How 2020 Impacted 2021’s Predictive Modeling

COVID-19 disrupted the world and predictive fashions in 2020. In 2021, we’re considerably higher ready to cope with excessive uncertainty.

Picture: momius –

Earlier than the pandemic hit, digital disruption was the foremost drive shaping the course of enterprise fashions and whole industries. As 2020 demonstrated, digital disruption appears comparatively tame now in comparison with COVID-19’s impacts on enterprise. The sudden and dramatic shifts in on a regular basis realities negatively impacted predictive mannequin accuracy as a result of they have been so inconsistent with historic knowledge.

“One of many actually massive issues that folks grappled with is the truth that they took without any consideration that the fashions have been constructed correctly,” stated Scott Zoldi, chief analytics officer at decisioning platform supplier FICO. “Clearly we have been in a time of giant stress in order individuals have been making an attempt to grasp tips on how to pivot their enterprise, as a substitute of asking, ‘How can I leverage the asset I’ve?’ they principally stated, ‘Let’s simply throw the mannequin out and construct a brand new mannequin’, which comes with a complete set of different points as a result of we basically then have fashions constructed on nonstationary knowledge.”

Adapting to the New Regular

Firms have historically had years of knowledge that may very well be used for predictive functions. Nonetheless, when all the world modifications so radically in such a short while — provide and demand, provide chain disruption, shuddered companies, keep at dwelling mandates — it is time to get artistic.

“A conventional method can be to take a look at gross sales and perceive traits. Now gross sales is just not a great predictor so it’s important to search for one thing else,” stated Dan Simion, VP of AI and analytics at international consulting agency Capgemini North America.

Dan Simion, Capgemini

Dan Simion, Capgemini

For instance, considered one of Capgemini’s airline shoppers is utilizing future bookings to foretell enterprise journey as a substitute of gross sales as a result of the demand for enterprise journey evaporated in 2020.

Whereas firms understood they wanted the power to adapt to alter rapidly, additionally they understood they wanted to reduce dangers through the use of knowledge to make selections.

“Attempting to foretell utilizing conventional statistical methods will get more durable and more durable since you want numerous knowledge factors and observations,” stated Simion. “Earlier than you’d have one statement every week or every day and you might go down to each hour.”

The identical factor holds true for different dimensions that may be decomposed into smaller items — zip codes as a substitute of nations or areas, for instance.

“That’s growing the levels of freedom, the variety of observations throughout the similar timeframe,” stated Simion.

Contingency Planning Is “In”

At a enterprise technique stage, organizational leaders have been warned that they wanted to do contingency planning at a wholly totally different stage than they’d earlier than. As a substitute of getting a plan A and a plan B, international consulting organizations have been advising shoppers to have a number of contingency plans masking totally different situations corresponding to lock downs and provide chain disruptions. Nonetheless, the identical kind of pondering did not trickle right down to the info crew in lots of organizations.

“We’re seeing a pickup in demand, particularly these days,” stated Simion. “The query was, ‘What’s the contingency plan?’ and now it is ‘What are my choices for delivery route if I am unable to ship via conventional routes? The place ought to I place my containers to account for that?”

Why FICO’s Predictive Fashions Weathered the Pandemic Higher Than Most

FICO had fewer challenges with its predictive fashions in 2020 than most different organizations. Then once more, prospects depend on its fashions to make vital enterprise selections corresponding to whether or not to situation credit score and at what stage.

Scott Zoldi, FICO

Scott Zoldi, FICO

“Previous to COVID, we have been all the time criticized. Why do your fashions take so lengthy to construct when this Fintech over right here can do it within the cloud [a lot faster]?” stated Zoldi. “We’d say to the shopper, ‘You and I each depend upon this mannequin and subsequently now we have to grasp it rigorously and now we have to construct it rigorously.”

A part of FICO’s secret sauce is a four-prong methodology that features:

  1. Sturdy AI, which focuses on mannequin efficiency and stability
  2. Explainable AI, which is about understanding relationships in a mannequin, together with what the mannequin is studying
  3. Moral AI, which entails testing to make sure moral outcomes
  4. Environment friendly AI ,which captures info from the earliest levels to:
  • Perceive the info
  • Do state of affairs testing
  • Determine whether or not the behaviors that drive the mannequin make sense
  • Perceive what to observe

Zoldi additionally underscored the significance of a governance mannequin or mannequin improvement governance mannequin.

“If you do not have a course of written down and codified to determine that from this level ahead, we’re solely going to make use of these applied sciences, have these sorts of individuals overview the mannequin, these are the requirements for what it means to construct a strong and accountable mannequin, and out of that might come issues the group would need to monitor to ensure the mannequin is performing correctly,” Zoldi stated.

In a forthcoming report sponsored by FICO, 90% of the CIOs, chief knowledge officers and chief AI officers surveyed stated they should make elementary modifications and funding in how they monitor their fashions.

“I believe if 90% of analytics leaders in these totally different companies say now we have an enormous quantity of labor to do in monitoring I believe that is in all probability one of many massive issues to take a look at in 2021,” stated Zoldi. “The opposite factor to concentrate on in 2021 is that if fashions are constructed correctly and punctiliously, you do not lose their predictiveness however their interpretation modifications just a little bit which means you may use a distinct rating threshold than you probably did earlier than.”

Associated Content material:

Making Predictive Analytics Work in an Unsure World

How IT Can Get Predictive Analytics Proper

IoT and Predictive Analytics: What We’re Driving Towards

Why Everybody’s Information and Analytics Technique Simply Blew Up


Lisa Morgan is a contract author who covers massive knowledge and BI for InformationWeek. She has contributed articles, studies, and different forms of content material to numerous publications and websites starting from SD Occasions to the Economist Clever Unit. Frequent areas of protection embrace … View Full Bio

We welcome your feedback on this matter on our social media channels, or [contact us directly] with questions concerning the website.

Extra Insights