Excerpt from roundtable during Tango’s virtual retail summit The Next Normal for Retail Location Strategy on May 19, 2020.
Modeling Techniques in a COVID-19 World
Pranav Tyagi – President & CEO, Tango:
“When you think about projecting the future, the historical tendency always has been to look at past behavior as a good way to forecast the future. Shai talked about supply chain – one of the big fundamentals in the supply chain is to look at store operations historically and then decide what is needed to sustain them or grow them at a certain percentage. In this environment, deciding which stores to reopen and when need to be well thought out. It’s not just a matter of ‘if you’re allowed to open, you should’. There’s a lot of non-traditional data sets, things that people have not used historically that need to be a factor in the immediate term, in terms of deciding when to reopen and which stores to.
“For example, we launched a COVID-19 Retail Benchmark Tool a few weeks ago that allows retailers to track and visualize several key metrics around their locations, such as cases per million, mortality rates, unemployment rates – including unemployment on an hourly worker basis, which is a key part of the labor pool that works inside these retail locations as well. There are some additional non-traditional data sets, such as cell phone activity. One of the interesting trends we’ve seen is the movement of those cell phones and how they’ve changed as different jurisdictions have opened and allowed for different types of movement within the retail and/or employment areas. But I think cell phone data traffic is an interesting data set that needs to be looked at. Also, changes to vehicular traffic from providers like Inrix would be an interesting one, as well as UberMedia, Inrix – those types of vendors.
“Understanding what is happening at a ground level at a county-by-county level is critical to making these reopening decisions. This data is not easy to interpret because what’s happening in an aggregate county may not be directly what’s happening at your individual location level – that there will be hotspots inside the county as well. One of the most challenging things for our customers has been to interpret and respond to the details of all the various governmental orders.
“And then most importantly, we also need to look at the changes in demographic behavior. Let’s say a retailer has the ability to open a location in a specific part of the country, but the COVID cases are increasing and employment is decreasing and their target customers are baby boomers. A location in this situation – even if you could open – is not a great candidate to open. Companies have to think about their sales forecasting and market optimization modeling a little bit differently. We feel that the ground has shifted very fundamentally, and past performance is not really a good predictor of the future. And then that means that traditional modeling techniques are no longer as relevant as what we believe right now because historical data is not a good way of predicting what a post-COVID environment is going to look like. So we need to incorporate time series data – data that’s changing on an ongoing basis.
“Annual updates used to work where demographic data was collected once a year, that’s not going to the change work here. Retailers don’t have the luxury of time. What they need right now is a highly adaptive model that can predict the future in a dynamic environment. Things change every day, and therefore, you need to use things like machine learning where models can consume an unlimited number of data sets, streaming data, constantly changing data. Customer buying behavior in the last week should influence what’s going to happen next week and beyond.
“These types of models can look at customer buying patterns and provide a far superior accuracy versus traditional models. And because of the technological advancements that have happened over the last couple of years, you can actually build these models in a matter of hours, where it used to take days, weeks, and months to do that. We believe that machine learning and artificial intelligence have a huge role to play in this changing environment from an analytical perspective.”
Want more? Get access to the full on-demand virtual summit here.
Interested in learning more? Schedule a demo.