In the current fast-paced and uncertain economic environment, businesses are struggling to adapt their planning and forecasting methods. The threat of inflation looms large and fuels the uncertainty.
When inflation rises, policymakers grapple with tough questions. Is this a transitory or permanent phenomenon? How is it affecting labor markets? How informative are the various inflation metrics and labor market indicators? The Policy Uncertainty Index illustrates how much uncertainty there is when it comes to inflation.
In such an uncertain environment, forecasting OPEX and price evolution becomes not only more important, but also more difficult for businesses. To create as much certainty as possible in an uncertain environment, the concept of "nowcasting" has emerged; it supports real-time predictions for very short time frames.
And this approach has gained a lot of traction among our customers! The year 2022 showed that businesses that rely solely on long-term forecasts, often based on macroeconomic data and expert consensus, can fall behind their competitors relying on nowcasting.
Can I rely on nowcasting for critical growth or investment decisions?
Despite the broad appeal of nowcasting, businesses need to determine when to rely on this approach for critical investment decisions. Different industries have different balance sheet structures, CAPEX/OPEX mix, and investment horizons.
For capital-intensive industries with 5-to-10-year investment decision windows, long-term forecasts and understanding of industry trends are irreplaceable. But even for these industries, manpower, OPEX costs, and lead time for capacity changes are significant drivers of agility and profitability.
For customers in fast-moving industries such as retail, where decisions on variable costs have a large impact on profitability, nowcasting has a lot of advantages. Having an accurate view of coming trend breaks or higher accuracy for short-terms trends can have a significant impact. Given that labour costs can represent up to 40 to 50% of OPEX, having such costs in sync with leading indicators on revenue can determine whether the yearly margin ends up at 10% or 15%.
We have seen periods when consensus forecasts pointed to a recession at a time when sales for a particular industry were at a record high. The use of nowcasting in this context can help get around this paradox of indicators pointing in opposite directions.
When used properly, nowcasting leading indicators can detect the right signals 2-3 months ahead and illuminate a clear path, despite what experts and long-term forecasts say.
Practical examples from Retail and Energy
For retail customers, Lynx has developed EcoBeats, a tool built from publicly available data such as FRED, Google Trends, and mobility data. It generates short-term predictions for macroeconomic indicators such as inflation, GDP growth, and consumer sentiment.
This data is then used in AI models to adjust predictions trained on customers' target KPIs such as revenues, inventory, etc. This approach enables our customers to adjust decisions for staffing, discounting strategies, as well as store openings and closures.
This last aspect proved to be crucial for many customers during the many months of COVID-related restrictions.
In the energy sector, we recently developed a forecasting model for a utility operator that predicts the hourly electricity consumption for a specific area for the next two days. Power utilities often need this kind of prediction as they want to take part in the short-term electricity or gas markets to trade the optimal amount of energy.
Increasing prediction accuracy by just a few tenths of a percentage point can result in significantly increased profits in the long run.
When building our model, we leveraged our work from an earlier project we had done for a grid operator in Europe. Electricity usage has become a lot more unpredictable in recent months, but the backbone of our original model still proved very valuable. The main idea was to convert the problem to forecast a multiplier between a given hour of the predicted day and the same hour on the previous day. This involves modelling "day-type transitions" as power usage will likely increase.
Nowcasting is not a new approach, but it can be applied in a revolutionary way. It is highly applicable, and often preferable in the current macroeconomic context where long-term forecasts have become less reliable. Trusting the most recent datasets means that sentiment-based decision making is slowly giving way to a more data-driven approach. In conclusion, businesses must consider when to use nowcasting to make critical growth or investment decisions in their industries.
