Andreas Gärtner of Nestlé on Forecasting: “Track the Mad Bulls!”

Forecasting product demand will never by 100% perfect but it is possible to reduce the error of judgment. Nestlé is revitalising its Demand Forecasting process in Europe, relying on analytics to predict the demand for low-volatile products. This is giving planners more time to forecast demand for promotions and the volatile products, the so-called Mad Bulls.

By Helen Armstrong  (Supply Chain Movement)

For almost 150 years Nestlé’s strength has been its closeness to customers. As the Swiss-based food giant has expanded around the world its desire to please customers locally, whenever, wherever and however has been a key to success. Today it produces 1 billion products every day from baby-foods to chocolate, sport nutrition products to bottled water and coffee. And with around 339,000 employees and 468 factories it is, to say the least, a huge company.

In 2001 Peter Brabeck, then CEO of Nestlé, recognised that as the company was getting bigger and bigger its size may become a liability. The global organisation needed a local component, which gave rise to the Nestlé GLOBE Centres, one each to cover Europe, the Americas and the Asia, Oceania and Africa. They were set up to accelerate performance by harmonizing business processes and to standardize and manage data. The GLOBE Centres would also standardize the IT systems, taking advantage of Nestlé’s worldwide scale whilst retaining the ability to act locally. At that time nearly every Nestlé country had its own hardware and software systems and the company hardly knew how many SKUs it was selling!

Future of Demand Planning

Much has been achieved in 12 years but the harmonisation process continues. Andreas Gärtner, Europe SAS Program Manager at Nestlé GLOBE Center Europe, based in Frankfurt, is responsible for the deployment of Nestlé’s Future of Demand Planning initiative in Europe. Speaking during the SAS Forum in the Netherlands, he explained how Demand Planning is in transition and how Nestlé is building on analytics to become a more demand-driven organisation.

“Still the majority of our forecasting is judgmental and therefore subjective,” says Gärtner. “Demand Planners make a Base Demand Forecast which is enhanced during the Nestlé Sales & Operations Planning Process with incrementals (for example information about promotions) supplied by Sales, Marketing and Finance. This results in the final demand plan.

“However, we have seen no further improvements in demand planning during the last few years in many markets. Most production is “made to stock”, often in large batches, so while an accurate forecast is essential it remains difficult to achieve. Forecasts are often on the high side leaving warehouses with unnecessary inventory and sales and planners in debate about how to close the gap. It seemed we’d hit a glass ceiling,” he says.

It was time for change. In 2011 Nestlé started to look at state-of-the-art statistical forecasting and predictive analytics to improve Demand Planning efficiency and accuracy. In March this year Nestlé signed a contract with SAS for software that will do just that. SAS as a high performance forecasting solution complements well the Nestlé Demand Planning solution which is based on SAP APO.  (Advanced Planning and Optimization).

Clear segmentation

“We know with forecasting that no one approach fits all so we need segmentation,” says Gärtner. It is an important part of the forecasting process. In SAS we created codes for segmentation of the product portfolio.” Products are segmented into four categories depending on the volume of the product and volatility of demand. To make these categories tangible they have been given names of animals:  Horses (large volumes, low volatility); Mules (small volumes, also low volatility); Jack Rabbits (low volume but jumping around); and Mad Bulls (large volumes and very volatile).

“It’s a good segmentation and it provides Demand Planners with easy terminology with which to communicate. By making it easier to explain why a forecast is created the planner is better able to challenge numbers presented from sales and see the gaps earlier.”
Gärtner explains that they also use “puppies” to describe new products, “mammuths” for old ones and “kangaroos” for promotion-only SKUs!

More time for Mad Bulls

“The forecasting performance for Horses and Mad Bulls was okay at Nestlé prior to introducing SAS. However, we’ve found that SAS can generate a forecast at SKU-national level for a whole market in a few minutes. That is much more efficient than the more judgmentally based processes used before, and with a similar and often higher accuracy and importantly less bias,” he says.

It means the planner doesn’t need to invest time on these products but instead can spend more time on forecasting demand for volatile products (the Mad Bulls) and on the increasing number of promotional activities. Here demand is less predictable yet there is an increasing need to forecast accurately per week and per customer. SAS will help Nestlé to control these Mad Bulls through the introduction of causal time series forecasting methods. These methods can incorporate information like price discounts, highly correlated with demand uplift, directly into the statistical models.

Getting acceptance

“The perfect scenario would be to use the analytics for the Horses and Mules automatically. But first we have to convince the planners to accept it because they have the final responsibility for the total forecast. They would only need to monitor the statistical forecast coming from SAS and possibly take corrective action if there was a known big change, such as a promotion, or when the statistical analysis fell outside a given threshold.”

Segmentation is done regularly, once per month, and alerts are produced if perhaps a Horse has changed to a Mad Bull. The planner can then look at the history and see what has changed to determine if it’s a trend or an exception.

“Then the next step is to have one single statistical model to simultaneously forecast the base and uplift in forecast. Sales provide the promotion tactics (price, period, promotion type, etc), and the statistical model estimates the impact on volume.”

Within Europe, Spain and France are the most advanced businesses in this respect. Demand Planners have explained this segmentation methodology to Sales who now accept the analytical forecast for Horses and Mules almost without a challenge. “They know it is from an objective source and it’s accurate. They now focus their monthly demand planning meetings on the Mad Bulls,” says Gärtner.