“Introduction to Business Analytics” held at the Indian School of Business on 24^th-June-2012 helped to demystify some common beliefs and preconceived notions regarding “Business Analytics”. The workshop revealed various interesting facts on the subject ranging from the underlying theories to innovative applications in various industries. Organised by Prof. Galit Shmueli, Professor of Data Analytics at the ISB, the session witnessed enlightening talks on the subject from eminent experts -

Mr.  Kuldeep Parikh, Strategy Lead, HP Global Analytics

Mr. Ajay Kelkar, Chief Operating Officer, HANSA Cequity

Mr. Kishore Rajgopal & Mr. Divyabh Mishra, Founders of CrowdANALYTIX, Bangalore

On a sunny Sunday morning a fully packed Khemka Auditorium witnessed CEOs, Directors, Entrepreneurs, and the students from ISB lending their ears to some of the finest speakers of the subject. The three hours of enthralling lectures, which were also streamed live to the students at the Mohali campus of the ISB, perfectly introduced the audience to the concept of “Business Analytics”.

The workshop began with an introductory lecture by Prof. Galit Shmueli, who provided a brief overview of the subject of “Business Analytics”. Prof. Shmueli explained how the term “Business Analytics” is being loosely used to refer to Business Intelligence, Business Modelling or Business Statistical Intelligence in the industry. She referred to the definitions of “Business Analytics” by SAS and IBM to explain the confluence of “Business” and “Analytics” for doing predictive analysis for the business. Highlighting the increasing importance of the role of Business Analysts in shaping the strategy of a company, Prof. Shmueli informed the students of the unexplored and untapped potential of “Business Analytics” in India.

A theoretical overview of the subject by Prof. Shmueli was followed by Mr. Kuldeep Parikh’s lecture on some key projects undertaken at HP Global Analytics. Mr. Parikh began his lecture with an allusion to Prof. Dan Ariely’s book “Predictably Irrational” and continued to explain the advantages of “Business Analytics” over intuitive decision making in rapidly changing business environments. He provided further insight on the subject by discussing real world examples of the various types of projects HP had undertaken to meet its internal strategic requirements.

The next lecture of Mr. Ajay Kelkar explained how Marketing could gradually transform into a more pull based approach and emerge as a service for the organizations. He went on to explain how data was increasingly becoming a key marketing tool for the corporate entities. According to Mr. Kelkar, for a company, “data equity” was not only becoming as important as “brand equity” but also helped enhance the latter further. He, however, noted that not all companies were able to effectively analyse the data to create better processes or to make better decisions.

The workshop was concluded with a very interesting lecture by the founders of CrowdANALYTIX- Mr. Kishore Rajgopal & Mr. Divyabh Mishra. CrowdANALYTIX is an online global community which invites analytics, data mining, and statistics experts around the world to brainstorm on business analytics problems. Mr. Rajgopal proudly discussed how they were able to find highly motivated people who would work on challenging problems and compete within the community to offer the best possible solutions. CrowdANALYTIX has its own way of recognizing the achievers who could range from young talented students to experienced analytics professionals. The founders of this innovative business model did not see their company as a competitor to analytics firms as they were able to promote the importance of analytics in business by offering high quality solutions at a very low cost.

The workshop helped the interested souls crack the jargon of “Business Analytics” and capture a glimpse of a gargantuan world of predictive analysis where analytical tools could perform – customer behaviour prediction, sales trend identification, marketing effectiveness study, financial analysis and much more. The widespread application of the subject unfolds a chain of opportunities for people who are seriously thinking of making a career out of this field. ISB offers two electives in this area – Business Intelligence Using Data Mining and Business Forecasting. Therefore, the interested students who have had a taste of analytical tools in their core courses and think that they have already mastered the science of “Business Analytics” are just in the words of Prof. Galit Shmueli – “Kids in the Toy Store”!

-Sammilan Dey

We would like to invite you to the 2nd Open Data Camp in India on Saturday, June 23rd, 2012, at the Indian School of Business, Gachibowli, Hyderabad.

This event is dedicated to all aspects of open data, from working with data, to getting it, and of course how to use it to create impact. This event is being organized by The Asia Analytics Lab @ ISB and the DataMeet group, an online group of data enthusiasts who hope to use data to create an impact in the lives of people living and working in India.

We hope that you can attend the event. Registration is compulsory to attend the event. Please register at http://opendatacamp-hyd.doattend.com/. If you would like to present, please send us the topic and a brief abstract. (e-mail to Arun_Kumar@isb.edu)

You can find more about the event @ http://odc.datameet.org/odchyd2012/

Regards,

OpenData Meet Organizing Committee

• What are the migration patterns across India?
• What are the drivers of migration in India?
• Do these drivers vary from state to state?
• What are the age groups of people who migrate due a certain reason?
• What do people do after they migrate?

[Click here to view the slides.]
The process of creating the visuals started with our initial prototyping of charts using a variety of visualization tools from basic Excel to more advanced R and to more interactive Spotfire, Tableau and Microstrategy. Our visualization partners at Gramener then used their proprietary tools to generate the final dashboards (final charts are shown in the slides).
In the planning are more visualizations for exploring the profiles (education, gender, etc.) of migrants, examining district-level migration, as well as visualizing migration using network graphs, connecting different locations. We might present these in the next Open Data unconference!

Map charts are data visualizations that combine a geographical map with a dataset that has information about geographical entities. For example, census-type data at the district level. The data are then displayed on top of the geographical map, such that coloring and other features can be used to represent a measurement of interest.

To create a map chart, one needs to obtain “shape files” which are a set of files with extensions such as .shp and .dbf. Shape files are available for purchase from various websites, and are freely available on others. One source that provides free shape files for many countries is DIVA-GIS. The administrative files for India include shape files at 4 levels:

Sub-district level map of India, generated from DIVA-GIS shapefiles

• files with “0″ refer to country-level, and will create a map of
• India as a whole
• files with “1″ refer to state-level, and will create a map of India with boundaries between states
• files with “2″ refer to district level, and will create a map of India with boundaries between districts
• files with “3″ refer to sub-district level — this will create a map with sub-district level boundaries, as shown in the image.

The choice of files to use therefore depends on the geographical level of your data. If the data are at the state-level, use the files with “1″.

All this is technically straightforward, but there is another issue that creeps up: What if the available shapefiles include some sensitive borders or disputed territories? Since the data analyst is the one obtaining the shapefiles (the software companies are smart not to provide them!), it is his/her responsibility to make sure that the shapefiles are appropriate for the intended audience.

The recent issue of Analytics (the INFORMS magazine) featured the first article in a series about “the rapid growth and widespread impact of India-based analytics”.  The first article, Emergence of global analytics centers of excellence (low bandwidth version here), is written by Indian-based HP Global Analytics vice president Rohit Tandon and  director Arnab Chakraborty. They describe in general terms the ways data analytics are creeping into various business sectors in India. One of their main “lessons learned” regarding leveraging analytics raises the critical need to

Create industry-academia partnership to groom the talent pool in universities as well as develop strong internal training curriculum to advance analytical depth

And this is exactly what the Asia Analytics Lab is about.

While it might be intuitive how management skills and knowledge are affected by the Asian context, how does the context affect data analytics? Is there really a need for “Asia Analytics”? Since data analysis involves understanding the source and context of how the data arose, the answer is absolutely. In this post I consider one example: Handling Missing Values.

Almost every dataset comes with missing values. Dealing with missing values is a critical initial step before any further analysis can be applied. There are a bunch of strategies for dealing with missing values, including removing records (rows) with missing values – this is the default of most software; removing measurements (columns) that contain many missing values;  imputing (fill-in) missing values using a variety of methods; and creating indicators of missing/non-missing to replace columns with missing values.

Choosing the right strategy for dealing with missing data requires understanding why missing values arise. For example, in data from a survey, why are many people not answering a certain question? Does the Western “no answer” notion apply to Eastern “no answer”? Does “no answer” in India mean the same as in Taiwan? If “no answer” is telling us something important (e.g., survey participants are hiding information), then creating a missing/non-missing indicator can be a useful strategy for exploring this hiding behaviour. But if “no answer” means “I don’t know but won’t admit to not knowing”, it is a different story altogether.

Apropos surveys: cultural conventions also strongly influence the way people answer on traditional 5-point and 7-point scales in terms of giving extreme ratings, understanding “good” vs. “very good”, not to mention different interpretations of notions such as “happiness”, “satisfaction”, and “friendship”. Special scales are therefore often devised (such as the Shagun scale for semi-urban/rural India). Country/culture-specific data collection mechanisms is a whole new story that also affects the resulting analytics. But that’s for another post.

Training algorithms on existing data is also language-dependent: the corpus of online text in some languages is still quite limited.