Analytics: Missing the woods for the trees

This article was first published in the Analytics India Magazine on June 23rd, 2015; Co-author: Sanjay Fuloria (Cognizant Research Center, Hyderabad)

http://analyticsindiamag.com/analytics-missing-the-woods-for-the-trees/

Big Data Talent Gap is a serious problem. Recognizing it, Universities are introducing courses on Analytics and Big Data. This has resulted in a larger supply of people who are well versed with data manipulation, handling and running codes. But, it has created a gap of another type. 

In an article on “Three Problems All Data Scientists Experience”, Drew Farris of Booz Allen Hamilton Inc. writes that “…problems go beyond technology and machine learning and are broadly encountered regardless of the task at hand: interpreting the problem, sourcing the data, and describing the outcomes”.

A lot of new joiners in analytics teams across companies face a serious problem especially while describing the outcomes. They fail to understand what their effort will lead to? This effort could be the software code they are working on or the project module that is assigned to them. Why only new joiners? Even employees with 6 to 7 years of experience find it difficult to look at the big picture. The institutes where they learn these analytics’ techniques are partially to be blamed. They are taught to play with the software. It could be coding or working on the dime a dozen graphic user interfaces that are available. They understand how to handle data, get the results and interpret the data. How this interpretation would lead to business gains or efficiency gains is not clear to them!

A simple example could be a segmentation exercise where the collected data is used to segment customers into various groups. These groups could be divided demographically or by using the customers’ choices and preferences. Once this segmentation is done, each segment can be profiled both on the basis of demographics and choices. Up to this point, all analytics greenhorns would do a perfect job. The next step is where complications arise. When they present this to the client, the client enquires about the usage of this exercise. They do not have an answer to this. If they can tell the client how each segment can be uniquely targeted using specific marketing campaigns and what amount of efficiency gains they would achieve, the client would be delighted. If this is done correctly, apart from the short term gain of client appreciation, they can expect long term career growth opportunities.

With so much of data available through various sources like smartphones, internet and social media sites, the requirement for experienced analytics professionals is bound to grow. The beauty of the situation is that this data availability is only going to increase with the advent of internet of things. In internet of things, devices will talk to each other with an app on your smartphone helping you to switch on your television and air conditioning just before you enter your home. A stage will come when the data of your home arrival times can be analyzed and the app will trigger the switching on of your devices   automatically without you even tapping it.

We also keep hearing of big data silos across data stores within the same organization. This happens because people with skills in data analytics do not understand which problem can be solved using the unified data. If they can be exposed to such problems and solutions, a lot of data can be unearthed from data warehouses and used productively.

There is an urgent need for institutions teaching analytics courses to equip their students with the ability to look at the larger business problem and then use their data skills to solve that. Instead of starting with the data, they should start with the business problem and while working on it they should not miss the woods for the trees. This can be done easily when the focus is on the business problem and not on the data.

Why does a rate cut matter?

This article was first published in www.rediff.com on June 23rd, 2015; Co-author: Anisha Sircar, Flame University, Pune 

http://www.rediff.com/business/special/perfin-special-why-does-a-rate-cut-matter/20150623.htm

The Wholesale Price Index for May 2015 stood at minus 2.36 per cent, continuing its downward trend since the last seven months. The Consumer Price Index stood at 5.01 per cent, well within the range of 2-6 per cent targeted by the Reserve Bank of India. With the release of these figures on June 15, 2015, the demand for a further rate cut have resurfaced.

The industry associations like Confederation of Indian Industries and FICCI have already issued statements to the effect that Reserve Bank of India should continue the “rate easing cycle” to “support demand”.

Repo Rate

Earlier this month, on June 2, the Reserve Bank had cut the Repo rate by 0.25 per cent, third time this year. Repo rate (short for ‘Repurchase Agreement Rate’) is the rate at which the central bank lends money to commercial banks.

When banks experience a shortage of funds, they may borrow money from the RBI. When the RBI increases the repo rate, it becomes more expensive for banks to borrow money, creating a ripple effect that affects businesses and individuals.

Higher the interest rates to acquire loans, lower the profits yielded, which may result in spending cuts and a slowdown in the overall growth of companies.

Similarly, an increase in repo rate results in banks increasing their interest rates charged for consumer loans as well, thus reducing the purchasing power of individuals. The diminished ability for consumers to spend discretionary money results in reduced demand for goods and services and affects businesses as well.

Stock Markets

Stock prices are a function of business operations and expectations people have viz. companies at different points in time. If a company is seen cutting back on spending or making less profit, the expectations of people from that company may go down, resulting in a lower demand for the shares of that company. With decreased demand for the shares of a particular company, the share prices start to fall.

When a macroeconomic factor, such changes to the repo rate, affects the entire market, the indices (like Nifty or Sensex) would go up or down, representing the impact on the market as a whole.

Therefore, when the repo rate is cut, the general effect is an increase in the amount of money in circulation, which makes the stock market a more attractive area of investment.

(However, it is important to note that repo rates are not the only determinant of stock prices and market trends. It must also be noted that the stock markets represent only the organised (listed) sector of the economy. And the indices only represent some of the largest companies listed on the stock exchanges).

Market Reaction

Usually, as in January 15 this year, a repo rate cut should create a positive effect in the stock market.

On June 2, the RBI governor cut the repo rate by 25 basis points to 7.25 per cent. Raghuram Rajan announced, "Banks have started passing through some of the past rate cuts into their lending rates, headline inflation has evolved along the projected path, the impact of unseasonal rains has been moderate so far, administered price increases remain muted, and the timing of normalisation of US monetary policy seems to have been pushed back. With low domestic capacity utilisation, still mixed indicators of recovery, and subdued investment and credit growth, there is a case for a cut in the policy rate today."

The markets had been expecting a rate cut and had started to factor its positive impacts in the stock prices even before June 2. See figure 1 below. 

                          

The market had moved up by 1.37% on May 29^th, 2015 and remained flat on June 1^st, in anticipation of a rate cut. However, in spite of the rate cut on June 2^nd, the NIFTY fell by 2.36% and continued its downward trend for the next five trading days.

Why did this happen?

This happened because of the cautious stance of the RBI. The fall in stock prices was because the investors had already expected the third repo rate cut this year and had factored it in, anticipating, in fact, a 50 basis points repo rate cut.

The uncertainty with regards to monsoons which may result in food inflation if not managed properly by the government, rising crude prices amidst considerable volatility and geopolitical risks, and volatility in the external environment, were cited as the reasons for a 25 basis point rate cut rather than a 50 basis point rate cut.

Therefore, overall guidance from RBI was not that of a ‘cheer leader’.

Rate cut again

RBI’s various statements and interactions with the media indicate that further rate cuts in the near future is unlikely. However, with inflation figures expected to be well within RBI’s target, there may be room for further rate cut. Though, RBI would be watching the monsoon and crude oil prices like a hawk before any decision is taken!

How the sub-prime crisis unfolded

This book review was first published in The Hindu on May 03, 2015

http://www.thehindu.com/todays-paper/tp-features/tp-bookreview/how-the-subprime-crisis-unfolded/article7165697.ece

Easy Money — The Greatest Ponzi Scheme Ever and How it is Set to Destroy the Global Financial System: Vivek Kaul

Sage Publications India Pvt. Ltd.

B 1/I-1, Mohan Cooperative Industrial Area,

Mathura Road, New Delhi-110044

Rs 395

The first and second books in the trilogy traced the evolution of money through the era of commodities and even people being treated as money, the rise of gold as money during the World War II, the role of oil, the Gulf War and the dotcom bubble burst.

Kaul continues the journey in this concluding book in the trilogy. The beauty of this trilogy is that it makes boring interesting. The third book is the icing.

After the dotcom bubble burst, one would have thought that central bankers across the world would have learnt from the mistakes of Alan Greenspan, the then [2001] Chairman of the US Federal Reserve. Alas, that was not to be. Greenspan himself continued with the “easy money policy that had created the dot-com bubble...The low interest rate regime created conditions that were ideal for a bubble; the only difference this time around was that real estate replaced stocks as the medium of speculation”. With these lines, Kaul eases into the post dotcom bubble burst era of the building up of the sub-prime crisis.

Greenspan’s warning

The startling revelation is that, just like Greenspan had warned people of the U.S. of “Irrational Exuberance” much before the dotcom bubble burst, he did talk about the housing boom way back in 2002 and that it “cannot continue indefinitely”. But as Kaul points out, he chose to do nothing about it. In fact, the easy money policy continued.

In fact, after the sub-prime crisis, Ben Bernake (Chairman of the US Federal Reserve) continued his predecessor Greenspan’s policy of easy money.

The chapter, “Some are more equal than the others” very interestingly analyses the growing income inequality and ‘borrowing substituting rising incomes’. The book follows on to explain, very rationally, the entire cycle of Chinese exports to the U.S., the consumption of goods by Americans on borrowed money, fuelled by low interest rates, and hence more demand, more exports by China, investing of the earned dollars by China in U.S. financial securities! This cycle kept the interest rates in the U.S. low and at the same time financed the U.S. budget deficit.

While low interest rates were enticing enough to borrow, the financial institutions came up with innovative ideas like 2/28 option ARMs, liar loans, easy lending terms and general thrust to lend to the sub-prime borrowers, resulting in the household debt being 140% of the household income before the sub-prime crisis hit.

How did this happen? Surely someone would have realised the huge risk to the system with such unsustainable debt levels? Why didn’t the banks act earlier? What rocked the party? Why did people start defaulting? In the midst of the crisis, why did the CEO of Citibank get a $95 million severance package when he was quitting? Read the book to know about all these interesting aspects of the crisis. September 15, 2008, the date Lehman Brothers filed for bankruptcy is the date the global markets went haywire and sub-prime crisis became the talk of the global financial system. The foundations of the crisis were laid decades ago, depending on how far back into the history one would want to go. However, how much ever far back one would like to go, there is provision for doing so in this trilogy. Just decide the period and start reading from then on.

Earlier, in Japan

What happened in the U.S. in the 2000s had already happened in Japan in the 1990s. What is needed is that a country not only learns from the mistakes of its own past, but also learns from the mistakes of other nations. This would be very relevant for emerging countries as they would start going through the phases which some of the developed nations have already gone through.

Lastly, it is worth pointing out that the research in the book is up-to-date with references to Thomas Piketty’s, “Capital in the Twenty-First Century”, published in 2014. Events in the global financial system till the first half of 2014 have been comprehensively covered by the book.

In the introduction, Kaul writes that he feels “the best books on the current financial crisis are yet to be written. They will probably start to get published around 2033 (25 years after the current crisis started)”. But till then, for the next 18 years, this might be the best account of the what, why and how of the crisis!

The foundations of the crisis were laid decades ago, depending on how far back into history one would want to go

Health Insurance Hospital Registry

This article was first published in the IIB Bulletin, Vol 1, Issue 4: Co-Author: Varsha, GS1 India
https://iib.gov.in/IIB/Articles/IIB%20Bulletin%20Q4%202014-15.pdf

Poor data impacts many areas in the healthcare system. One of the areas that has an impact on Healthcare Analytics is the way hospitals are identified and stored in the various databases. In the case of the Insurance Industry, each Insurer has their own naming convention for Hospitals. For example, Table 1 shows that five different Insurers can name the same hospital in 5 different ways in their databases.

Table 1

Database A	Database B	Database C	Database D	Database E
ABC Hospital	The ABC Hospital & Emergency Services	ABC Hospitals Pvt. Ltd	ABC Hospital Group	ABC Group of Hospitals

In the above illustration one cannot be certain if all the names are referring to the same entity or if they are all different entities, without painstaking manual intervention. Using the list as it is would not give a clear picture of the number of claims, average claims, top diseases in a period in a particular Hospital, total insurance claims paid per Insurer to the hospital, and many more such statistics.

To overcome this issue it is recommended to identify each entity (hospital) with a standard and unique number. Think of it as a mailing address: an identifier for a single location in the world that is globally unique to that location. No other organization, agency, or affiliate can use it to identify their locations, but all parties can and should use it to identify that location.

The Standard adopted globally to identify a location using a unique and unambiguous number is a GS1 Global Location Numbers (GLNs) based on the GS1 System of Standards. Utilizing a GLN can help improve data integrity. In turn, it will help reduce cost and time spent on data cleaning and making it more reliable.

Such a system enables global and unique identification of products and locations, as well as the continuous, automatic update (i.e., synchronizing) of standardized information across all stakeholders. Unique identification provide the necessary foundation for achieving the best results when using complementary applications like automatic data capture, e-commerce, electronic record management, etc.

Insurance Information Bureau of India has undertaken a project to identify each Hospital in the Health Insurance Providers Network. GS1 India would allocate a GLN to each hospital, which is a unique, 13-digit number for a specific location. Implementing GLNs simplifies the exchange of information and provides the opportunity to manage accurate and authenticated data more effectively.

The GLN, or the globally unique ID would not only identify a specific location, but also provide the link to the information pertaining to it (i.e., a database holding the GLN attributes such as postal address and GPS co-ordinates of the location, services offered at that location, key contact person at that location etc.). This is a key advantage of using a globally unique identifier because all information can be held and maintained centrally in a database or registry reducing the effort required to maintain and communicate information between multiple parties on a national or global basis.

This enables various stakeholders to simply reference a GLN in communications, as opposed to manually entering all of the necessary party/location information. Using a GLN to reference party/location information promotes efficiency, precision and accuracy in communicating and sharing location information.

Figure 1

Several countries like UK, Australia, Austria, North America etc. use GLN’s in their procurement processes to enable efficiency and transparency to deliver better patient care.

The use of GLNs provides a method of identifying locations that are:

·         Unique: with a simple structure, facilitating processing and transmission of data;

·         Multi sectoral: the non-significant characteristic of the GLN allows any location to be identified - regardless of its activity

·         International: location numbers are unique worldwide.

By identifying hospitals with GLNs enables interoperability with other GS1 Healthcare Registries in the world, building global visibility of Indian healthcare facilities, services and capabilities for international patients

However, the most immediate impact of the Unique Identification would be on the quality of Analytics. Only when hospitals are properly identified, logged and data generated on health aspects from them are reliable, can any meaningful analysis be carried out. A list of unique hospitals will be beneficial to Hospitals, Insurers, Govt. Agencies and also the Public.

·         Claims payment can be accelerated

·         Fast, reliable and relevant Analytics

·         Geography based trends, patterns of disease occurrence, cost patterns, etc.

·         Footprint is visible

·         Will aid in the Fraud Analytics efforts of IRDAI

Ministry of Health and Family Welfare is working on standardizing treatment procedures and costing templates. Efforts are being made by IRDAI-FICCI to categorize hospitals. Unique Hospitals would complement all of these projects as well.

A simple illustration may be seen in Table 2 where the outlier analysis throws out more meaningful results when the hospital is correctly identified.

Table 2

	Cost of treatment for Disease type Cholera
	Database A	Database B	Database C	Database D	Database E
	ABC Hospital	The ABC Hospital & Emergency Services	ABC Hospitals Pvt. Ltd	ABC Hospital Group	ABC Group of Hospitals
Claim Paid 1	16,016	2,093	33,115	24,299	39,113
Claim Paid 2	16,577	27,929	22,919	19,366	26,343
Claim Paid 3	12,122	23,767	30,916	29,279	26,000
Claim Paid 4	16,134	25,958	31,108	21,147	15,500
Claim Paid 5	10,280	15,981	1,99,400	26,828	25,000
Average claim paid per hospital	14,226	19,146	63,492	24,184	26,391
Overall Average claim paid	29,488
Highlight Outliers where Claim paid or amount claimed is above/below +/- 50% of the average for the hospital
	Database A	Database B	Database C	Database D	Database E
	ABC Hospital	The ABC Hospital & Emergency Services	ABC Hospitals Pvt. Ltd	ABC Hospital Group	ABC Group of Hospitals
Claim Paid 1	-	Outlier	Outlier	-	-
Claim Paid 2	-	-	Outlier	-	-
Claim Paid 3	-	-	Outlier	-	-
Claim Paid 4	-	-	Outlier	-	-
Claim Paid 5	-	-	Outlier	-	-
If the Hospital is identified as the same hospital in all databases, the average claim paid will be Rs 29,488/- across all 25 claims.
	Database A	Database B	Database C	Database D	Database E
	ABC Hospital	ABC Hospital	ABC Hospital	ABC Hospital	ABC Hospital
Claim Paid 1	-	Outlier	-	-	-
Claim Paid 2	-	-	-	-	-
Claim Paid 3	Outlier	-	-	-	-
Claim Paid 4	-	-	-	-	-
Claim Paid 5	Outlier		Outlier	-	-