(Article)Business Analytics, what is Analytics?

​While definitions abound, and each of them captures the essence of the term, let us start  with the simple one:

Analytics is the application of mathematical and statistical techniques to  data, in order  to  discover patterns and co-relations or  to  make   models that  predict,   thereby enabling fact-based decision making  or planning  within the organisation.

Or even simply, Analytics is deriving insight from information, and using it for the benefit of the organisation. But, why has  analytics become so important today?

“The  use   of  analytics  has   pervaded  all  aspects  of business. For  business today,   analytics is  no  longer an   option  but  an   imperative.  However,  as the  ever increasing pace  of  change  drives   businesses to  be more  nimble,  the skill and  sophistication behind  an organisation’s analytical capabilities are  distinguishing the high performers from the rest. With a surge in the use of  predictive   analytics, organisations  are  increasingly buying to anticipate tomorrow rather  than explain yesterday. As  a  result,   analytics-driven solutions  are helping  to transform business across functions.  To stay competitive, companies  are  looking  at  effective  ways to infuse  analytics into every  part of their organisation". (From the article ‘Analytics Everywhere: Using numbers to drive business transformation’, published in Accenture Business Journal for India, 2016,  Issue 2, by Mr. Arnab Chakraborty and Mr. Mahesh Narayan).

As the world around us goes more and more digital and our customers interact  with us through  digital channels, and not through physical interactions      (alternate Channels as compared to branch visits to transact), we may be ‘losing’ the customer, in the sense that the one- to-one  relationship, whereby we  could  understand the customers’ need and wants  and service them across the counter, is becoming less  possible at a physical  level. For  example, nearly  80%  of customer transactions in State Bank of India happen through alternative Channels (ATM, online banking,  mobile banking  etc.) as opposed to Branch  Banking.  This means 4 out of 5 transactions are   through   digital  mode   and   you  do  not  ‘see’  the customer.  To actually  get  a view of the  customer and fulfil his needs, analytics is a must.

This  shows how  the  customer has  gone  digital.  With over  900  million mobile  phones in the  country  people are  now  comfortable transacting digitally. Around  300 million smart  phone users, given  a choice,  are  most  of the time transacting through  a digital mode.  The second aspect is the  huge quantities of data  being  generated. The   data   explosion  is  so   large   that   the   amount  of data  generated over  the  last  few  years is  more  than the  amount generated  throughout our  history.  While technology has  played  a huge  role in this massive data growth, it has  also  helped in capturing and utilising this data. The reduction in data storage and technology costs and growth of networks has  made Analytics easier.

Analytics itself has grown as a science (and art) with more skilled  people and  enhanced tools  and  models being available. The  use  of analytics has  to be  all pervasive across products, processes, Human  Resources, fraud, and  risk to name a few. It has  to be across all channels so that not only is all data captured and  updated, but is available to the employees, and, more importantly to the customer at his preferred channel.

As we live in a world facing  continuous disruption  and competition, it is necessary to understand the customer and  meet  his  needs when,  where  and  in the  manner he  wants  it. If today,  you want  a pizza,  you just call. If you want a book or music,  you order  online. If I want to see a movie,  even  my movie tickets  and  snacks are  all available for purchase online. However,  many industries and  institutions   are  still playing  catch-up. Banks, for example; If you need a loan you generally have  to end up in a branch (and mostly more than  once). 

As  Bill Gates has   famously   said,   you  need  Banking but  you  don’t  need Banks!  Disruption  is  everywhere, as shown  by Pay  Pal,  M Pesa, Airtel money  etc.  The way of doing  business is changing to mostly  serve the customers. Here are few examples:

•      Air BNB –World’s biggest hotel chain  provider does not own a single hotel.

•      Uber – World’s largest taxi operator does not own a single taxi.

•      Facebook  –  World’s  most   popular   media   owner does not own any content.

•      Alibaba.com – World’s most  valuable retailer  does not own any inventory.

To help understand Analytics in a simple but comprehensive way,  we can  do no better  than  to look at the success factors  that make  Analytics work. These have  been shown  in, ‘Analytics at Work’ a book by Mr. Thomas H. Davenport, Mr. Jeanne G.  Harris  and  Mr. Robbert Morison.  They  have   grouped the  factors   as DELTA (the  Greek  letter  Δ or δ). These factors,  using analytics,  can   change  a  business.  The   full form  of DELTA is:

D   for Data  (of good  quality  and  retrievable for use), Breadth, Integration, Quality.

E   for an Enterprise wide usage (approach to managing analytics).

L    for Leadership in Analytics (passion and Commitment).

T    for Targets (First Deep  then  Broad).

A   for Analytics (Professionals and Amateurs).

First and foremost is data. The data  comes from a huge variety  of  sources and  has   to  be  stored in  a  proper manner for analytical  usage. Data is the new oil, as they say  nowadays. It is the base for all that follows. Let us take  State Bank as an example. The Bank gets  internal structured data  from 58 different sources including customer data  (Demography, Geography, income, gender etc.),  data  from the  contact centre, Complaint Management and Lead Management Data, Transaction data   (70  million,  transactions  a  day),   Product Data, Distributer  Data,   Channel Data,   NEFT,  RTGS,  ECS, employee data, Forex Data.... the list is growing. This   data   can   be   Captured  –  to  undergo  an   ETL (Extraction,  Transform and  Load  Process). The  Bank has   nearly   300  TB  of  this  data, with  huge   amounts moving in every  day.

How is this data used for analytics?

Few  examples are  below.  The  most  easy  usage,  to get  quick wins,  is to use  it for descriptive or statistical analytics. For example, how many SME loans  are being charged at  below  the  minimum  rate  (believe  me,  this can actually happen and if corrected, can result in huge gains  immediately).

Which Home  Loans  have  not registered the  mortgage in the  system, how  many  customers (specially  senior citizens),  are  without nominations, submission of subvention claims to the Government for example, Agricultural gold loans, staff accounts without the  staff identifier   (Provident  Fund  numbers  in  State  Bank’s case) etc.

Further, analytics would help in bringing in new accounts using data  mining on the narration field of the customer, to track how many  housing loan repayments to X Bank are  going  through  customers’ accounts  maintained by A Bank  through  RTGS,  NEFT,  ECS  transactions etc., car  loan repayments to Y Bank, or Insurance Payment to  Z  Insurance  Company. These  customers  can   be contacted with a  view to see if  their  accounts can  be held by the Bank too.

A very  important  usage is to create a  Customer One View (COV). Here,  the  static  details  of a customer are captured like name, address, mobile number, PAN number, Aadhaar Number  etc. The customer’s account holding with the Bank  are  also  captured – like types  of deposits, (FD, SB, CA) or loans (Home, Car, Educational, Personal etc.),  his  transaction preferences  like using ATM, online  banking  etc.)  and  also  his  insurance (life and  general) demat holdings  etc.  A value  is derived  at, using  all this data, and  the  customer is rated  (as  say, Platinum, Gold, Silver, Bronze etc.) and a tool run to see the  next  best  course of action  vis-a-vis  the  customer. The   tool  throws   up   the   products  which   should   be offered,  considering the customer’s income,  age,  place of stay,  existing  portfolio etc.  The various  methods like e-mails,   branch interface, call  centre, online,  or  ATM etc. can be used to convey  the product  to the customer. With agile  technology available, it could  even  go  as a pre-approved loan.  SBI, for example, has  a tie up with Flipkart  to  give  EMI, to  over  a  million SBI  customers when they purchase from Flipkart, in real time.

The data  can  be used for process analytics (now much cash to be kept at over 50,000 ATMs of the Bank). Each ATM is monitored analytically  on  usage patterns (like first days  of the  month,  holidays,  festivals  etc.)  and  an optimum cash loading figure is arrived at. This results in huge  savings.

Fraud  Analytics (velocity checks, i.e. a card  being  used too frequently for example, checks on a card being used simultaneously or closely  at Chennai & Mumbai,  fraud patterns) is an important  Analytical capability.  Analytics can  also  predict  loans  at risk, accounts likely to attrite etc. through   models which basically  run through  similar accounts which have  defaulted or attrited  over the past two years, and  use  the  findings  to run  the  symptoms across existing  accounts, which throw up similar trends in some of the accounts (Action can be initiated to check attrition/defaults).

Risk  analytics is  a  vast  area. Risk  models are  highly developed and  can  predict  probability  of default,  loss given  default  etc.  Quick,  real  time availability  of data enables the  early  warning  system to  work  efficiently by taking  up with the  borrowers in time.  Analytics  has become big in regulating and  compliance areas  also. The  need to  submit  accurate and  timely  data   to  the regulators is a must.  Late  or wrong  submissions have repercussions. The ADF (Automated Data  Flow) is now being  mandated by the Controllers.

Data   in  hand   is  however,  very  dynamic.   More  data needs to come  in and  it has  to be  cleansed  regularly. Even  for the  structured internal  data, gaps exist  in all organisations.  When   data   entry   happens  at  various places by  a  large  number of people, it can  result  in anomalous entries, due  to lack of time,  understanding etc.  At the  entry  level,  further,  new  data  is needed to be  updated constantly. In a  Bank  which  has  been in existence for a large  length  of time (two centuries plus, in SBI’s case) legacy  issues exist in data. Moving from manual banking  to  bank-master (Branch  stand alone computer system) to core banking can result in data loss. Also a couple  of decades earlier,  cell phones or PAN or Aadhaar did not exist,  so older accounts may not have the  data, which  has  to be  obtained and  incorporated. Even  date  of birth can  be an issue in rural India.

To this internal data, we have  to add external structured data.  This  data   from  Credit  Bureau  and   if  possible from  electricity,  cell  phone companies etc.  is  gaining in  criticality.  For   example,  while  without   the   credit bureau  data,  a  bank   may   be   pushing  for  a  Home Loan  to a  customer, the  Credit  Bureau data  may,  on availability and  incorporation, indicate  a customer who has  defaulted with another bank(s). The  electricity  bill amounts give guidance for prospective borrowers. If he is paying a bill of `200 a month, a high end car loan may not be  warranted. So,  also  giving a small car loan to a borrower paying bills of `20,000 a month be relooked at, considering other  factors,  of course.

For analytics to succeed in any  organisation, it has  to be  accepted and  used across the  verticals.   Analytics has   to  select  applications with  relevance to  multiple business areas, and the analytics team  has  to work with Business and  IT jointly. The  team  has  to develop the analytics strategy and  a road map  for all business units and attempt to extend analytical  tools and infrastructure broadly and deeply  across the enterprise.

The leadership of the institution has  to have a buy-in, for analytics to do well and  produce results. Just  to have  a full fledged  data  warehouse and analytics team  (which has   costs), needs foresight   and  guidance of  the  top management, to undertake analytics and ensure usage. Business will only go where  results can  be  seen and where these can be traced and monitored. “Low hanging fruits” are available to start with, but the analytical  group has also  to focus  on strategic initiatives, value  creation and  building  distinctive  capabilities that  will enhance competitive differentiations.

We  also  have  un-structured data  to  take  into consideration. There  is again  internal and external unstructured  data.  The   internal   un-structured data comes in the form of emails,  comments from appraisal reports, auditors’ remarks etc.

The  external un-structured data   available is  limitless and a big source is the internet.  It includes social media posts  on   facebook,   linkedin,   pinterest,   instagram, twitter,  Youtube   and   others.  Also,  increasing  use   is being  made of crawling engines and  listing tools, which go across the  web  (with defined  geography, say  India web)  and  generates output  on special mention  words. For  example, if  you  were  to  use  a  listening  tool  and put  Home   Loans   and   Bank  as a  mention,   it  would generate reports on all the  times  the  words  were  used in conjunction over  a  set  period  of time.  But this  data being  in  un-structured format,  cannot be  stored in  a traditional data  warehouse. So, now a ‘data lake’ is used to capture such  big data. Big Data  is a combination of some ‘V’ characteristics.

1.   Volume: Big data  is high volume. The volume which can  be considered big will of course depend on the firm’s size and the industry it operates in.

2.   Variety: There  are  multiple data  inflows. As stated earlier,  these would be structured, partly structured or unstructured data. The data  could be from within or external.

3.   Velocity: The speed of generation and movement of data  is growing. A good system will try to capture the data  as close  to its time of generation as possible, as this  allows  business units  to make  real  time  or near  real time use  of the data.

4.   Veracity:  The   purity  and   reliability  of  the   data. In spite  of best  efforts  some data, by their  nature will remain   unpredictable, like the  weather or  the economy data  and this has  to be taken  care  of. To  quote   from  an  IBM study,   published by  the  IBM Institute   of  Business  value   and   the   said   Business School   at   University   of  Oxford,   “companies  clearly see big  data  as providing  the  ability to  better  predict customer behaviours and by doing so, improve the customer  experience. Transactions, multi-channel interactions,  social   media,   syndicated  data   through sources like loyalty cards, and  other  customer-related information  have  increased the  ability of organisations to create a complete picture  of customers’ preferences and demands – a goal of marketing, sales and customer needs for decades”. Another  area of data  generation and  analytics is going  to be  from deep learning  which an  article  in  ‘The  Economist’,   defines as an  artificial intelligence technique  in  which  a  software system  is trained  using  millions of examples, usually  called  from the internet  which bring us to the ‘A’ of DELTA.

Today analysts are a set of skilled people in great demand. They  are  amongst the  highest paid  professionals and all ranked universities are  providing/planning to provide courses on  analytics. These  range  from  Ivy League Universities to our  IIMs and  are  being  undertaken  by more  and  more  management  institutes. Grounding in Mathematics, Economics and Statistics, along with technology knowledge makes for a  good  analyst. The

IIM, Kolkata  has  a  course on  Analytics  in conjunction with IIT, Kharagpur and  Indian Statistical Institute, Kolkata. The course is for two years and spreads across the  three   institutes (6  months  approximately at  each place)  with an industry project  work also.

As  the  quality  of  analytics and  the  analysis they  do will be  the  major  factor,  while  selecting the  team   of business  analysts,  technical analysts,  data   scientists etc, personnel qualifications like MCA, B.Tech, statistics degree holders, B.E.s would prove more apt, especially if   they   have   undergone an   analytical   course,  and also  have  experience in analytics work (although it is difficult to get many such  people in a young industry like analytics).

An analyst is much  more  than  the degree he holds.  He needs to have  good  communication skills,  as well as have  an understanding of the Business Domain  and  be able  to keep  commercial interests of the organisation in mind.  They  should  also  have  the  ability to sift through data  and discern the patterns. Of course, model building is a  core  asset. Not only does an  analyst need to do analytics,  he  has   a  large   role  to  see that  business understands the  power  of analytics. He  has  to  guide business to use  the analytical  findings and this requires time,  patience and  effort.  Domain knowledge is a  big plus, as are project  management skills.

As the world around us changes much more rapidly than any  time  before  in history,  as disruption  becomes the norm, and as newer  technologies like internet  of things, artificial intelligence, robotics,   block  chain  technology and  a  host  of other  developments are  on,  the  Analyst has  to play the key role in taking institutions  to the brave new world. As Mr. Peter Sondergaard has  said  “Information is the oil of the  21st    century,  and  analytics is the combustion engine”. Let us make  the engine run well.

Courtesy: Kajal Ghose (Former Chief General Manager, State Bank of India)