Fresh Food Demand Forecasting: How to Produce Accurate Demand Forecasts for Perishable Products

Fresh Food Demand Forecasting: How to Produce Accurate Demand Forecasts for Perishable Products

Dorota Owczarek - May 7, 2023

As any grocer or retailer knows, fresh food is a high-stakes game. With limited shelf life and a host of variables that can impact demand, accurately forecasting how much inventory to stock can be the difference between success and failure. But fresh food demand forecasting is no easy feat - it requires a specific approach that takes into account the unique characteristics of perishable products, as well as a variety of internal and external factors that can influence consumer behavior.

In this article, we’ll explore the challenges associated with fresh food demand forecasting and offer practical tips and techniques for producing accurate demand forecasts. Whether you’re a grocer, retailer, or food industry professional, this guide will provide valuable insights into how to maximize profits and minimize waste by accurately predicting fresh food demand. So buckle up and get ready to dive into the fresh food forecasting challenge!


Accurate demand forecasting is crucial for success in the perishable products and fresh food industry due to the short shelf life and various external factors that can impact demand.

• Fresh food demand forecasting requires a specific approach that takes into account unique characteristics of perishable products, such as product appearance, hidden out-of-stocks, and fluctuating delivery times.

Key variables to consider when building a reliable fresh food demand forecasting model include seasonality, calendar of holidays and events, weather, pricing, product availability, consumer preferences and trends, marketing and advertising, competition, and transportation and storage requirements.

• Machine learning algorithms can help fresh food retailers and grocers optimize their operations, maximize profits, and reduce waste through accurate and timely demand forecasting and inventory management automation.

• Companies that implement AI solutions for demand forecasting and inventory management can benefit from reduced expenses, improved customer loyalty and engagement, and a competitive edge in the industry.

• nexocode AI experts have extensive experience in supply chain management and can help businesses in the retail industry implement AI solutions for accurate demand forecasting and inventory management. Contact us today to discuss your needs and find out how we can help.

The High-Stakes Game for Fresh Food Retailers and Groceries

demand prediction allows retailers to optimize their warehouse management and prevent both under and overstocking, but with fresh food, it is a totally new-level game. When the clock is ticking, inaccurate predictions can generate a significant financial loss. 

While other retailers would store the stock and try to gradually sell it out, for instance, using dynamic pricing techniques, food retailers cannot do so. They have to comply with various regulations which determine when the food can be sold and what to do with it afterward. Even if it seems perfectly fine, there is no chance of selling it when its shelf life comes to an end. 

If not sold on time, the food has to be handled in a specific and often strictly regulated manner, which, of course, implies additional costs. Thus, it’s far beyond a warehouse problem, like in the other retail niches. That’s why accurate predictions can be a game changer for businesses working with fresh food.

Why Fresh Food Needs a Specific Method for Forecasting Demand?

The forecasting methods used in retail predict demand based on historical and real-time data, taking various variables into account - from market and social media trends to even weather. Having learned what triggers changes in demand from the provided datasets, the models can predict them in the future. 

In the case of forecasting current and future demand for fresh food, the mechanism is similar. However, the pressure on accuracy is much higher since forecasting errors will have much more implications for the fresh food retailer than for any other business, forcing them to remove products from stock without a chance of selling and then dispose of them in the right way.

At the same time, there are various variables that impact the demand for fresh foods and their predictability, which are often not taken into account in other retail niches.

Product Appearance

Fresh food products are highly perishable, and their appearance can rapidly deteriorate. That, of course, may affect sales, as the customers will avoid choosing them, even if they know they are perfectly fine in terms of taste and quality. Once the product’s appearance deteriorates, the chances of selling are in a freefall. That’s why it’s crucial for fresh food retailers to forecast demand accurately to avoid overstocking and spoilage. They can actively prevent the deterioration of their products’ appearance by controlling the storage and shelf conditions and reducing food’s exposition to touch.

Hidden Out-of-Stocks

Another factor that enforces more accurate predictions is the possibility of having hidden out-of-stocks. Out-of-stock situations can occur even when inventory levels appear sufficient, as fresh food products may be damaged or unsellable due to factors like bruising, discoloration, or other issues. Forecasting enables retailers to prevent such situations, including a predicted percentage of out-of-stock fresh products in the estimations. 

Fluctuating Delivery Times from Farmers and Producers

Fresh food products are often sourced from local farmers or suppliers whose supply is subjected to internal and external parameters conditions. Weather and other unexpected events can lead to unpredictable delivery times and varying availability of particular products throughout the year. Whereas in other niches, they impact mainly the demand, in the case of fresh food, it’s the supply that gets affected. That’s why a forecasting model built for fresh food has to take the delivery times into account, drawing insights from historical data.

Lack of Data on How Products Look and Perish on Shelves

Fresh food can change its appearance in a relatively short time - every hour counts. Stocked in the morning, in the evening, it may already be in an entirely different shape. It doesn’t help that customers are often in direct contact with the products, packing and lifting them to compare the quality. 

That may speed up the deterioration of the appearance of the food while making it difficult to control how it looks on the shelves. Contrary to products with a longer shelf life, which are often tightly packed, fresh ones can lose their appeal easily if they are not organized in the right way. With computer vision paired with other AI solutions, retailers can solve this problem, but meanwhile, they need to include this factor in the estimations. 

Factors Affecting Fresh Food Demand Forecasting

We have already proved that fresh food needs a slightly different approach to forecasting than the majority of products, but which factors specifically affect its demand? Here are some key variables to include when building a reliable demand forecasting model.

factors affecting fresh food demand forecasting

factors affecting fresh food demand forecasting


Seasonality plays a key role in the supply predictions of fresh food. Just as the supply, the demand for fresh food varies depending on the season. Before globalization gained such a scale, the supply and demand would be more or less aligned. Fresh foods always taste the best when it’s their season, and due to their high supply, the prices can go down, so naturally, customers would opt for them. 

However, with globalization, they grew accustomed to having access to various types of fresh food all year. That affected the relevance of seasonality as a variable. Regardless, the habits still impact the preferences of the customers, and so does the quality of the seasonal food, so it should be taken into account when forecasting demand.

Calendar of Holidays and Events

This variable impacts sales in general, usually forming a classic pattern – just before the holidays and events, customers buy more than usual, and during them, they buy less. Thus, any demand forecasting model should include it. However in highly competitive retail industry, fresh food retailers get more impacted by it than any others due to the nature of their products. Besides adjusting their inventory and supply chain management to the sales peaks, they also have to think about the limited shelf life of the food. Precise historical sales data in this field is thus crucial for preventing financial loss.


Even though the farmers and suppliers can use preventive measures (like greenhouses, etc.) and get partially independent of the weather factors, droughts, frost, and plagues can leave retailers understocked. Some vegetables and fruits require specific conditions to grow, leaving them entirely dependent on weather circumstances. 

Relying on the weather records, the forecasting model can identify the fluctuations in weather conditions and the way they impacted the supply in the past. Natural disasters are difficult to predict - however, the predictive models can use recent data to detect more problematic events in the supplying regions, allowing retailers to make quick decisions to avoid supply shortages.


Pricing affects the demand, regardless of which product we’re referring to. In order to use this variable correctly, fresh food retailers should understand their target groups very well, since they may react to fluctuations in price in different ways. For instance, pricey imported avocados may sell out faster in a shop placed in a rich neighborhood that targets customers aiming for a healthy lifestyle than in a supermarket or convenience shop. 

The share of particular target groups should be included in the estimations to reflect the demand in the most accurate way. On the other hand, prices established by the suppliers may rise due to particular weather conditions, so the model should combine these variables to predict demand accurately.

Product Availability

Product availability is another variable dependent on weather conditions and supply chains, but there’s more complexity to it, particularly in the case of fresh food. The availability of particular foods is often tied together, and the predictive model should reflect it. It’s simple – certain fresh food are considered to go well together, so they are often purchased in the package. Take the yogurt, banana, and a fresh bun for breakfast or the salad and a sandwich for lunch. These duos or trios, of course, depend on the local customs – for instance, Polish customers will likely pair young potatoes with kefir or cottage cheese with spring onion and radish. Including this aspect in the estimations may radically improve their forecast accuracy too.

Trends have an enormous impact on demand, and the forecasting model should take that into consideration. They often boost the demand for certain foods to impressive levels, whether it’s avocado toast, shakshuka, or a basque cheesecake, to then suddenly drop, leading to overstocking. Fresh food retailers can multiply their margins if they observe these trends and adjust to them with their inventory management and marketing efforts. In order to stay competitive, they should use forecasting systems that include these. They can do that by integrating external data sources, most of which would be social media. 

Marketing and Advertising

Trends shape the demand externally, but the retailers themselves can also drive it and control it through thoughtful use of marketing. If they decide to promote particular products, the model should reflect that - otherwise, they could be sabotaging their own efforts and, thus – multiplying expenses. If the advertised products sell out in an instant, customers who come to buy them after seeing the ad may migrate to another provider, with lost revenues, especially if it happens repeatedly.

Short Shelf Life

As we have already mentioned, the short shelf life makes demand prediction much more challenging, leaving less margin for forecasting errors. The model should be taught about the shelf life of the particular foods to make the estimations as accurate as possible. It is worth taking into consideration that different aspects may impact the length of a product’s shelf life, including the conditions of the transport, the complexity of the supply chain, the weather, or even the distribution on the shop shelves (products that are beyond reach may quickly deteriorate in quality).


The presence of the competing fresh food retailers in the area, their target groups, marketing efforts, pricing, and product availability of their suppliers – all these competition-related aspects may determine the product demand, so any accurate forecasting model should include it.

Transportation and Storage Requirements

Fresh food often has special requirements when it comes to transport and storage. It’s usually much more susceptible to damage and may quickly lose its vivid color and properties if not stored correctly. Some products, like eggs, may easily break, while others melt when not stored at the right temperature. The forecasting model should include this aspect in the estimations, predicting the likelihood of the food getting spoiled and suggesting a supply that the retailers can actually cope with given the required conditions.

AI-based replenishment software with demand forecasting options that automatically trigger restocking orders for medical products. Just like fresh food, some pharmaceuticals have short shelf life and need accurate forecasting solutions to minimize waste.

AI-based replenishment software with demand forecasting options that automatically trigger restocking orders for medical products. Just like fresh food, some pharmaceuticals have short shelf life and need accurate forecasting solutions to minimize waste.

Related case study: Optimizing drug distribution and inventory activities for a hospital pharmacies network

To improve current large-scale procurement processes, a pharma company approached us to use applied analytics to stock and distribute drugs among US hospitals.

Our challenge? Maximizing savings by streamlining the procurement of medication across the hospital network and their pharmacies. Read more about this case study.

The Machine Learning Advantage: Enhancing Fresh Food Demand Forecasting Accuracy

The demand gets impacted by numerous factors that often interfere with each other and depend on one another. In the case of rule-based programming, we would have to know and understand them all in order to establish the rules correctly and come up with accurate estimations. In machine learning, it is the algorithm itself that finds the correlations between the different variables and demand, improving itself with every feedback. 

You have seen how long is the list of variables that should be taken into account in fresh food demand forecasting.  Traditional methods make it harder to improve estimations over time and add tons of manual work, as you need to redefine the rules every once in a while.

How predictive models for demand forecasting work

How predictive models for demand forecasting work

Benefits of ML-Based Inventory Forecasting for Fresh and Ultra-Fresh Products

Using machine learning forecasting algorithms, fresh food retailers and grocers can keep up with customers’ expectations while optimizing their operations and maximizing profit. Having just the right amount of fresh food in stock, companies are capable of building customer loyalty and engagement. Accurate and timely forecast prevents them from generating excessive waste, which is very common for fresh food prone to quality deterioration, and facilitates optimal inventory allocation decisions. 

Machine learning forecasting models help enterprises tackle the challenges of short shelf life, supporting the daily work of inventory managers and purchasing directors. If they waste less supply, they save money on products that will never sell, their storage, and management. 

Limiting waste also works in favor of the food companies’ image. In recent years, the problem of food waste has become an object of public debate. New regulations appeared to encourage businesses to reduce the amount of produced waste. Today, a big percentage of customers put importance on that aspects when choosing between food suppliers. That means accurate forecasting could give food companies a competitive edge.

Machine learning forecasting algorithms also enable automation in replenishing and managing perishable inventory. With the ability to predict demand with greater accuracy, retailers can set up automated systems to restock their inventory, reducing the need for manual intervention and freeing up valuable time for their teams to focus on other tasks.

In addition, machine learning can be used to identify and track key performance indicators (KPIs) related to food sales, such as turnover rates and sell-through rates. This data can then be used to optimize promotions planning and determine the best times to launch new products, maximizing sales and revenue.

The Bottom Line: Accurate Demand Forecasting is Essential for Perishable Product Success

Predictive analytics is becoming increasingly popular among retailers as a way to optimize business processes while reducing expenses, and food retailers are starting to follow. If you would like to gain access to all the benefits listed above, we can help you create a model that takes all your needs and the specifics of your business into account. Reach out to us if you would like to discuss the details!

About the author

Dorota Owczarek

Dorota Owczarek

AI Product Lead & Design Thinking Facilitator

Linkedin profile Twitter

With over ten years of professional experience in designing and developing software, Dorota is quick to recognize the best ways to serve users and stakeholders by shaping strategies and ensuring their execution by working closely with engineering and design teams.
She acts as a Product Leader, covering the ongoing AI agile development processes and operationalizing AI throughout the business.

Would you like to discuss AI opportunities in your business?

Let us know and Dorota will arrange a call with our experts.

Dorota Owczarek
Dorota Owczarek
AI Product Lead

Thanks for the message!

We'll do our best to get back to you
as soon as possible.

This article is a part of

AI in Logistics
51 articles

AI in Logistics

Artificial Intelligence is becoming an essential element of Logistics and Supply Chain Management, where it offers many benefits to companies willing to adopt emerging technologies. AI can change how companies operate by providing applications that streamline planning, procurement, manufacturing, warehousing, distribution, transportation, and sales.

Follow our article series to find out the applications of AI in logistics and how this tech benefits the whole supply chain operations.

check it out


Insights on practical AI applications just one click away

Sign up for our newsletter and don't miss out on the latest insights, trends and innovations from this sector.


Thanks for joining the newsletter

Check your inbox for the confirmation email & enjoy the read!

This site uses cookies for analytical purposes.

Accept Privacy Policy

In the interests of your safety and to implement the principle of lawful, reliable and transparent processing of your personal data when using our services, we developed this document called the Privacy Policy. This document regulates the processing and protection of Users’ personal data in connection with their use of the Website and has been prepared by Nexocode.

To ensure the protection of Users' personal data, Nexocode applies appropriate organizational and technical solutions to prevent privacy breaches. Nexocode implements measures to ensure security at the level which ensures compliance with applicable Polish and European laws such as:

  1. Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing Directive 95/46/EC (General Data Protection Regulation) (published in the Official Journal of the European Union L 119, p 1); Act of 10 May 2018 on personal data protection (published in the Journal of Laws of 2018, item 1000);
  2. Act of 18 July 2002 on providing services by electronic means;
  3. Telecommunications Law of 16 July 2004.

The Website is secured by the SSL protocol, which provides secure data transmission on the Internet.

1. Definitions

  1. User – a person that uses the Website, i.e. a natural person with full legal capacity, a legal person, or an organizational unit which is not a legal person to which specific provisions grant legal capacity.
  2. Nexocode – NEXOCODE sp. z o.o. with its registered office in Kraków, ul. Wadowicka 7, 30-347 Kraków, entered into the Register of Entrepreneurs of the National Court Register kept by the District Court for Kraków-Śródmieście in Kraków, 11th Commercial Department of the National Court Register, under the KRS number: 0000686992, NIP: 6762533324.
  3. Website – website run by Nexocode, at the URL: whose content is available to authorized persons.
  4. Cookies – small files saved by the server on the User's computer, which the server can read when when the website is accessed from the computer.
  5. SSL protocol – a special standard for transmitting data on the Internet which unlike ordinary methods of data transmission encrypts data transmission.
  6. System log – the information that the User's computer transmits to the server which may contain various data (e.g. the user’s IP number), allowing to determine the approximate location where the connection came from.
  7. IP address – individual number which is usually assigned to every computer connected to the Internet. The IP number can be permanently associated with the computer (static) or assigned to a given connection (dynamic).
  8. GDPR – Regulation 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of individuals regarding the processing of personal data and onthe free transmission of such data, repealing Directive 95/46 / EC (General Data Protection Regulation).
  9. Personal data – information about an identified or identifiable natural person ("data subject"). An identifiable natural person is a person who can be directly or indirectly identified, in particular on the basis of identifiers such as name, identification number, location data, online identifiers or one or more specific factors determining the physical, physiological, genetic, mental, economic, cultural or social identity of a natural person.
  10. Processing – any operations performed on personal data, such as collecting, recording, storing, developing, modifying, sharing, and deleting, especially when performed in IT systems.

2. Cookies

The Website is secured by the SSL protocol, which provides secure data transmission on the Internet. The Website, in accordance with art. 173 of the Telecommunications Act of 16 July 2004 of the Republic of Poland, uses Cookies, i.e. data, in particular text files, stored on the User's end device.
Cookies are used to:

  1. improve user experience and facilitate navigation on the site;
  2. help to identify returning Users who access the website using the device on which Cookies were saved;
  3. creating statistics which help to understand how the Users use websites, which allows to improve their structure and content;
  4. adjusting the content of the Website pages to specific User’s preferences and optimizing the websites website experience to the each User's individual needs.

Cookies usually contain the name of the website from which they originate, their storage time on the end device and a unique number. On our Website, we use the following types of Cookies:

  • "Session" – cookie files stored on the User's end device until the Uses logs out, leaves the website or turns off the web browser;
  • "Persistent" – cookie files stored on the User's end device for the time specified in the Cookie file parameters or until they are deleted by the User;
  • "Performance" – cookies used specifically for gathering data on how visitors use a website to measure the performance of a website;
  • "Strictly necessary" – essential for browsing the website and using its features, such as accessing secure areas of the site;
  • "Functional" – cookies enabling remembering the settings selected by the User and personalizing the User interface;
  • "First-party" – cookies stored by the Website;
  • "Third-party" – cookies derived from a website other than the Website;
  • "Facebook cookies" – You should read Facebook cookies policy:
  • "Other Google cookies" – Refer to Google cookie policy:

3. How System Logs work on the Website

User's activity on the Website, including the User’s Personal Data, is recorded in System Logs. The information collected in the Logs is processed primarily for purposes related to the provision of services, i.e. for the purposes of:

  • analytics – to improve the quality of services provided by us as part of the Website and adapt its functionalities to the needs of the Users. The legal basis for processing in this case is the legitimate interest of Nexocode consisting in analyzing Users' activities and their preferences;
  • fraud detection, identification and countering threats to stability and correct operation of the Website.

4. Cookie mechanism on the Website

Our site uses basic cookies that facilitate the use of its resources. Cookies contain useful information and are stored on the User's computer – our server can read them when connecting to this computer again. Most web browsers allow cookies to be stored on the User's end device by default. Each User can change their Cookie settings in the web browser settings menu: Google ChromeOpen the menu (click the three-dot icon in the upper right corner), Settings > Advanced. In the "Privacy and security" section, click the Content Settings button. In the "Cookies and site date" section you can change the following Cookie settings:

  • Deleting cookies,
  • Blocking cookies by default,
  • Default permission for cookies,
  • Saving Cookies and website data by default and clearing them when the browser is closed,
  • Specifying exceptions for Cookies for specific websites or domains

Internet Explorer 6.0 and 7.0
From the browser menu (upper right corner): Tools > Internet Options > Privacy, click the Sites button. Use the slider to set the desired level, confirm the change with the OK button.

Mozilla Firefox
browser menu: Tools > Options > Privacy and security. Activate the “Custom” field. From there, you can check a relevant field to decide whether or not to accept cookies.

Open the browser’s settings menu: Go to the Advanced section > Site Settings > Cookies and site data. From there, adjust the setting: Allow sites to save and read cookie data

In the Safari drop-down menu, select Preferences and click the Security icon.From there, select the desired security level in the "Accept cookies" area.

Disabling Cookies in your browser does not deprive you of access to the resources of the Website. Web browsers, by default, allow storing Cookies on the User's end device. Website Users can freely adjust cookie settings. The web browser allows you to delete cookies. It is also possible to automatically block cookies. Detailed information on this subject is provided in the help or documentation of the specific web browser used by the User. The User can decide not to receive Cookies by changing browser settings. However, disabling Cookies necessary for authentication, security or remembering User preferences may impact user experience, or even make the Website unusable.

5. Additional information

External links may be placed on the Website enabling Users to directly reach other website. Also, while using the Website, cookies may also be placed on the User’s device from other entities, in particular from third parties such as Google, in order to enable the use the functionalities of the Website integrated with these third parties. Each of such providers sets out the rules for the use of cookies in their privacy policy, so for security reasons we recommend that you read the privacy policy document before using these pages. We reserve the right to change this privacy policy at any time by publishing an updated version on our Website. After making the change, the privacy policy will be published on the page with a new date. For more information on the conditions of providing services, in particular the rules of using the Website, contracting, as well as the conditions of accessing content and using the Website, please refer to the the Website’s Terms and Conditions.

Nexocode Team