Request demo
Please leave your contact information and we'll be in touch regarding your demo shortly
By submitting this form I give my consent to DSLab to process my personal data in accordance with privacy policy

Why fresh food needs a different approach to demand forecasting?

Characteristics of perishables and associated demand patterns to take into account when forecasting demand for fresh assortments
Fresh food is among the most powerful differentiating factors for modern grocery retailers. It drives sales, basket size, and traffic to the store, comprising up to 40% of grocers' revenues. Trouble is, perishable goods are extremely difficult to forecast. Fresh food accounts for 70% of food wasted in the world and billions of dollars of retailers' losses each year. Standard demand forecasting approaches are no good for fresh food. In this article, we discuss the characteristics of perishables and associated demand patterns that retailers need to take into account while forecasting demand for fresh assortments.

Daily forecasting and replenishment

Fresh food - like dairy, bread, meat, and vegetables - is perishable. With a shelf life of 1 to 7 days, fresh food loses the market value every day putting raising pressure on the planning department.
As freshness is key, perishable goods require daily forecasting and replenishment. Demand for some items can change so rapidly that the forecasting system needs to be able to account for and adjust to daily, intra-daily, and at times, hourly changes in demand at store level.
Highly accurate granular forecasts are important, however, there is more than accuracy to it. After all, it's not the percentage of error but the cost of error that makes a difference. Hence, the goal of every forecast is to keep an optimal balance between on-shelf availability and stock level based on grocers' business goals and KPIs.

Hidden out-of-stock

With perishable items, customers optimize their behavior for freshness. Take, for example, fresh milk with a shelf life of 7 days. We have two identical bottles of milk - the same brand, the same price, the same fatness - but one expires in 7 days and the other expires in 5 days. While for retailers' accounting systems these two bottles are the same - actual stock, good for consumption -, the majority of customers tend to choose the "fresher" option.
Eventually, items with closer expiry date become a hard sell leading to hidden out-of-stock and write-offs. On a bigger scale, it means that new shipments sell better than previous ones.
Typical for fresh food, this type of customer behavior is uncommon for non-perishable goods and thus is not taken into account in the standard demand forecasting approaches.

Seasonality and price fluctuations

Demand seasonality is nothing new. With fresh food, however, products have a seasonality of their own. Think fruits and vegetables available only in a certain period of the year. During these periods of time consumer demand switch between different categories. For example, in the northern hemisphere during summer months customers prefer seasonal goods - like strawberries and apricots - to available year-round apples and bananas.
To satisfy customer demand and avoid excessive write-offs, retailers need to account for seasonal demand switches that are excluded from the standard demand forecasting process.
Seasonality in pricing is another influencing parameter in fresh food forecasting. No doubt global supply chains make it possible to enjoy even the most exotic fruits no matter the season, however, off- and on-season prices may differ ten-fold.
While being common for fresh food, such price fluctuations are often seen as an anomaly in the standard forecasting systems.

Product appearance

When it comes to fresh food, appearances matter. External imperfections and signs of aging in fruits and vegetables often result in major markdowns. While this allows retailers to minimize food wastage, it also cuts margins. To avoid it and adjust replenishment strategies, retailers need to constantly monitor the appearance of the produce.
Modern computer vision technologies can automatically monitor the exterior of fruits and vegetables, defining their age and forecasting best-by date.
This data can also be taken into account in the demand forecasting process to further optimize replenishment strategies and inventory-allocation decisions.

With demand forecasting, there is no one-size-fits-all solution. At DSLab we've developed a forecasting solution tailored exclusively for perishable goods. It takes into account all the specificities of fresh food and associated demand patterns such as seasonality, price dynamics, storage conditions, hourly stock on hand, next delivery date, advertising, public holidays, and even weather. Bayesian methods of machine learning at the core of our solution align with business requirements and allow retailers to assess the potential influence of out-of-stocks and write-offs on key business metrics and find the balance between on-shelf availability and stock levels depending on business goals and KPIs.

To learn more about our approach and technologies behind DSLab demand forecasting for fresh food, read the article Fresh food forecasting. Challenges of demand forecasting for fresh and ultra-fresh assortments
Lada Trimasova
Head of Predictive analytics group at DSLab

More posts you may find helpful

Contact us to learn more about ML-based forecasting and how it can benefit your business
Contact us
Get in touch to see how AI can benefit your business
By submitting this form I give my consent to DSLab to process my personal data in accordance with privacy policy