Businesses worldwide are scrambling to respond to the COVID challenge. The International Monetary Fund (IMF) predicts the global economy to shrink by over 3 percent in 2020 – the steepest slowdown since the great depression of the 1930’s.
The World Trade Organization (WTO) economists believe the decline will likely exceed the trade slump brought on by the global economic downturn of 2008-09.
As a consequence business leaders need to take rapid decisions on controlling costs and maintaining liquidity while readjusting to the new realities. The need to redefine correct priorities while staying agile and iterating changes on the fly, is what will ultimately decide on who will live to fight another day.
The role of Data, Analytics and AI/ML are taking centre stage during the pandemic, collecting vital COVID 19 data, such as transmissibility, risk factors, incubation period and mortality rate. This data is being used for visualizations, and creating mathematical models that is guiding governments on current disease detection and control strategies.
Current media reports are flush with cases highlighting how companies like IBM, John Hopkins University, Blue Dot, Insilico Medicine, Medical Home Network, Google’s Deep Mind, Covid-net, Esri etc. are all leveraging Data and ML Models to combat the situation.
These efforts though commendable are all burdened by the new challenges that enterprises face today: –
- Lack of historical data on context since this is a new situation
- Model training and simulation becomes quite challenging
- Older models becoming invalid or need course correction
- Real time responsiveness needed due to the rapidly evolving scenario
- Existing issues of data quality now exacerbated hence need for more robust solutions.
With this as the context, let’s explore the challenges for businesses especially for the banking sector worldwide. In the Pre-Covid world, the banking sector had already been grappling with how to respond to the digital revolution and how to innovate and be relevant to its customers.
Let’s begin by asking the following questions:
What promise do AI/ML technologies hold in the Post COVID scenario for a bank? And more importantly, what are the necessary data building blocks that organizations need before they can accelerate on the digital transformation journey?
How can the existing data, analytics investments be leveraged better and how to course correct on the data analytics front to respond to these new challenges?
In simple terms,two key ways that banks can combat this crisis is by generating business and recovery scenarios based on forecasts and predictions, and secondly, setting in place tools and frameworks that facilitate early discovery of business trends, that generate alerts and share business-critical information across an organization’s operating landscape.
The assertion of basing business continuity management action on data analytics is all good and well taken – but, exactly how doable is it in the current scenario?
Let us explore this through a few tangible questions that banks, investment companies, insurance companies, real estate firms and the other industry players face presently.
In the COVID times of global uncertainty, do banks know how customers behave when interest rates change? What kind of retail products (savings, wealth, insurance) do such pandemic scenarios promote (or demote)? Does the financial industry understand the impact on the customer and commercial lending sector? Will the banks have definite sales projections, or business volumes for the upcoming festival season and the new year? What about extrapolations and forecasts for credit card spends? Will the insurance industry have customer growth or demand numbers by product lines? Predictably enough, health insurance sector may see growth, but by how much? Will higher infection zones correlate to larger loan risks and loan delinquencies? In light of the exponential growth of digital payments and e-wallets, are there any new business models or profit sharing mechanisms that banks can conceive?
The list of questions can be interminable. But their answers indeed are changing in the new normal. And, in ways we do not understand deeply enough or in quantitative terms.
No matter how you see the COVID scenario(Is the glass half full or half empty?),the fact is that our legacy data structures, databases and data models aren’t capable to provide accurate insights for us to rely or predict the COVID future. However hard we may try we cannot know the current situation’s impact basis last year’s data. This mean that the situation calls for quick analyses on the mountains of different data-sets coming in. Right Now. Let us explore the key areas where urgent action is needed on this front:
Data Sources , Data Currency and Data Quality
Since the past trends are unlikely to help, and the fact that incoming data is likely to be unstructured, unclean, qualitative and anecdotal, financial organizations have to gear up for creating capabilities to analyse real time, broad-based, ad-hoc and granular data. This dynamic kind of exploration will undoubtedly bring its own challenges like deciding – what are the various data types, data sources, data structures that will need to be consolidated?
Adoption of Data Virtualization technologies becomes more important to tap into sources as they exist. For example,joining between an older Datawarehouse and a current CRM database might become very urgent to generate certain critical reports without creating permanent data structures. This can be achieved more easily by Data Virtualization tools.
In terms of Data Currency and Latency, already many banks are moving to real time or near real time Data Ingestion platforms based on the likes of Apache Kafka or AWS Kinesis. Leveraging new methods of detecting Data quality issues like anomaly detection using ML is becoming very critical given that traditional methods are time consuming. Also Self correction of data issues will become centre stage as we move forward.
Data Science, Analytics and the promised land
Next, let us examine a possibility brought on by the situation with an example. Say, there is a rural consumer lending company specializing in loans for buying farm equipment. And there come in a slew of government changes that mandate how rural loans are to be structured, or say announce changes to the moratorium conditions, or perhaps reduce the interest rates. Now for the same company, all such changes have suddenly opened a completely new market. Undeniably, the company will need to quickly respond for the new loan sizes, new loan types, new risk profiles, new collaterals and the like.
Another scenario could be the consumers’ response to the health scare, is seen in the shrinking of the two wheeler loan market and a simultaneous rise in the small car loan segment. Along with the automotive industry, how can banks analyse and validate this hypothesis?
In reality,Data Science and Machine Learning tools promise was to make all the above a seamless experience. However the reality is that most of the data science work is still not enterprise grade and is evolving at a slower pace than envisaged in terms of adoption.
Banks would need to urgently look at the following aspects of data modelling, model management, model deployment and maintenance far more seriously. The elements of model drift and the validity in terms of dimensions and algorithms may need to be reassessed as well. For deployment, it is important to test and deploy on a larger scale for more users by leveraging the real power of distributed architectures like Apache Spark.
There are various tools in the market today that assist in streamlining this process, the key ones being Data IKU, Data Robot among others. If implemented well, these along with the right team of data scientists and engineers can reorient the existing investments in Data science and ML to readjust to the new requirements much faster.
Investments in Data and Analytics Technology – Getting bang for the buck
There is another matter of considerable importance: what if the COVID situation is a temporary one lasting for 12 – 18 months (let’s say, it turns out to be a temporary normal as opposed to a new normal). If any investments are made in data storage or processing or new analytics tools, will these data technology investments turn out to be use-and-throw?
In fact, for this dynamic-analysis-environment to churn out insights for newer decision making models will need fresh thinking around the existing tools, newer frameworks, and flexible hardware investments. Pre-COVID, most financial institutions relied ‘on premise’ models either on enterprise data platforms or Open Source Big Data platforms on the Hadoop ecosystem.
Banks had not adopted cloud platforms for data storage mainly on account of regulatory and Data privacy norms. But given the present maturity levels of cloud solutions, it presents a cost effective and a time opportune strategy for financial institutions to move to the cloud with its native data analysis, and data visualization solutions.
As mentioned earlier, the rate at which data is churned and analysis carried out today will need to be exponentially ramped up or down. The Cloud vendors have come out with excellent distributed processing engines comparable or better than the Open Source options whether it is for ETL/ELT or just processing complex ML jobs, Query Engines, In Memory processing and large scale data lake platforms. However given the costs of the cloud, it is important for banks to evaluate a mix of Open Source and Cloud native tools with cloud infrastructures as the back end.
The goal here is not a complete replacement of the existing investments in the On Premise data infrastructures , but to emphasize that this is an opportunity to quickly leverage the cloud or a hybrid mix of tools in this situation.
Conclusion
Finally, the players which win this battle of ‘ad-hoc and rapid analyses’ required to stay viable in the COVID economy will turn up as leaders of the new normal.
Conversely, the failure to see the COVID situation; popping open ‘new market segments’ or ‘splitting traditional ones’ or the rapid reengineering of business models; will severely handicap organizations’ growth and market capture strategies.
This was originally published on www.globalbankingandfinance.com website and is being reproduced here.
