In order to better understand and use this data, John Galt Solutions recently hosted a webinar, COVID-19: Cleansing History & Managing Impact with Dr Barry Keating, Professor of Economics from the University of Notre Dame and Jeff Marthins, Senior Business Consultant at John Galt Solutions. We wanted to answer three main questions: what’s wrong with existing data, what kind of data businesses need to collect, and how to process this data for the most accurate predictions? This article is an extension of our webinar.
You can watch the webinar for free here and read more about our findings by requesting our detailed article on the topic. Fill out the form on the right and we will send you a copy - or read along for a high level overview.
Most businesses have relied on pattern matching routines like time series models that rely on past years’ data to find trends and seasonality to make predictions about the future. The “problem” lies in the stability of past years. The overall world economy has gone through a steady growth phase for nine years with few outliers that disrupted the entire marketplace. Therefore, the routine and stable business patterns and data that companies relied on in the past to create time series models will not work for future models.
In order for 2020 data to be used post-COVID, recent outside data needs to be collected and applied to the predictions to make the analysis more in line with our current global market conditions.
Outside data refers to data pertaining to external factors that cannot be controlled by the business itself but it still influences business decisions. To find relevant outside data, ask the following questions:
Answers to questions like these give businesses clear and useful insights into the market and thus any data pertaining to similar questions needs to be collected.
Some of the common sources of outside data that are very relevant today include the S&P 500, COVID-19 cases at a regional or state level, Google Mobility reports, social media mentions, etc.
Since pattern matching routines (like time series models) can only make predictions by analyzing past data or by “looking backwards”, it isn’t ideal in the midst of a pandemic. A better alternative to time series models is called multiple regression.
Multiple regression “looks forward” by allowing businesses to input outlier variables like the number of hospitalizations or extended restrictions in your state to create much more realistic predictions. For a data-driven company, this method will far outlive the pandemic as it integrates external data regularly to create a much more complete picture, thus becoming a much more reliable method of analyzing data and creating supply chain planning models.
The idea of collecting and integrating outside data to existing models might seem overwhelming and even disruptive to many companies. In fact, getting started with demand sensing is one of the hardest steps for a company, especially in such tumultuous times. However, this is all dependent on your method and tools.
Make your transition to becoming a data-driven company fast and seamless with our Atlas Planning Platform, a unified continuous end-to-end supply chain planning solution. With Atlas Planning Platform’s automated machine learning, you can easily automate planning decisions, adopt multiple regression, detect changes and patterns by bringing in external data, and balance demand and supply in a post-COVID world.
To learn more, please fill out the form to download the full article on how to cleanse your data post COVID. You can also watch the webinar recording COVID-19: Cleansing History & Managing impact.
If you’re interested in learning more about Atlas Planning demand sensing and analytical capabilities and how it helps other businesses in your industry, please contact us here.