Companies are moving to the next phase of pandemic planning, with one eye on the recovery period and the other on what their business and industry will look like as they emerge from and go beyond the crisis. But the future is a blind spot for many companies. Prior business forecasting methods, based on past customer demand, no longer suffice. Prior assumptions must be challenged and, in some cases, deemed invalid.
For companies that do not change their forecasting methods to meet changing demand, scenario planning for both near-term operations and long-term capital allocation will be fatally flawed.
During a recent EY webcast on forecasting for recovery scenarios, only 9% of participants said they were “very confident” in their company’s ability to forecast demand for products or services. In fact, 35% said they were either “not at all confident” or “not very confident.”
The pandemic recovery will be driven by medically defined phases
The lack of confidence is justified. Unlike past downturns, the current crisis is both an economic and a medical one. The COVID-19 recovery will move through medically defined phases; a three-year plan could change in three days, based on when the spread of the virus slows (now), when companies can ramp up reopening, especially as testing and tracking protocols can be put in place (next) and, eventually, when there is immunity, either through vaccines or herd immunity (beyond). Even beyond the medical recovery, human movement and interactions may return, but in completely different patterns and ways that shift entire industries to new models.
The challenge is that the intensity and length of each phase is still unclear, and we will see changes in customer demand across each phase. To prepare for these changes, forecasting must become a core competency, with an emphasis on analyzing data from multiple and sometimes novel sources to understand not only your customers’ plans, but also the potential change in who your customers are and how you deliver value to them. This means that waiting for the market to return to normal conditions as if it will be a V, U, L or W-shaped return to the demand curves can miss the bigger changes happening in the market.
Companies can take several steps to change how they forecast.
Change how you look at your customer
Customer behavior and spending potential are changing dynamically across many industries. The key question is which of those changes will become permanent and how long the temporary changes will last. When will consumers be willing to travel by air, enjoy a sit-down restaurant meal or gather in a movie theater? Will businesses be able to pay rent, and will companies cut down on office space long-term? Will patients be willing to go to a hospital for elective surgery?
In the EY webcast poll, 77% said that changes in customer behavior were the key risk for their company when it comes to forecasting, significantly more than those who identified liquidity and capital restraints (the next most frequent answer).
Knowing their customers have become even more important as their circumstances and behavior have massively changed, how should companies adapt?
Change your data analysis techniques
The information used in the past may have become too static, too imprecise or no longer predictive. For example, low gas prices would normally correlate with increased restaurant traffic, but that relationship has obviously broken down.
In some businesses, social media analysis can be used to improve data forecasting. Outside data on pandemic hot spots, weather data, government regulations, mobility data, consumer sentiment and other measures can be run through regression or more advanced artificial intelligence (AI) neural network models to see what can best be used to augment and inform company forecasts.
One consumer company we worked with brainstormed to decide which data they needed to augment their typical syndicated sales data and other inputs to develop a more accurate, timely forecast. Among the inputs they chose were weather data, cell phone tower data, social media mentions of their products and ZIP-code level unemployment data. Their finance team then performed a regression analysis to see which metrics would show causation. The result was more accurate forecasts that are now updated in hours, rather than weeks.