general knowledge facts on a daily basis, helps Bezos’
company gain a deeper knowledge about the things you are interested in, in
particular with the sales driven device Echo Look. This latest product has a
built in camera, which allows you to take a picture or video of your outfit comparing
it with the previous ones and suggesting you which one is more fashionable. The
suggestion is generated by the alliance of a machine learning algorithm, a team of
“fashion specialistgb6 ”
and other customers that can rate the outfit in the picture. Here again Amazon
is able to know the preferences of its customers improving its services suggesting
related clothing in different colours or models (Goode, 2017).

One of the biggest ethical issue concerning these kind of listening personal
assistants is generated by the audio recorded by the device and sent to the
server when triggered with the “wake” word in order to produce the answer
needed. What if these devices are always listening and recording even if not
triggered? The Seattle-based company declared that will never share the data
collected by these devices to third party but there’s been a case where amazon
echo has been used by the police to solve a murder case in Arkansas because
possibly recording key facts during the night of the homicide (Sampathkumar, 2017gb7 ).
Another significant problem is that with the increasing business of the
Internet of Things (IoT) a big amount of devices will be exchanging data among
them and hackers could possibly hijack all these new devices stealing data, recording
audio or video without that consumers are aware of it.

When
handling this big amount of data collected through the website and different
devices, the privacy of such sensitive info became an important factor for
customers and the company.
As mentioned before, Amazon state that whatever is collected, stored and
processed through its devices isn’t released to third party and it is kept in
their cloud server as long as the user decide to delete the query history or
pictures/videos. The same regulation is not applied to the data gathered
through the website such as account data (customer first name, family name,
email, address date of birth), cookies and search history etc. Amazon works
along with other affiliated businesses like Marketplace sellers with which
share customers info in order to proceed with the transaction.
Other data is shared  with companies or
individuals : “fulfilling orders, delivering packages, sending postal mail and
e-mail, removing repetitive information from customer lists, analysing data,
providing marketing assistance, providing search results and links (including
paid listings and links), processing credit card payments, and providing
customer service. They have access to personal information needed to perform
their functions, but may not use it for other purposes.” (Amazon, 2017)gb8 .
All the information provided by customers to the company are encrypted using a
Secure Sockets Layer (SSL) software which protect the data transmission, ensure
the server you are connected to is actually the correct server and confirm that
the data that is requested or submitted is what is actually delivered.

 

In
conclusion, Amazon base a big part of its revenue over the data that could
possibly collect directly or indirectly from customers and potential client,
strengthening its algorithms and AI Machine Learning with the purpose of having
an effective targeted recommendation system to increase sales and profits,
securing at the same time the sensitive info gathered.

 

 gb1http://rejoiner.com/resources/amazon-recommendations-secret-selling-online/

 gb2 https://www.amazon.com/gp/help/customer/display.html/ref=footer_privacy?ie=UTF8=468496#GUID-1B2BDAD4-7ACF-4D7A-8608-CBA6EA897FD3__SECTION_87C837F9CCD84769B4AE2BEB14AF4F01

 

 gb3https://www.cs.umd.edu/~samir/498/Amazon-Recommendations.pdf

 gb4https://www.mckinsey.com/industries/retail/our-insights/how-retailers-can-keep-up-with-consumers

 

 gb5https://aws.amazon.com/it/blogs/big-data/generating-recommendations-at-amazon-scale-with-apache-spark-and-amazon-dsstne/

 gb6https://www.theverge.com/2017/7/6/15924120/amazon-echo-look-review-camera-clothes-style

 

 gb7http://www.independent.co.uk/news/world/americas/amazon-echo-murder-investigation-data-police-a7621261.html

 

 gb8https://www.amazon.com/gp/help/customer/display.html/ref=footer_privacy?ie=UTF8=468496#GUID-1B2BDAD4-7ACF-4D7A-8608-CBA6EA897FD3__SECTION_3DF674DAB5B7439FB2A9B4465BC3E0AC

Third party service provider