How to keep fresh products on the shelves
Accurate forecasting optimizes customer service, minimizes inventory overstocks and lays the groundwork for effective marketing at Nestlé
A billion units roll off Nestlé production lines every single day. This number illustrates the sheer quantity of goods produced by the world’s biggest food company. To deliver on its promise of “Good Food, Good Life,” Nestlé has brought to market a whopping 10,000 products aimed at improving consumers’ lives with better and healthier foods and beverages.
To ensure the right amounts of those products make it to the shelves and into customers’ hands, Nestlé relies on forecasting. After all, even the best marketing promotions can backfire if the shelves are empty when the customers show up for their favorite foods.
It comes as no surprise that Nestlé’s interest in closely managing the supply chain and keeping inventories within tight limits is proportionate with the size of its operations. Its sheer size makes planning on a global scale highly complex. Product categories, sales regions and an abundance of participating departments combine to weave a tangled web.
It’s also the nature of the food and beverage industry that makes operational planning a challenge. Seasonal influences, being dependent on the weather to provide a good harvest, swings in demand, other retail trends and the perishable nature of many products make it difficult to plan production and organize logistics.
We’re now able to drill down through customer hierarchies and do things such as integrate the impact of promotions and special offers into the statistical models.
Head of Global Demand Planning Performance and Statistical Forecasting
Tied down by conflicting KPIs
“Supply chain management is a well-established, recognized stream and process at Nestlé,” explains Marcel Baumgartner, who leads global demand planning performance and statistical forecasting at Nestlé’s corporate headquarters. “Our professionals take care of transportation networks, run efficient warehouses and are the first point of contact with customers. One area of focus is planning – or, more precisely, demand and supply planning.
According to Baumgartner, this process tackles two important metrics: customer service levels and inventory levels. One can improve customer service levels – defined as the percentage of complete and on-time deliveries – by expanding inventories. But that ties up capital, and it’s often difficult to find storage space. The freshness of the product suffers as well.
In this industry, products are processed in very large batches to keep unit prices low, ensure quality and take advantage of raw ingredient availability. This make-to-stock production strategy contrasts with the make-to-order principle frequently seen in other sectors such as the automobile industry. “To have the right quantity of the right products at the right place and time, we rely heavily on being able to predict the orders our customers will place as precisely as possible,” says Baumgartner.
Other business metrics, such as budgets and sales targets, are also important factors. The overarching goal, according to Baumgartner, is to be able to “take proactive measures instead of simply reacting.” To accomplish this, Nestlé focuses on strong alignment processes, stronger collaboration with customers and the use of the proper forecasting methodology.
Statistics vs. instincts
There are two main options for generating forecasts. The subjective method is mainly dependent upon on the estimation and appraisal of planners based on the experience they draw upon. The statistical method approaches the forecasting problem with data.
Before using SAS, Nestlé was primarily using SAP APO’s underlying forecasting techniques, together with models from the open-source statistical software R, integrated into APO. Those forecasts were then revised by the Nestlé demand planners. SAS enhances this, and thus complements SAP APO perfectly.
Statistical forecasting tends to be more reliable if sufficient historical data is available. “But one thing has become clear to us — you can’t predict the future with statistics by simply looking at the past. It doesn’t matter how complex your models are.”
So it’s not the statistical methodology that’s the problem for Baumgartner and his team. The critical factor in this complex environment is being able to assess the reliability of forecasts. Two elements have attracted the most attention within this context: dealing with volatility, and SAS.
“Predictability of demand for a certain product is highly dependent on that product’s demand volatility,” says Baumgartner. “Especially for products that display wide fluctuations in demand, the choice and combination of methods is very important. SAS Forecast Server simplifies this task tremendously.
Of particular importance for demand planning are the so-called “mad bulls,” a term Nestlé uses to characterize highly volatile products with high volume. A mad bull can be a product like Nescafé, which normally sells quite regularly throughout the year, but whose volumes are pushed through trade promotions. A simple statistical calculation is no more useful in generating a demand forecast than the experience of a demand planner for these less predictable items. The only way out is to explain the volatility in the past by annotating the history. Baumgartner and his team rely on the forecast value added (FVA) methodology as their indicator. The FVA describes the degree to which a step in the forecasting process reduces or increases the forecast error.
More knowledge, less guessing
According to Baumgartner, SAS® Forecast Server is the ideal tool for this scenario. The solution’s scalability allows a handful of specialists to cover large geographical regions. And selecting the appropriate statistical models is largely automated, which is seen as one of the strongest features of SAS Forecast Server. “At the same time, we’re now able to drill down through customer hierarchies and do things such as integrate the impact of promotions and special offers into the statistical models.”
The results paint a clear picture. In a comparison between the conventional forecasting method and SAS Forecast Server procedures – for the most part using default settings – the results showed that Nestlé often matches and improves its current performance for the predictable part of the portfolio and thus frees up valuable time for demand planners to focus on mad bulls.
Last but not least, Nestlé emphasizes that even a system as sophisticated as SAS Forecast Server cannot replace professional demand planners. “Particularly for mad bulls, being connected in the business, with high credibility, experience and knowledge is key.” With more time available to tackle the complicated products, planners are able to make more successful production decisions. And that means really having enough Nestlé ice cream at the beach when those hot summer days finally arrive.
Ensure the right amounts of products make it to the shelves and into customers’ hands. Manage supply chain, plan operations and organize logistics on a global scale based on a variety of influences and factors.
Reliable forecast methods free up time to focus on demand planning for highly volatile products. More successful production decisions ensure products are available when customers want them.
Nestlé is the world’s biggest food company. More than 330,000 employees work at 469 locations in 86 countries to generate annual revenues of more than 90 billion Swiss francs. These sales figures make Nestlé the global market leader by a large margin.