1.   Application  of  computing  in  business  has  grown in expanse and depth very rapidly and added lots of new dimensions in operations, service delivery and business management. Analytics is one such specialised area that can be used to enhance customer experience as well as business manager’s capabilities for directing business efforts with focus. When you call a marketing site and get answered in your preferred language with greetings by name, reckon that analytics have played in the background. You look up a marketing site to view some preferred items , say a TV for example; next time you go to your mailbox to check for emails, advertisements appear here and there on the screen , about those preferred TVs and brands- click it and you are in the marketing site  with the specific  item on the screen. Surely analytics is playing a part in all these.

2.   Data volume gets very big in today’s human activities at business, say like banking, because business delivery and customer operations are on computer, in almost all activities – which means much more information on customer activities are getting captured in computerised environments, leaving trails and records in terms of computerised records. To  be  able  to  understand  and  direct  business efforts internally, or, provide rich and meaningful components and contents in customer interactions, all such data related to customer transactions and behaviour, as also external environmental / market data, are required to be captured, understood, studied and analysed. From all such data, selected portions are extracted, suitably restructured for ease of quick retrieval to help in queries made on this big pile of data, and combined into new data elements forms a new and different database. This process is ETL – Extract Transform Load. These information of different items or activities etc., are  then put in a different database ; these databases form the base of the activities like   storing of   information, retrieval and processing to discover trends, rules, patterns of   customer behaviour and some more related information. The various types of activities and handling in this area are termed as Data mining, Data Mart, Master Data Management, etc.  Under the overall domain of Data Warehouse.  Structured and organised data are made of data elements that follow strict rules of length, content type, permissible range of values, etc. However, there will be much other information which may be in texts, pictures, sounds etc. that may be useful to study and rate. For example, apart from price, a design or colour shade of merchandise, may have  reasonable influence  on customer preference without customer being able to articulate the same. However, by studying customer preferences, the back end system of a seller can have an insight to use for production and marketing strategies. This class of data  will not follow a fixed structure, syntax, size or even the form – these constitute what is called ‘unstructured data’.

After we have these data, i.e., original business data and then reconstructed as mentioned above, we need to analyse them  to find meaningful insight. This part of activity is Analytics. There are other activities  after  this  to  use  the  findings  for internal understanding, relate  such  findings to the business facts observed (proposing models and testing them for validation), rating of various factors so found and create and test business strategy, customer interaction, marketing strategy etc. Incidentally, always a customer or selling a merchandise, are not the target of the exercise. We bankers may do  an  Analytics  of  our  MSME  loans  portfolio, study the repayment histories, the appraisal and sanction procedure, post sanction acts of bank, market  conditions,  overall  external  indebtedness of borrower, family earning and loan histories etc.,

to  perhaps  arrive  at  a  desired  accurate  formula for making provisions for bad loans. Because the data is diverse and huge - thumb rules or simple averages or projections based on one or two easily measurable factors, will not do. And, if we desire to know while doing a Money Market deal, probability of this deal to cause the bank to exceed any agreed risk exposure level or limit at the whole bank level, then the Analytics and the resulting action have all to be real time, within the activity session of the deal. The narrations above are of course a simplified and narrow bird’s eye view only, the gamut of activities and challenges to understanding data in reality, are much  bigger and difficult.

3.   Analytics as a part of the integrated data driven operations of an organisation, will usually consist of classifying, segmenting, grouping of data, computing some values (of result, trend, etc.) representation of the same on screen by tables, charts and graphs of different types, dashboards, scoring tables, or similar any other graphic ( for on- screen) presentation  providing interactive program for business management to study, change a few parameters and see the effect on the result, etc. For example, we can see the impact of   a change in, let us assume, transaction  charges to be levied for services  - - by trying with various different values of transaction charges, and note the expected changes for the same on the profit or market  share; this can be as a graph or bar chart or any other desired format of output on screen that appear about immediately (after entering the varying inputs). This class of activities are often called ‘what-if” exercise. The practice of visually seeing a change in output as an impact of complex business factor interplay, - is called Visual Analytics. Sometimes depending on the domain or platform to be studied or evaluated , a genre of analytics is named – e.g.– Cloud Analytics, Banking Analytics, Risk Analytics, Loan Analytics, in commercial communication. Basically selection of parameters to study, the data elements to be chosen, the  features  of the output, the  business  domain specific  data  elements, and  the format of the result to be shown – may often have some specialities or usage norms; they use business rules and concepts of that domain, and can get bundled and sold under such  specific  names like in the  foregoing  example; there are no hard and fast rules however.

The major target of analytics is to understand the dynamics   of   market   factors,   operational   entities, etc., and then be able to predict market response  or customer impact,  and  then  finally to  provide  suitable links, handles, offers in customer interactions in terms of presentation of the interactive  internet  screen  for customer; the customer can be internal – (like in our example  of  the  loan  portfolio  understanding  above)

– who are   expected to invoke a favourable action – (like  customer  gets  enthused  to  purchase  an  item). With such results for a larger number of sample of customers, the internal team may get helped to select or propose the underlying algorithm and build up a model for implementation and testing The  types  of analytics that  are  specifically  tailored  to predict results or outcomes to help business plans and strategies, have grown into a distinct genre and are referred as Predictive Analytics. As mentioned above, there are many other typenames based on purpose or         business domain (Financial Analytics, Big Data Analytics, Customer Analytics for banks, Risk Analytics for banks – which are  termed and  marketed as specific products by vendors to service providers / banks / business) or on similar segmentations. The major driving forces behind banks going in for analytics may be a few – most notables are:

a.     Regulatory Reforms, asking for more and more data based information from banks.

b.       Profitability/ Cost cutting  in view of  increasing competition.

c.       Achieving Efficiency in operations.

d.    For better Risk management.

e.     To obtain   better insight into business data and customer preferences – these can be customer segment-wise also, providing a farther segmentation.

f.     Attempt to redesign business processes.

g.      Fraud Control.

h.     Loan delinquency avoidance.

i.      Customer satisfaction assessment and enhancement.

j.       Call centre or workforce efficiency.

k.     Cross selling, customer acquisition, etc.

Hardly, an all pervasive project to kick-start many studies and activities in many domains will get done simultaneously, because business dependence is complex based on multiple factors. Any model or strategy, should   better   be   piloted and tested in parts, by adjusting different parameters one by one, and the overall business system allowed to grow with these, in steps of changes to help stabilisation and correct understanding of effects of a change in each of the many factors in a business situation.

4.   Banks handle huge data, and need to do more, which  they may not  be normally doing – say for example while we study loan defaults on the basis of accounts or customer numbers; however, study of relationship of loan delinquency with customer’s family/lifecycle issue history or projected competition of alternative service providers that may affect banking usage of customers, etc, is not easy, as, dependable data itself may not be there, or their relationship to business  results are not understood well. Over and above, the thinking and capabilities required for data  crunching and  finding patterns in huge volume of data, are not in the core competence areas of bankers. On the other hand, technocrats are not expected to have the business domain knowledge. In this backdrop, it may be appropriate to see how best a simple banker can get along with Analytics in the best interest of the organisation.

5.   The various available products of Analytics in the market as they are, suggest that the developers behind them have gained a reasonable insight in the underlying business. The teams of technology experts   and       business  process experts  from the providers’ sides have   developed  these products. The most distinguished and established organisations like Gartner or Forrester rate the capabilities of vendors that get accepted more than for any other ratings in the industry. These ratings tell us company-wise capabilities based on various factors that they explain in these   rating   releases. However,  if we bankers plan  to consider a specific genre of product, it will be good to look into the views on the particular product and domain and check that the functionalities and deliverables are in line with what is our plan and our own domain. We may not at the outset,  be able to spell out or fully plan the outputs or the resulting product to procure,   like we can normally do when we procure a server or few discs or some equipments, or some fixed  functionality   products  like  MS  Office,   etc. There will be some exploratory components in the solution and the outputs. There can be a facility in the solution provided - for user operated (by the banker who is implementing this solution) day to day analysis, report, parameter changes, etc., on a regular basis, or as and when required.  It may be useful to adopt a few core outcomes, like capacity addition  for  understanding/analysing  /  reporting etc for management support, as the desirables; specifics of the solution  can  get defined  and  refined as we go from here. The banker’s team involved in the  initiation of the  specific  analytic  must  have members knowledgeable in the business process of the underlying specific business area operations, for which an analytic solution is planned to be deployed. In most of the situations, the analytics vendors (that include big names like SAS, IBM ,etc worldwide, as also quite a few niche solution providers are there in the top bracket) have, through the assignments handled, collected knowledge and practices of the business domain and embodied the same in their solutions. So, as such many vendors would be approaching banks with specific  solutions – say  on credit risk management, or fraud risk management, etc. These are to some extent ready, that a bank can procure, learn operations, put values of parameters,  get  trained  in,  and  start.  This  may prove to be a very easy and comfortable option for a functionary in the bank side, because, depth of their expertise or conceptual clarities may not be very great always, due to frequent movements in banks or limited or no experience, or also, the areas are new and growing, or, scope of theoretical grounding and exposure to global knowledge and practices are  limited.  Whatever  it  maybe,  these  together may lead to a situation of vendor dependence for operating expertise and also, thought leadership, This may not be helpful for knowledge enrichment and capacity creation in the bank. It is a good idea to expose the bank team for an analytics task / project, to theoretical concepts and industry best practices- preferably in the domain desired to be controlled or predicted with Analytics. For example – if we need to  take  up  an  analytics exercise  to  find  what  all to do to improve capital adequacy and block and mitigate factors that erode capital adequacy, even if a vendor arrives with a ready model and solution to fit into – it will be  useful  to field a team  from the bank side consisting of   business domain people thoroughly knowledgeable in the concepts of capital adequacy,  Basel committee norms and directives, RBI directives, models in industry  use  in this field, etc., testing and validation concepts like stress testing and other global practices, and also internal working in the bank to the extent  that covers how from all the business departments practically which business figures   and   data   emanate and   get  fed into required capital adequacy computations, and, to what extent. The team also should have one or two Information Technology person(s) who are thoroughly conversant about which data elements pertinent to this domain are sourced from which accounts or operations in IT, if there are processing issues in IT that may have scopes to have bearing on the data values (say some values  are repeated from old data if new data is not updated and some others are left blank if new data is not received – the dependability of the data quality  gets differently affected in these two cases), and similar inside views. The IT persons are also to act as bridges with IT for interfacing or aligning any analytics input or output from or to the main banking system (core banking) or its subsidiary systems. Apart from proper manning and business knowledge gathering on the issue to be subjected to an analytics exercise, the usual project management that the banks do, often in their own practiced ways – will have to be in place as usual  for the analytics project also.  However, Analytics  being a bit advanced in concepts and far more advanced in IT-  in terms of processing capabilities and methods than the usual applications that get added besides corebanking, the processing of the analytics activities are to be in the analytics technical domain mostly. However, we need to have some insight and some understanding in gross terms, about the working models and components of analytics.

In most cases, analytics should lead to Predictive Analytics that should predict outcome (example - in which case the chance of a borrower failing to repay will become high), and suggest actions and produce the appropriate actionable (say a special notice to borrower, or, a special inspection schedule for the loan officer can  be  produced by the  system, or the account can be included in providing for doubtful accounts to a decided extent) , and very desirably- the system  should automate the process to a good extent – leaving it for human approval or revision

if desired. The basic purpose is to use superior technical capabilities with control and focus on business goals – not getting overwhelmed or led by technology. Also, providing clean data, appropriate data, data  that  can  be  verified to be  correct  – are very important, as otherwise analysis, modelling, and  predictions  based  on  such  information  will not be useful to business. For the analytics to be useful, the banker is primarily responsible to have clean and correct data in the system. This sounds obvious, but is hardly come across. Incorrect, incomplete, and inconsistent data has been there to an unacceptably high proportion in many banks. Rapid expansions, conversion from manual to branch-based computerisation and thereafter to core banking could not take care of these gaps fully because the older systems and the later systems did not have the same data elements, and often the older data elements were not captured at one place so that many gaps resulted while converting to later versions. Banks have through special drives of data cleansing and de-duplication covered some ground. However, for a particular group data to be adopted for Analytics to provide us with insight and suggest actions  –the   first  requirement  will be   a   special check and cleaning of the data, as also, conscious decisions as to what default rules will apply in case of inconsistency, the actions to be taken for them, and the impact of these imperfections on the results should be understood, and used while appreciating the Analytics outcome.

6.   Before coming back to the issue above, it may be useful   to again understand the gamut of   use of data  for  business  understanding  and  directions and the entire universe of data warehouse, data mart, Business Intelligence, analytics, visualisation, modelling, predictions, etc., to understand the place of analytics in these and its role.

a.       Gartner          defines         Business          Intelligence (BI) to be a wider activity that “spans the people, processes and applications/tools to organize information, enable access to it and analyze it to improve decisions and manage performance” In  this  context  Analytics is  defined   as  “packaged BI  capabilities  for a particular domain or business problem” [Gartner  IT Glossary]. Other  definitions  put Analytics as a science of analysis, or, tracing of

things to their source, or mapping information to its original causes or principles. In other words, analytics is a way to understand causes of and connections among business events, business conditions, outcomes. So, analytics is expected to enable business managers to appreciate causal relations that are not easily visible, lead to insights some of which may be found  to  be  crucial  or  significant.  This  leads to right business decisions that  are  difficult to derive in normal course in view of the usual deluge of multidimensional information faced in business from diverse sources.

b.     In  this  data  driven  discovery  of  possible business truth, often  a full fledged  creation of Data Warehouse, with few Data Marts, is in place. Often  only a  specific  area is taken  up for analytics or in wider terms – BI. The start in either case will be with data collection, capture, appreciation, transformation and then creation of the database for this warehouse or BI system. In this data capture stage itself, there can be rules and approximations based on experience based stored rules, or established practices and algorithms. Particularly if data is not standardised, say in handwriting/picture/ graphics/sounds,  or  languages  with  usages not complying with the computer stored ones, or if data elements do not show compliance with  established  social  norms  –  in  terms of demographics, financial standards and occupation/assets, etc. Data from there now will be  subjected to  classification, segmentation, and study of the same has to be done. The rules  to apply  in these  classifications, finding relations,  investigations   will  have  to  come more from the bankers and less from the technocrats  –  as  bankers  are  expected  to know which item is meaningful to link with which business phenomenon -           say deposit growth may be dependent on customer’s age, earning, liabilities, quality of            service of the bank unit interacting with the customer, market dynamics related to the customer’s profession, customer’s awareness and banking operational capabilities, local factors, seasonality etc. as also,  priority among such factors. The outputs of these processes are to be presented for management view and considerations. This is in fact the specific  part that in the whole chain of this data based decision support attempt, is the analytics stage per se. Analytics   will consist of activities like creation, presentation and analysis of tables, trend graphs or graphics,  executive dashboards, reports, scoring, Balanced Score Cards, priority lists, etc., the way desired. It should also offer tools to manipulate the data elements or some combinations or derived results to see the overall effects of such proposed changes. These  activities  lead  to  understanding  that will be modelled, tested, validated or rejected or   amended   and   retested.   The   outcome will constitute business understanding and knowledge. From this, actions will ensue as the business managers may decide. These actions  can  be  very  diverse  like  changes in pricing plan, changes in customer facing communication or repeated push of the promotion messages for preferred products around customer activities – not necessarily in the store or website of this business entity. For example, if I was looking up some sofa sets in Amazon or Flipkart or similar other online merchants, (the  names are  purely to cite an example and does not suggest any preference or recommendation) on my laptop today, it is most likely that for quite a long period I may find  sofa-set  pictures and   prices   from  such sites on my screen interspersed with other subjects being looked up – that may be a blog on sanitation system in my city or seeing my electricity bill in my mailbox for which the ISP is neither of these merchants. The tendency is to map customer behaviour as close to real time as possible and push the content that have a chance to be settled for; the icons or messages pushed in the web for this, are also provided with means to carry through the preference into desired actions i.e. sale for the item, or joining into the activity being promoted, by invoking a link to the exact page and item of the merchant website or the organisation promoting an item based on the analytics in the background. This gives an idea that there must be collection of customer activities, quantifying them under some types or scores, and automated rule creation in the background, application of such rules to the customer activity information and then  an  automated  rule  based  adjustment of customer deliverables (like on screen repeated messages or promo items). These are the outcomes from the analytics, and the rule creation etc are, often programmatically done – new inputs help add to the rule stack. These types of activities are by the software in place; these software have logics to add to knowledge base and rule base – often they are called self learning, or, heuristic etc. A genre of programmes and algorithms employed in this respect collectively      go into the domain of ‘machine learning’, ‘artificial intelligence’ and similar types of very specialised software. 

c.     While it is interesting and mind boggling, we bankers will really have to use analytics for our  business  with  control,  comfort  and  be able to direct it to our goals – irrespective of the  marvels  of  the  software  or  technology, and our lack of meaningful knowledge of its internal working. The market often packages ETL tools, BI, analytic applications   together to form a base product that is marketed and added on or enhanced; the whole package may get named as an analytic package, which is not our concern. Mostly the packages in the market come with some pre-built analytic packages – built upon previous experiences and interactions with other users / customers. This is good in one way that, this brings lots of experience and use-cases ready to make a start, as we may not have a handle on exactly what will be our desired components of a solution, or how exactly in an analytics environment our needs gets translated into algorithms, products and logic. However blind reliance on vendors and available products, attempting to fit into available constructs, and partially managing external add-on processes to lead to the desired outcome, may not be a good idea in the long run, as, the automation, the speed and ability to interact in-session or on-line will get lost, and, interfaces may pose to be source of error, delay and lack of control.

d.     From all of the above, the learning that we bankers need to take is that, if a number of factors are more probable to lead to a certain outcome that is not understood normally, can be understood if our basic business concepts and targets are known and prevails through the analytics exercise. The products and underlying  technology  are  complicated  but we need to be able to articulate our needs and understanding, and accept propositions based on experience and tested results. A general saying these days are that University teachers are required to teach and guide in topics or ideas they did not study as a student, or even did not teach ten years ago. The same open mind and continuous learning is highly recommended for bankers to be able to handle and guide in these new areas, instead of being blindly waylaid by vendors. It is an observation that despite use of frontiers of technologies, often a BI/ Analytics project does not deliver much   good,   or,  benefits are  not  reasonably derived for the organisation. This mostly will be due to less than required understanding of task, inadequate requirement articulation, inadequate information on existing system, lack of infrastructure or driving of the project from the bank side, and gap in diligence from vendor side. Communication, monitoring and oversight from bank help reduce these gaps. For bankers, attempts to understand the scope and dynamics of BI/Analytics, and willingness to learn, expand views and accept the philosophy of iterative trial, learning, and improvement are expected to help use BI better.

7.   The approach to adopting analytics will have to be like for any other project. A simple list can look like below -

a.     Fix which business activity to cover, expected outcome, and who champions and who steers the project.

b.    Internal   discussions   to   assess   what   are the areas to enquire about, have deeper understanding, and the possible demands on time, cost, staff skill and knowledge, expected benefit.   A business and   IT mixed  group  as mentioned earlier need to dwell upon and act.

c.     Decide, and make staff and seniors aware of the oncoming project requirements, impacts, and expectations from employees.

d.    Look  for  solutions,  read  up  market  reports, invite vendor presentations.

e.     Once   the   area,   nature   of   attempt   and outcome  are  more  or  less  clear  internally, start usual project routines of procurement, team   formation,   training   of   core   project people, development of project deliverables, requirements in cost / manpower / time etc, commence project with continuous dialogues with vendor’s  team, regular  review, verification against milestones etc.

8.   The areas that are supposed to pose  big demands are a few :

a.     Hardware, IT set-up. The volumes that need be crunched are very heavy compared to what usual business does, as lot of external and market information, and unstructured data may be required to be crunched to understand and validate even small business trends or events. The hardware allocated to this activity will need big capacity and high processing power; the solution vendor will indicate the requirements. Even  big memories and  specialised data  flow processes are adopted like-

1.     Processing - use ‘in-memory processing’ – for big volume of data, the entire data is taken in the RAM together and processing happens for the whole avoiding ‘fetching’ from disk during the processing in small lots to process and return, to cut down time required;

2.     Even the programming languages used – say ‘R’ - are different from what banking software use.

3.     Use  of  special  genre  of  software  or processing methods – like ‘machine learning’,

4.     And    many    others    like    specialised processes to handle Big Data.

b.     Technological skill – the Analytics software or process handling skills are more specialised than many other existing banking software items, and solution vendor need to post a capable team for that. Getting bank employees to learn and take over can be started only partially and in a slow phased manner, if desired.

c.     Time and cost – the products take time to understand and work with, as also are costly.

As to time, it is better to have a long term view and  take  up reasonably identifiable  parts  of  it one by one over time,  instead of a big bang approach to cover all activities or areas under a big Data Warehouse (reservoir of all data extracted  and    converted    and  organised), data centric solutions,/processes , a few Data Marts  (a  smaller  section  of  data  reservoir for   a   particular   business   domain/vertical, with capabilities of data extract,   reports and dashboards), Analytics in few areas, etc., all together leading to a decision system (also involves – modelling, testing, improvement, stabilisation, resulting in amended business processes).

d.    Space.

e.     Manpower.

f.     Control – of resources, target shifting, rigours in scenario creation and testing.

g.    Testing  and  validation  –  acceptance  and implementation of discovered business rules should be after thorough testing and validation. Despite technical capabilities, any software will produce for which it is programmed, and also ability to cover all possible effects of our action on market or human (customer) behaviour.

9.   Analytics adds huge power and speed to information processing and actions; this is very helpful in the present environment of data intensive activities in all walks of life, with ever increasing data volume. Bankers need to have a primary idea of what all constitutes Analytics and what all it can do, and keep focus on their own objectives in use of analytics to derive  benefit.  Some banks are  already employing Analysts, Data Scientists, etc to handle their data warehouse from the bank side or undertaking some analytics projects. This is welcome and is expected to make analytics more widespread and understood. Long back few banks started employing in-house IT personnel, from where the banks’ IT capabilities have   moved   very  far  now.  In  this  sunshine field of data based decisions, the same is perhaps the future.