The Covid-19 pandemic has disrupted everything from consumer behavior to supply chains and the economic fallout is causing further changes. The data analytics field faces a complicated problem: How to use past data, and predict future behavior in the face of uncertainty.

Organizations have moved towards predictive analytics in the last several years, as they use data to forecast future trends. The pandemic’s unpredictability is however now posing a big challenge in predicting the future and hence the concern on the reliability of past data in decision making for present and future.

“Today, from personalized marketing to supply and demand patterns, finance to customer service, historical patterns of behavior and assumptions are the basis of predictive algorithms. Historical data has always contributed to the main drivers and these models produce tremendous value and powerful decision support capabilities. But, Covid-19 has impacted us in every aspect of life, from how we live, to how we work, to how we spend, to how we commute, and hence disrupted data patterns significantly,” said Prashanth Kaddi, Partner, Deloitte India.

Data models rely on historical data to predict the future and deliver conclusions. However, the data sets do not take into account events as pandemic or lockdown. Many sectors rely heavily on predictive analytics to meet demand and because of this disruption in data, the businesses are being impacted largely. But an array of minor fixes, modifications and new techniques in the data analysis process can help organizations build a robust framework and overcome the Covid-19 data impact in short as well in long-term aspects.

While Covid-19 has been a significant impact event across the board, it has impacted every industry in varied ways – so data scientists need to dive deeper and look at a case to case basis with added industry insights. The critical element here is to separate the “Covid effect” which is a result of the disruption from the “base effect” which would have happened otherwise through analysis.

Lendingkart has taken this challenge as another opportunity to simulate these factors and multiple scenarios in its self-learning underwriting model that will be updated inline with growth and changes in the ecosystem. This will also strengthen our risk assessment capabilities and build robust early warning systems for such testing periods of time. We plan on developing and driving hyper customized products for our customers in partnerships with corporates by integrating APIs to offer best suited financial products to MSMEs like supply chain financing, marketplace lending, and cross-sell products in coming times to best recoup post this wave subsides,” said Abhishek Singh Chief Analytics Officer at Lendingkart.

Lendingkart is constantly looking at its customer data and segmenting them into actionable sets. In its core product of working capital loans this segmentation leads to multiple benefit areas, like categorization of customers to provide the right loan offering to them, enabling varied collections strategy for different sets, and reducing overall portfolio risk.

“Another ability is data on the end-to-end customer journey. This enables us to use analytics for operations improvement and marketing optimizations responding to the changing ecosystem within time. Leads could be prioritized using channel information and product information rather than on potentially unreliable generic or manual feedback or ongoing sentiments that might sometimes reflect a different picture This helps us to focus on the right customers segment upfront and hence save on many inefficiencies,” Singh added.

Managing this atypical data

It is a challenge for a data scientist to build a predictive model, during this Covid-19 pandemic due to the data patterns being disrupted. However, there are certain techniques to navigate out of this situation which accounts for the unforeseen volatility as well.

Godrej Housing Finance is looking for holistic solutions to incorporate any odds that may occur. It is following an agile approach enabling a technological and cultural change that drives enhanced collaboration and automation.

“The organization’s leadership team will need to recalibrate business strategies for the changing landscape, work on new data partnerships, convene interdisciplinary teams with sufficient diversity, and more to minimize the impact on business. The faster an organization can detect unexpected market changes, test new ideas, and adjust, the more successfully it can respond,” said Shalinee Mimani, Chief Risk Officer, Godrej Housing Finance.

In unprecedented situations like this, predictive models can run in autopilot only if such situations have been handled earlier, else models need manual interventions to get the most value from them.

Business Leaders using the outputs of these models often need to augment model outputs and interpretation with human judgment. Analytics leaders must also reassess how we learn from history and weigh in situations differently based on their context. Simpler models, which are more transparent and explainable, also help, but one must constantly adhere to the changes and refine them with changing time,” said Kaddi from Deloitte.

He suggests the data scientist to use the following way to ensure robust models with accurate outputs in the time of disruption:

1. Expanding the organization’s data sources and focusing on where they can obtain insights rather than relying on lagging information

2. Improving decision-making process as now organizations can’t let their models run on autopilot.

3. Enhancing and modifying the input data for a model to account for peaks and troughs.

4. Remodeling with new variables to capture effects of disruptions and creating a robust framework which is agile enough to create refined models faster and cost efficient.

Pre-Covid-19 models have tremendous capacity to give management important insights to help navigate the crisis and the next normal. But the challenge here is not simply a technical one for data scientists to solve.

Source link