How to Forecast During Unprecedented Uncertainty

By no means do I consider us out of this pandemic. There is plenty ahead with unexpected twists and turns; however, now that we are past the initial phase, it is an excellent time to reflect to prepare for what's next.

Never Forget the Underlying Sentiment.
Contrary to many of the experts, our clients did not discard their forecasts when this pandemic started. But getting through that initial shock . . . sucked. And while things have settled, it’s not over. There will be plenty of unexpected twists and turns. But this time, we all, including our clients, are better prepared. Below are key lessons to take in managing and improving your forecasting results during any time of stress and uncertainty.

Collaborative Communication.
Because of these circumstances, open, bi-lateral communication is vital. Before COVID, we typically held weekly or bi-weekly meetings with our clients. As soon as COVID hit, we started having daily touchpoints where we'd all share what we knew, what we didn't know, and what we all were speculating, as well as what we were doing. These times require honesty and that can be brutal at times. But that enabled us to be creative and come up with solutions to manage these situations. But back to my first point. We need each other.

Patterns were strange. Why?
Communication was crucial because none of us had seen these patterns before. We needed to interpret the ’why’, ’what is causing them’, and ‘what to do’. Since we work with a diverse client base, everyone was experiencing things differently. Our apparel retail clients closed their stores, but not everyone shared the same surge in online business. Our Food & Beverage CPG clients experienced unexpected growth and stockouts, other Consumer Products clients underwent massive channel sales shifts with some categories up, some down, and others not affected at all.

Here are a few simplistic examples, which with hindsight, are easy to explain:

  1. We saw a massive increase in mascara sales and a drop in lipstick sales occurring in different states and times. The cause: mask mandates. People show their eyes, but not their mouths.
  2. Personal grooming products and hair clippers immediately surged. The cause: hair salons and barbershops closed, so people started buying their own for home.

Forecasts were not Discarded.
As mentioned, contrary to many of the experts, our clients did not discard their forecasts. They didn't blindly follow them, either, because some forecast accuracies went to zero (Prior point: honesty). But not every forecast was inaccurate. Some retained their accuracy while others, to our surprise, were better. But all of them provided insights. The questions were, "What were they telling us? Why were they telling us this? What's the next action to take?" The best answers weren't by machine or man, but the combination of man and machine.

A lot of trials. Even more error.
Science is about experimentation, trial, and error. And during this period, there were many trials and even more errors. We had these fantastic ideas, and then we would try it . . . and fail miserably. Those were humbling experiences. On the other hand, sometimes we’d throw stuff at the wall to see what would stick and, amazingly, some things worked brilliantly. But that is science, and sometimes progress happens that way: happy accidents.

Elegance in Simplicity.
For complicated situations, you think you need complicated solutions. We love trying out new techniques, the latest algorithms, and jumping into all the models. We are an AI company. But there were times where we gained incredible value from doing extensive data analysis and descriptive analytics. Why? It showed us what was going on and identified data issues. We didn't get caught up in making an existing solution solve the problem. We focused on finding a solution to the problem, whatever that solution was.

We Became Better.
While this experience was extremely stressful, I take great satisfaction in how we responded. We proactively worked with our clients on a common goal, increased our communication, troubleshot problems, and employed a lot of experimentation. As a result, we've improved our ability to predict consumer demand. Our reporting and analytics have improved. And finally, we've enhanced our own AI and machine learning techniques to tag these types of events, consider leading market signals, and respond even more quickly to consumer demand changes. While we're never sure of what's next, we and our clients are better prepared.