top of page

Using Unique COVID-19 Signals to Help Businesses Reframe Risk

Author: Maor Shlomo, CEO, Explorium

One of the first things that became obvious once world governments started enforcing stay-at-home orders and country-wide quarantines was that businesses would have little to no time to adapt. To go from fully operational to completely shuttered — and more importantly, to see entire industries do the same — introduced a completely new level of risk and volatility to most companies’ plans.

At a time when machine learning and data science have become increasingly popular across industries, most companies’ existing data has been invalidated because it can’t account for the current reality. At Explorium, we understand the power of data to help chart a smarter course, and we have dedicated serious efforts over the past few weeks to find a solution for organizations that are suddenly flying blind.

The result has been a new way to help industries assess risk by using data about the COVID-19 pandemic shaped into unique signals. These new COVID-19 signals offer a quick and easy look for companies to understand how the pandemic has affected not just them, but any industry and company they interact with. Let’s dive a little deeper into how our signals enable businesses to navigate this new landscape with more confidence.

Building new signals from the ground up

Our COVID-19 signals were built from a variety of sources, and are layered in order to help our users gain a more holistic and comprehensive picture of the current landscape, and of companies’ and industries’ risk tied to the pandemic. We understood early on that to offer a full view of the situation, we needed to focus on internal factors, external policy and industry factors, and geographic indicators.

Taken together, these layers let us give companies and organizations a clearer picture of the overall risk they face, both in terms of the industries they work with, and even specific companies they work with as vendors, clients, or partners. Let’s explore our signals a little deeper:

The first layer: the company level

At the most basic level, organizations need to get a better idea of what’s happening with the businesses they’re working with, whether they’re clients, providers, or partners. The first signals are individual, adding visibility into organizations’ industry, history, social media activity, whether they can continue operating online, and even sentiment analysis around them.

These individual signals let companies start building a risk profile about their own customers that includes a variety of data such as whether they can continue functioning during the shutdown.

The second layer: industry and policy factors

The next step is to factor in those external policies and components which also add to a company’s risk. For instance, a B2B lender may have risk models in place that can already account for internal risk factors but may not have the ability to add in the impact of external policies. Our signals account for a variety of policy-level factors that add to organizations’ and industries’ risk.

The first new signal we created is the “market effect” metric, a proprietary index that offers a view of how each industry has been impacted by the pandemic. For manufacturers, the market effect can help better understand their supplier risk, as they could identify potential high-risk nodes in their supply and production chain. Moreover, it could help them pivot and continue operating to provide key services.

In addition to economic impact, organizations need to understand the effect government policies have on an industry. Therefore, we layered in signals regarding government actions for each industry. This helps users determine whether a company or industry is considered “essential” according to the most recently available governmental guidelines. Another important datapoint organizations need to understand is industries’ and companies’ financial viability during the crisis, so we measured whether they are eligible for government assistance based on the bailout and support packages being legislated and implemented.

So, for instance, a lender may be able to determine whether a company is high-risk based on more than just its previous ability to pay. Let’s say the lender is reviewing a loan application from a new prospective business. While the company’s traditional financials all seem above board, it suddenly found itself asking for a significant sum than it seems it could handle. By comparing its internal data to our COVID-19 signals, the lender could determine whether this new loan request is reasonable, or whether the company is likely to default due to external pressures.

The third layer: geospatial factors

Finally, even in places with less restrictive policies, geographic factors may present an added layer of risk. Our geospatial signals measure geographic impacts that tell our users much more about contextual risks and opportunities for different industries. We wanted to factor in the risk of COVID-19’s spread to different regions into our calculations. For instance, even a company in a low-risk industry — healthcare services, for example — may have a higher risk if it’s in a region that’s heavily affected by the pandemic’s spread.

Moreover, organizations can make better decisions if they can see where the pandemic might spread to, and where potential hotspots may emerge. We focused on measuring the growth factor of COVID-19 cases by country, as well as the predicted growth rate. We also measured geospatial factors on a micro-scale, incorporating footfall data to track consumers’ movement patterns to identify areas that may have higher business activity than others.

Another important measure in our signals is whether businesses in a specific area — and especially offline-heavy businesses — remain open or closed. In this same vein, we track footfall data by region, which gives us greater clarity on consumer patterns. For example, we can understand who is going out, and how many people are staying indoors, which can let us know how consumer-facing businesses like restaurants and retail stores are impacted. Finally, we added in data about consumers’ online and offline shopping habits to give us a better understanding of overall profitability.

Aggregating signals for greater industry visibility

Our COVID-19 signals add greater clarity for anyone looking to understand the impact of the pandemic on their industry. Let’s look at a few examples of how some of our users have applied them to their own models:

  • Lending: B2B lenders need to understand default risk, but most of their models have been thrown into disarray as the definition of risk changes in volatile times. Our signals offer a stabilizing stream of data that add greater clarity and help organizations understand whether potential borrowers may need more help, or even if they should receive loans based on their new risk profile.

  • Supplier risk: Manufacturers need to understand how each aspect of their supply chains and their supplier networks are affected by any development, and COVID-19 has been a major shock to every system. Our COVID-19 signals let them understand where potential disruptions, outages, and bottlenecks might form, and quickly pivot to keep their production and deliveries operating in these crucial times.

We understand that right now, businesses need clarity. As our old data becomes less able to predict successfully, adding new signals that are directly relevant to the situation lets us give everyone greater visibility and the ability to make smarter decisions to continue operating and suffer only minor losses.

You can see our signals in action here, and learn more about how we’re working to bring greater understanding around the pandemic.


About Explorium:

Explorium's automated data discovery and feature generation platform automatically connects a company's data to thousands of relevant premium, partner, and open data sources to extract an optimal feature set based on model impact. Explorium is creating a new category as the first company to empower data scientists with end-to-end automation of data discovery and feature generation — fueling superior models and driving real business impact.

352 views4 comments
bottom of page