26
1 Evgeniya Burakova Masters of E-Commerce Business Analytics Portfolio: Exploration of Business Analytics Solution

Business Analytics Portfolio

Embed Size (px)

Citation preview

Page 1: Business Analytics Portfolio

1

Evgeniya Burakova

Masters of E-Commerce

Business Analytics Portfolio: Exploration of Business Analytics Solution

Page 2: Business Analytics Portfolio

2

Abstract

Innovations in BI from various providers broaden analytic capabilities enabling almost every department of an

organization to participate in the extraction of relevant information. The data involved in data mining goes through

the process of extraction form a dataset, integration, transformation into an understandable structure for future use.

Due to the large number of sources together with the diverse nature of data different analytical tools are applied. It

is important to recognize that data coming from crowdsourcing, transactions, clicks (stream mining), and client data

is transformed into knowledge and therefore could be used in building future strategies and explain certain events.

Some analytical tools are lack of certain functions and represent certain type of data worse than others. This portfolio

comprises the screenshots of the examples how the software operates and what information could be derived.

Content

Sap Lumira

Page 3: Business Analytics Portfolio

3

IBM Cognos

BusinessObjects Analysisi

Business Objects Explorer

Business Objects Design Studio

Watson Analytics

Sap Infinite Insight

Page 4: Business Analytics Portfolio

4

1.1 SAP Lumira

The user is able to create customized dataset and include the attributes he is interested in by integrating several datasets.

The ability to acquire data from multiple excel spreadsheets, merge those files, prepare and start analysis can provide rich insight

for a business consultant. The reason for combining the dataset could be the missing values of one of the datasets and therefore

not enough “foundation for analysis”. On the other hand, once the data is combined Lumira smooth some picks and negative

instances because of the data scope. Before data is uploaded the system asks for the ambiguity to be resolved, so the user controls

the blending.

Figure 1-Two datasets combined

Page 5: Business Analytics Portfolio

5

Geographical hierarchy could be

displayed on the map of the USA

(Figure 2). The bubbles represent

the roaming vs non-roaming in the

different states. By clicking on

indexes on the right certain types of

revenue (that come from roaming

vs non-roaming) will be

highlighted while the opposed

faded. This will show the user the

state with the largest portion of the

revenue (the highest degree of the

circle).

Figure 2-Revenue by Regions

Page 6: Business Analytics Portfolio

6

This visualization shows PostpaidRev is

contributing most to revenue and Data plan is the

most popular. Understanding product

segmentation is important when developing value

proposition as marketing mix (product features,

price, place, and promotion) will differ between the

products. As we can see telecommunication

strategic focus should be on the product with the

highest revenue. Users acquire data plans as

oppose to messaging or talking plans as most of the

applications allow to do both talk and message but

those aps require internet.

When it comes to “place” component of marketing mix the strategic location should be considered based on the highest

consumption and revenue. The geographical location tool reflects the revenue by states (Figure 4). The next step is to find

minilmal valuable product , i.e. the startegic locations by ranking top 5 states (Figure 5).

Figure 3-Stacked chart for plan types by revenue

Page 7: Business Analytics Portfolio

7

The ranking could be applied to the types of the plan and presented as a score card. The canvas and the size of the blocks represent

the contribution to the revenue providing the manager informatio about the most popular plans.

Figure 4-Revenue by states Figure 5- Top 5 states by revenue

Page 8: Business Analytics Portfolio

8

Formation of clusters from a large number of instances enables a user to apply similar strategies. Having distinct groups make the

management easier. We can then explore different parametres of the cluster such as Cluster Density and Distance to understand

distribution of the clusters (Figure), Feature Distribution (Figure) or Cluster Center Representation to check on which attributes certain

cluster have high/low value (Figure)

Page 9: Business Analytics Portfolio

9

The scatter charts of store clusters plotted between various pairs of dimensions is useful for comparison purposes.

Page 10: Business Analytics Portfolio

10

The dataset on sales can be used to derive such information as material with no sales data. Total value for different dimensions or get

information on some unique value by applying according filtering options. The comparison between two values could be easily

visualized with either a line chart or column chart. The graph could be also helpful to show dramatic changes or build a scatter plot for

all the values achieved in order to see possible trends (by quarter). Sorting and ranking enabled to identify smallest ad highest results or

top 3 values, for example.

Page 11: Business Analytics Portfolio

11

1.2 IBM Cognos

Figures 14a and14b represent sales by customer types. The new crosstab cab be easily created by dragging and dropping the dimension

as column/rows depending on which aspect of the sales the user wants to concentrate. For this scenario business analytics can sort the

sales in descending order to see which customer type brings the highest revenue to the business. What makes this board even more

informative is the margin revenue vs cost and the margin score. This additional column with colored “signal” provides the user with the

quick overview of the areas that require more attention. As the revenue is sorted by descending order and the scorecard allows to

immediately notice negative values for retail channel (high costs vs sales), we can see that the biggest problem is with the “Retail”

customer type despite its highest revenue production as compared to another channels. We can use bar chart to visualize the findings

(Figure 14b).

The introduced elements of IBM Cognos BI are used to understand the comprehensive corporate performance management solutions.

One of the prerequisites for working with the software is to understand organization's reporting needs: whether it is about visualizing,

predicting. The color serve as alerts for the senior managers to quickly identify the areas that need more consideration. The

storyboards could be used to present monthly sales of materials. With column charts, pie, crosstabs, area, point charts, tree maps a

user can discover which distribution channel has the weakest performance in terms of the margins (cost vs revenue) for example.

Dataset for exploring the tool contains information about material, customers’ types, year, territory, sales channels which serve as

dimensions of the dataset and revenue/sales/quantity is set as measures.

Figure 14 a- Sales by customer types Figure 14b- Sales by customer type (bar chart)

Page 12: Business Analytics Portfolio

12

The next storyboard reminds of the matrix where product types are reviewed customer wise. The crosstab contains many figures

therefore again margin score are good alerts of the performance. Below the crosstab are the independent graphs for each customer types.

Each product`s performance is reflected by the color which is assigned and indexed on the right of the graph (computers & tablets-navy,

mobile-yellow etc). User can select measure by which the sales could be compared and by changing to score, ie margins, the sales

performance could be shown.

The filters showed that mobile sold online was more successful compared to retail marketing channel as it, first, requires no labor costs

and second, due to the nature of the product (as customer is more inclined to purchase devices online since extra research in terms of

the alternatives, prices might be needed). This might be the one of the rational a marketing analyst will provide to complement the

information derived from Cognos. The graph helps to analyses which product were the most/least successful for which channels.

Surprisingly, portable electronics that were successful for all distribution channels did not have successful sales figures for retail. The

Figure 15a- Sales by customer and product type

Figure 15b- Sales by customer and product

types measured in scores

Page 13: Business Analytics Portfolio

13

managers might continue to track the performance and the case the situation does not change the expanses spent for portable electronics

should be transferred to computer and tablets segments which has substantial revenue losses. Marketing campaigns might be included.

What makes Cognos even more powerful is the capability to predict future trends/ sales. Based on the values of what is going on in the

business one year to predict how the business would perform considered adjustment in sales/cost a manager can develop marketing

tactics to achieve forecasts.

Right allocation of budget and achievement of the revenues that we set as the predictions will change the performance of not only retail

but other channels as well (we can see that the revenue has increased above 6 m among all channels).

Figure 16-Sales prediction

Figure 17-Sales

prediction

Page 14: Business Analytics Portfolio

14

1.3 Business Objects Analysis

Figure 18 provides an overview of

the performance for each year.

Having sorted Sales Quantity in

descending order, gives a user an

information about the year with the

highest sales quantity (2007 in this

case) as compared to the lowest in

2009.

SAP Business Objects Analysis, a plug-in for Microsoft Excel and PowerPoint, makes the data analysis more profound and pretty

easy due to Excel`s familiar to an average business user design. It connects to query designer to directly extract data from data

warehouse so the BI content is pre-defined in Microsoft Excel or Microsoft PowerPoint. Under design panel a user can find

analysis, information or components. Business Analyst can define conditional formatting or conditions by measures. OLAP

techniques such as filtering, slicing and dicing are used to perform data analysis on:

• The city/department with the highest revenue and visualization

• The year with the highest sales quantity

• Department with the lowest sales

All was done by applying filters and sorting by measures.

Figure 18-Financial information by years

Page 15: Business Analytics Portfolio

15

The next figure presents the

table that contains info on

year, material code and

material name as

Dimensions and financial

data as Measures. Sorting by

measures allowed to identify

not only the year but also the

material with the poorest

performance in 2009.

Figure 19-Sorting the products by sales

Page 16: Business Analytics Portfolio

16

Excel Analytics helps to understand the overall revenue of

each products during 6 year period. Again by sorting in

ascending order we find that water bottles had the lowest

revenue. Year and total revenue by the countries are presented

in a line chart (Figure 21) and we can clearly see that though

starting almost with the same revenue 2007 Germany

improved its overall results over 4 years

Figure 21-Overall results by country and year (line chart)

Figure 20-Product by revenue

Figure 22-Overall results by country and year

Page 17: Business Analytics Portfolio

17

1.4 Business objects explorer

Page 18: Business Analytics Portfolio

18

Page 19: Business Analytics Portfolio

19

1.5 SAP Business Objects Design Studio

The software provides an environment to create apps and dashboards in Design Studio is a powerful yet easy to use platform. In the

business world where mobile manager requires immediate response business object design is useful as it supports both desktop and

mobile theming. Since no additional download or installations but an HTML5 capable browser are required to open a dashboard

either on smartphone or tablet, companies find Business object design cost effective and hassle free tool. Once the properties are set

up for the future application, sales dataset is uploaded in object design

Page 20: Business Analytics Portfolio

20

We have to determine the layout of

the future app and include objects

such as title, graph, switching

between currencies and filtering

options. The following is possible by

dragging and dropping components

from the left-hand menu. Having

created this interactive app a

manager now can make it accessible

in the web. Design studio allows a

user generate the QR code for the

link

Figure

Figure 5- App design development

Figure 6-QR code generation

Page 21: Business Analytics Portfolio

21

The user friendly interface is created considering CRAP

design principles. We used the default black theme for

mobile app. We can enable the user to apply the filters on

customer/material/sales organization or check the key

figures. In the scenario when comparison of revenue

between customers in Germany North is requested, key

figures is selected as revenue and sales organization as

Germany North to check the most successfully sold material

with the most frequent customer

Figure 7-Global Bikes App

Page 22: Business Analytics Portfolio

22

1.6 InfiniteInsight

We used InfiniteInsight to understand the contribution of variables on the propensity to have an accident. Since the demand for the

insurance service is high it is worth identifying those triggers and perhaps customize the service according to the demographics of the

customers or other attributes. As we can see on Figure there is a negative correlation between the number of children in the family and

The software allows to understand and predict a phenomenon as well as describe a data set, by breaking it down into homogeneous

data groups and clusters. Association rule is used in order to determine basket analysis which could enhance customer relationship

as a vendor will be able to offer better deals or launch a marketing campaign based on the information derived. InfiniteInsight can

also support future decision making by matching the attributes of a new instance with the past data in order to see the probability

of the vent happening.

The examples below illustrate the application of the tool on insurance industry, retail sector

Figure 9-Probability of an accident (children as an indep variable) Figure 8-Probability of an accident by different variables

Page 23: Business Analytics Portfolio

23

the chance of the accident occurrence (as the number of children increases the risk of accident decreases). Therefore, parents in larger

families tend to be more cautious drivers. Figure explicitly reflects that the number of children is the most determining factor.

The third most influential factor is gender. We can then understand that

influence in more detail. The well know myth that woman cause traffic mess

could be destroyed with this graph. Female are actually safe drivers as

compared top male.

The software allows to construct the decision tree using the

attributes and see the probability of accident occurrence. We used

the factors analyzed above (children and gender) and now can

make a statement that a woman with 1-3 children is in a lower

risk group (6.16%) as compared to a man (12.19%).

Unfortunately, we cannot drill down using more than two

attributes. The simulation function allows to make quick decision

based on the known statistics and the personal info of a client.

The probability for a claim of a student of 22 year old man driving

a Sedan without any children is 19.05 % (the driver belongs to

the group “Claim=Yes” is 19%).

Page 24: Business Analytics Portfolio

24

The software allows to upload a new dataset of customers and using the

prediction algorithm calculate the probability of claim for those

customers.

The examination could be applied to other industries as well. For

example a bank analyst can find the factors that contributed to customer

retention. We can explore how a region influence on the propensity to

leave a bank

Page 25: Business Analytics Portfolio

25

The recommendation based on the algorithm

function determibes the best option for the

customer.

In the predictioN InfiniteInsight shows the

accuracy of the model.

The results from the analysis could be used to generate an

executive summary in the form of PPT.

Page 26: Business Analytics Portfolio

26

With the confusion matrix function we can predict the correct outcome by setting the desired values Imagine you have budget to contact

10% of your contacts. If you contact your customers randomly, you will only reach 10% of the customers that purchase. However,

with a good predictive model you will be able to increase the success rate.

Infinite Insight allows to conduct Link Analysis on

the Members and Products. We can either suggest

what might appear in the shopper basket by setting

customer Id or find the most complimentary

products. We can see the shared community

(Figure). People are likely to buy panini with coke

(Figure)