Sales Forecasting with COVID-19

"We decline to provide forward-looking statements as a result of the uncertainty surrounding the ongoing pandemic." Sound familiar? Pretty much every earnings announcement includes some version of those words in the past few months. Yet businesses need to continue to operate and a very large part of ongoing operations is developing a credible forecast for inventory replenishment, labor scheduling, and sales and marketing support. So how can companies forecast in a way that provides credible assumptions for these activities? Perhaps more importantly, how can companies "flex" forecasts when underlying assumptions prove false to adjust in a rapidly changing marketplace?

Start with scenario modeling. If you don't have your own scenarios, most investment banks and consulting firms have their version of the truth and you can find many if you're a client/prospect for free. Caveat: almost all of these feature a rapid V-shaped recovery and I have yet to find one with a realistic timeline to vaccine (no vaccine has ever been developed in less than 5 years, with MMR leading the pack at 5.5 years; typical is 10-12 years). That doesn't mean your business won't recover in a rapid V-shape. It just means that those that do will have businesses either unaffected by COVID or positioned to adapt and thrive.

A good set of scenarios contemplates a range of potential futures, rooted in macro-economic metrics (as noted above, available from many sources, skewed overly optimistic), developed with clear stories (how the scenario came to be, what a day in the life of the consumer looks like, and the implications to your business), and demarcated by clear indicators for the scenario in which the business is operating. Against these scenarios and metrics, forecast baseline industry and company sales. This is your starting point.

COVID presents a fact pattern rarely experienced in history; therefore, the baselines produced by regression will be wrong (by a lot). How do we adjust our forecasts for this rare event? Start with behavioral proxies mapped to COVID-driven trends relevant to your product categories. For grocery, I knew that there was a fundamental shift between restaurant dining and retail grocery during lockdown. Some of that shift would persist due to COVID concerns until COVID is largely eradicated and at varying levels within the population (i.e. some people are less risk-averse and/or less risk-aware). In addition, the trend towards healthy-eating would be accelerated by public awareness that dietary changes can improve immune response.

To adjust baseline forecasts for these trends, I looked at weekly category and restaurant sales data from sources like Nielsen, IRI, and restaurant associations, along with annual healthy retail category sales trends from sources like Euromonitor. Extrapolating these curves through fitted exponential decay and Bass diffusion models provided adjustment factors that modified baseline sales created through macro-economic ensemble models.

For more specific sub-category models in client work, I spent more time in data engineering from a variety of initial data sources to fine tune historic data across countries and to adapt COVID recovery and macroeconomic recovery scenarios for each country, leading to long-range, multi-scenario planning forecasts that supported a disproportionate investment thesis in certain product categories.

It's more than possible to forecast and plan in uncertainty. The above is one framework for how I've approached this across a few projects. Companies that plan well stand to steal share from competitors that are less prepared.

---

Ed has 15+ years of experience consulting for companies to develop insights from data. He leads data science and reporting teams, programming machine learning models in python, building executive alignment around metrics, and managing product development. He also helps clients build analytics capabilities to self-serve insights. His experience spans Retail, CPG, Pharmaceutical / Life Sciences, Health Care, Insurance, and Telecom/Media. He has a strong “come with solutions” attitude and a commitment to process improvement and ethical leadership.