Success in today’s data-oriented business environment requires being able to think about how API, Data & Analytic concepts apply to particular business problems. Data and data science can provide value in the context of Bank’s business & competitor strategy and to meet demands of customer experience.
Google Finance Australia reported that majority of Internet users in Australia start their search looking for a mortgage online, and spend between six and 11 hours researching the mortgage before they select a potential provider and reach out to them. When they do contact a mortgage provider, increasingly it will be via the website, rather than by walking into a branch or phoning the call center. The myth that customers require a branch to buy a mortgage is just that, a myth. It is more than likely that the majority of mortgages sold today were actually selected by the customer online, and the branch was just a step in the application process.
It has become clear that Bill Gates’ quote of old about us needing banking, but not banks, has never been more likely an outcome of the technology and behavioral-led disruption we find ourselves in today. Banking was no longer defined or hemmed in by a physical distribution network, or physical artifacts. The banking system emerging out of the global crisis would be one that was highly utilitarian, pervasive, mobile, and seamlessly engineered to work when and where we needed it.
In the end, many of the banks that were household names during the 20th century will simply cease to exist as they are displaced or consolidated in the system-wide disruption that is soon coming. New players are emerging now that are taking ownership of the customer experience through revolutionary new techniques that attack the fringes of “banking” and payments.
The new value is not being a “bank”. The new value is understanding the context banking products and services play in the life of the consumer, and delivering those products and services on that basis. The customer will expect and demand this type of integration. The customer will have no patience for a bank that insists he comes to its “place” before he can have access to banking. Mobile payment is one of the most disruptive areas in the Banking segment. Whatever bank or company operates the mobile-money system will be able to leverage the data for its own purposes (with the right partner, a bank, for example, could offer consumers discounts on a vacation to a favorite destination in addition to offering a savings account to let consumers hoard money for the trip).
There are a bunch of start-ups emerging right now that use such context data to target consumers with ads that are highly relevant. However, matching this to previous card usage data or to purchases such as an airline ticket for future travel makes the pool of data available from within a bank highly sought after.
Cardlytics is a well-known provider in the space. Cardlytics essentially provides an offer-matching capability and mines card data on an aggregated basis to match merchant codes with offers that might be of interest to the bank customer. This, of course, requires that the bank have a relationship with multiple merchants so that offers can be successfully served to a customer.
Some banking products are highly contextual. In fact, many day-to-day banking products are. Here are some examples of the context of core retail banking products:
1. Mortgage (at a potential home or with a realtor)
2. Car lease or loan (at a car dealership or when purchasing a car)
3. Credit card (potentially at a mall, or getting ready for a trip overseas)
4. Travel insurance (when booking a holiday, or at the airport)
5. Student loan (when enrolling at college or university)
The opportunity lies in using this data and extrapolating a customer opportunity with respect to banking. This requires you to know your customers well enough at a minimum to know what offers are relevant to them, and what merchants, locations, etc., their frequency. This data can, of course, be gleaned from existing card usage analytics, and cross-referencing with mobile data.
Say a customer has previously used his credit card to shop at Myer’s. A bank might pitch him a discount on his next purchase at David Jones instead. He won’t need to print a coupon to redeem an offer, he’d just swipe his existing credit or debit card , and receive the discounts as a statement credit after he makes a purchase.
Predictive & Cognitive selling, Triggered offers
Distributed point of impact
As technology continues to progress, point-of-sale equipment will also continue to migrate to the cloud, Internet of Things devices or integrate with our app phone’s capabilities. So the point of impact goes beyond simply the point of sale or a Branch. It includes wherever the sales journey might be initiated with a possibility of closure.
A classic example would be CBA’s Albert which runs on open Pi android platform which will have an open architecture, giving you the opportunity to develop new and innovative apps for businesses, across a range of industries.
With respect to new Banking principles, push based marketing needs to change to pull based, point of impact and service selling. Segmentation and customer intelligence through behavioral analytic is the key to this type of messaging capability. More than simply segmenting customers, banks will need to understand how customers behave, what they do, how often, and through which channels. Currently banks don’t even understand which transactions go through which channel for existing customers. Marketing must understand the why and how, and ask customers what they want. Some banks probably imagine that they do this already, but most of them certainly don’t use that data effectively to sell or match offers to individual customers.
Design & Model your APIs to be Context aware
Under the umbrella of big analytics, context analytics denotes the incremental context accumulators that can detect like and related entities cross large, sparse, and disparate collections of data. The data collection includes both current and historical data. The completeness of the data context enables analytics to correctly assess entities of interest. Creating data within the appropriate context delivers higher quality models. Higher quality models applied to contextually-correct data can lead to better mission decisions and better outcomes.
>>> Understand the API value chain & identify APIs
What business assets are going to be provided through the API? What information, services, and products will be available? Of what potential value could these assets be to others? How will the owner of the business assets benefit from the API? For a mortgage API, the assets are the mortgage data and the points of interest and impact.
>>> Identify channel centric attributes to build the context
For example geographical location information could be an attribute for mobile advertisement context
>>> Identify attributes which can provide individualized customer experience
>>> Identify at-risk customer behavioral patterns & attributes before they are likely to churn
>>> Identify influencing factors in a time-sequenced historical data at scale. Simple classification techniques doesn’t solve all business problems.
>>> Group and classify of products as they pertain to multi channel interactions
Simple Products – No advice required – Credit Card, Current/Savings Account, Personal Loan, General Insurance etc.
Informed Purchase – Advice sometimes required – Mortgage, Life insurance, Overdraft etc.
Complex Products – Specialist advice required – Securities, Investment Funds, Mutual Funds, Deriatives, Structured Products etc.
>>> Build Machine learning and analytic capabilities at API layer and leverage classification techniques to identify and build patterns. Build your own prediction APIs
General concepts for actually extracting knowledge from APIs, which undergird the vast array of data science techniques and predictive analytic tools.
—- Identify informative attributes — those that correlate with or give us information about an unknown quantity of interest
—- Fitting a numeric function model to data by choosing an objective and finding a set of parameters based on that objective
—- Controlling complexity is necessary to find a good trade-off between generalization and over-fitting (Try to identify a data set, leverage existing models like IFW or IFX), Nobel Laureate Ronald Coase said, “If you torture the data long enough, it will confess.”
—- Calculating similarity between objects described by data
>>> Design & expose prediction APIs for App builders. Ultimately the prediction APIs needs to be exposed to API developer to build smarter and predictive apps.
Attracting new customers is much more expensive than retaining existing ones, so a good deal of marketing budget should be allocated to prevent churn. It is easier to engineer the API data to apply the existing data mining tools than to match existing tools.
Cross-sell to existing customers
Create point of impact marketing campaigns & personalized experiences based on context
Predictive & Cognitive selling e.g. simple event triggered offers
>>> Significant Balance change (Investment Needs) : Customer’s account holdings increased by large or significant amount ($ 200k to $ 500k, $ 500k +). Lead delivered to banker next day who contacts customer with offer e.g. a financial planning appointment.
>>> Large Transactions (Withdrawals/Deposits) : A transaction out of the ordinary for that customer, e.g. greater than average for last three months.
>>> Term Deposits (Renewals/ Upgrades) : Relationship Manager contacts customer to renew or increase deposit, perhaps offering better rate through a structured product or something similar.