Beruflich Dokumente
Kultur Dokumente
MYTH OR REALITY
A dissertation submitted to
CRISIL
By RISHABH GUPTA
MBA 2017-2019 DIME,IIT KANPUR
On 29th October, 2017
Index
Content Page No
Executive summary 2
Introduction 2-3
Pitfalls in 6
implementation
Myths vs realities 7-8
Conclusion 8
References 9
The world has shifted from living in the present to foreseeing the future. It’s an era where huge
amount of data is available, which is analysed, modelled and deployed to increase consumer
base, prevent fraud, invest money, and retain employees. Predictive analytics is the talk of the
era. Almost every industrial sector has developed algorithms to analyse the data available and
predicting what the future holds for them and what steps they can take to grow continually.
Financial sector has employed predictive analysis to model the available structured data and use
it to create models for various purposes which have been discussed. With every opportunity
comes a challenge. Industries employing predictive analysis faces pitfalls and myths. As I have
proceeded further in this document, I discovered how predictive analytics have helped various
industries to grow and solve some of their major setbacks. I have tried to trace out some common
myths about the method of predicting future outcomes which prevails. There on I have discussed
the realities of implementing this method to get the best out of it. At the end of the document I
have tried to put forward my views about it.
INTRODUCTION:-
As per the World quality report 2015-2016, published by Capegemini, Sogeti and Hewlett
Packard enterprise, which found that 66% of respondents in the automotive industry have
business /analytics specialists in their testing organizations, and 47% use predictive analytics to
construct test strategies. Across the health and life science industry, the report says, 67% employ
business /analytics specialists. This is just a drop in the ocean of opportunities provided by
predictive analytics in a world where people and companies are going to any extent of taking an
advantage over their competitors.
Gartner segmented the usage of different stages of data with its value and the difficulty in
obtaining it at Y axis and X axis respectively.
How can we
Make it happen?
What Diagnostic
Happened? analytics
Value
Descriptive
analytics
Difficulty
The process of predictive analytics can be broken down into segments all of which are shown
below.
Various industries are implementing data analytics to improve their processes, increasing the
revenue, fraud detection, decreasing crime, retaining employees.
Financial industry employs an ocean of data and that too structured data so it aims to make the
most out of it. Big data analytics will bring about significant changes in value generation for the
financial services industry. But at the same time a fact which cannot be overlooked is that too
much data and not enough insight or analytics to understand it can cause digital overload. By
employing predictive analysis financial service providers can design ways to offer their
customers more personalised experiences. A vast amount of data is generated by various
channels when they are browsed for services like loans, saving products etc. Predictive analytics
offers a opportunity for these interactions to become increasingly personalised for the customer ,
based on the combination of their past preferences ,real time needs, and pattern of spending and
investing money. Let’s see how it can be done.
The Asset recovery intelligent system prototype is an asset tracking system which uses
semantic technology to identify fraud. ARIS identifies keywords, summarizes and infers new
facts to create a knowledge base of all people and parties involved, all networks and roles.
Let us go through some instances where the models of predictive analysis has been applied:-
Using Big data technology and machine learning algorithms are constructed to take into account
the current state of the market and identify similar transactions which might be of interest to the
investors. In 2014, Goldman Sachs led a $15 million round for big data analytics start up
kensho. Kensho’s predictive analytics capabilities have enabled leading financial institutions to
drastically shorten processing time for analysing the impact of global market scenarios on
various asset classes. This helped firms to cut costs and reduce inefficiencies. In 2015 a firm
Blue Yonder secured the backing of global private equity firm Warburg Pincus to accelerate the
company’s ability to enable its clients to use predictive analysis. The platform provided insights
as well as statistical models for automated decisions. Hedge funds are the first one to employ
artificial intelligence in predictive applications. In the field of investments a designing software
Bridgewater associates has automated the day to day management of the firm including
investment activities. A firm Overbond, a digital primary bond issuance platform, uses
predictive analytics to make better informed decisions within primary bond issuance.
Transparency market research reported that the market for predictive analytics software will
reach $6.5 billion by 2019.
Pitfalls faced by organizations while making data driven decisions.
MYTHS VS REALITIES
So far we have discovered how predictive analytics is used by various organizations to analyse
current and historical facts to make predictions about health care, financial collection activities,
customer retention and many more. Although Predictive Analytics is a powerful optimization
technique, it is often misunderstood, and thus misused due to the misconceptions or myths
prevailing among the users .Below are some of the myths which the users generally believe in
and often land up with wrong solutions.
Myth 1: Predictive analytics produces “perfect” predictions and is always the best
technique.
Reality: The users of predictive analytics often overlook the fact that models are estimations
which sometimes does not take into consideration the intangible or abstract data . Let me
explain this with an example of a decision tree model being implemented for predicting
risk associated with extending credit to individuals. The decision tree classifies risk on
the basis of three variables which are supposedly tangible (income, renter status, and debt
status).
No Yes Yes No
Now suppose a person ‘X’ who is a struggling artist, with low debt ends up with extra alimony
driving his total income above $ 20000. This way ‘X’ ends up in the category of good risk,
which in reality is not the case.
Moreover to apply predictive analytics in the present scenario the conditions needs to be optimal
so instead of straightaway formulating the predictive analysis model, the more simpler
techniques that are less resource intensive like profiling, correlation analysis, trend analysis can
be implemented.
Myth 2: Big data is a silver bullet.
Reality: Though pool of information and prospects can be a real competitive advantage but if
this overload of data is not processed and edited in a valuable manner, it can negatively affect the
analytics to conclude altogether a different solution. Businesses can reach better decisions
through better data management and analytics. Therefore even small amounts of data if correctly
processed can mean a huge leverage in zeroing down to more approximate prospects.
Reality: As we have discussed earlier in this paper, more data improves the predictions but as
time goes on, new data should be added into our model. Sometimes, the model might require to
undergo changes due to cultural shifts, demographic changes, and other events that might create
a need to radically change the model.
Myth 5: Data always speaks the truth so all business decisions should be based on what
data tells us.
Reality: Data does not lie. This is a 100 % correct statement but there may be multiple factors
that affect the quality of data driven insights like poor quality of the source data, environmental
assumptions, and erroneous approach in building predictive models, etc. So making solely data
driven decisions might prove to be wrong and deteoriating for the organization.
Business intuition along with data backed facts and figures proves to be handy and useful for
long term benefits. Therefore the real objective of predictive analysis is not to substitute business
intuition but to help evolve from “gut–feel based” decision making to “evidence based decision”
making.
One such example is the failure of Google Flu trends which was a web service started in 2008 to
predict flu outbreaks in about 25 countries. The logic was to detect some keywords in google
search queries like flu, cold, fever. But at the peak of 2013 flu season, the project GFT failed by
a whopping 140 % .This happened because the keywords used to analyse like cold, cough, fever
etc. were common to other searches also. This led to the demise of the project
Conclusion
Now a result which now seems more to be fact is that predictive analytics in combination of
machine learning enables businesses to anticipate emerging needs and adapt before their
competitors do, instead of scrambling to react after change becomes obvious . Unlike traditional
analytics and reporting tools that provides a perspective on what’s already happened or is now
happening, advanced analytics applies algorithms, and its tools offer a see through view to
predict future outcomes. As a testament to the power of predictive analytics in finance, total
investment in the industry has exponentially grown. But at the same time the challenges faced by
the organizations and myths prevailing in the market about predictive analysis needs to be
addressed carefully so as to make full use of the power of prediction. The sole objective of
introducing predictive analytics was to make smart decisions backed by an appropriate data. But
these misconceptions defeats this sole objective leading to opposite results. We discovered that
how just the data with no business knowledge does not guarantees success. How knowledge of
tangible or abstract values changes the results of the analysis. This technique to grow is not only
for big companies and can make a impact on SMB’s too. A telecommunication provider used
analytics to customize offers for millions of customers and reduced its customer defections to
others by 28%.An algorithm which was originally used to predict earthquakes, was implemented
to predict the probable hotspots for crime. The results were astonishing in Los Angeles where
burglaries reduced by 33% and 21% reduction was observed in violent crimes. Hence
predictive analytics is a technique which holds a capacity to foresee future outcomes and
accordingly firms design their processes to earn profits. But every user needs to face the
challenges of big volumes of data, sorting the quality data and analysing it using correct
statistical models to forecast most appropriate outcomes.
REFERENCES
1. https://techbeacon.com/predictive-analytics-disrupt-software-development
2. https://www.mckinsey.com/business-functions/digital-mckinsey/our-insights/making-
data-analytics-work https://www.computerworld.com.au/article/617969/unleashing-
value-big-data-financial-services-industry/
3. http://www.vitria.com/industries/financial-services/
4. https://ontotext.com/practical-big-data-analytics-financials/
5. Some common myths about predictive analytics-
http://birchmangroup.com/en/debunking-myths-about-predictive-analytics/
6. http://www.bi-bestpractices.com/view-articles/4713
7. http://www.xconomy.com/boston/2012/12/17/the-5-myths-of-big-data-analytics/
8. https://www.thebalance.com/challenges-managers-and-organizations-face-with-data-
4100645
9. https://www.thefinancialrevolutionist.com/opinion/predictive-analytics-in-financial-
services