The impact of weather on the demand forecast

How Artificial Intelligence (AI) combines weather and sales data to improve demand forecasting

“If only I’d known,” said demand forecasters after a major weekend weather change left empty space on many shelves across the region. Pick the weather event, pick the category impacted, it happens all of the time. But what is it exactly that demand forecasters needed to know? While meteorologists don’t always get it right, their accuracy has improved significantly for a 5-day forecast, and they have a pretty good view for a 14-day forecast. But it’s still not enough information for demand forecasters who need to achieve the complex goal of balancing supply with demand in a fast-moving retail environment where volatility of demand is the name of the game. Reducing stock cover, waste and obsolescence while improving availability to maximize sales for weather sensitive products is an ongoing challenge, but there is a missing link.

Mapping the weather forecast onto the demand forecast

The missing link is the ability to map the weather forecast onto the demand forecast with a degree of accuracy that is based on fact. Knowing what the demand will be rather than guessing what it might be is what demand forecasters need. But without the data and the processing capability, it’s a stormy path. This is where Artificial Intelligence (AI) comes in, clearing the mist and fog away to provide clarity on the demand for products that are weather sensitive, right down to cluster and store level.

Artificial Intelligence marries weather to sales

The way AI works with the weather forecast is quite simple – the weather is used as one of the inputs to create a demand forecast. Along with all the other data that can be fed into the process, historical data about the weather can be combined with sales data for the same period, to show demand forecasters how much bottled water, beer, or any other product they sold the last time that the temperatures soared in late summer. Spotting patterns that might go unnoticed by both traditional statistical forecasting and the human eye, AI takes the mystery out of demand forecasting, although it will never take the mystery out of the weather.Viewpoint: Demand Forecasting and AI

The human feel of weather

AI takes into account all aspects of the weather, recognizing the subtle difference between rain and rain with high humidity, and the difference that this can make on sales of different products. This is the more human feel of the weather and not simply just the bare facts. We know that a sunny day at 15 degrees in the summer feels very different to a sunny day at 15 degrees in late autumn. The feeling of that along with the activity and shopping needs are entirely different for the same set of parameters. Location also plays a part in demand forecasting, as weather conditions can vary dramatically across different regions of the same country, at any moment in time. Other factors that provide input might include sporting events or national holidays where AI can marry the change in demand with any variability due to weather to provide the best short-range forecast.

Extreme weather and demand forecasting

Extreme weather occurs in many regions and it’s part of normal life, with the seasons dictating the likelihood of hurricanes, blizzards or heatwaves. Exactly when these extreme weather events will occur and the potentially dramatic effect they will have on demand are critical questions for retailers operating in those regions to answer. Whatever the weather, AI is the answer.

Learn more in How AI and machine learning improve forecasting accuracy in an age of retail disruption.