Can AI and Blockchain Solve the Food Loss and Waste Problem?

The Atlantic featured an interesting article on how to reduce food waste by hacking expiration dates through the use of AI. The article explains how chemists started tackling the problem of oxidative rancidity by using AI to suggest faster and more efficient means of determining effective antioxidant combinations. 

This, of course, is one among many articles that is meant to say: “Hey kids, look! AI can solve X.” 

So today, we will take a look at why the expiration date is such a big deal, and worthy of the use of AI and why, in my opinion, AI won’t solve (alone) the bigger problem of Food Loss and Waste. 

Food Loss and Waste

Food Loss and Waste (FLW) is a serious issue which affects every part of the global food supply chain. The problem isn't just the extra food being discarded, it's the financial, environmental, and societal impact it has. To better illustrate what I mean, McKinsey estimates that around $600 billion worth of food is lost globally during or right after harvest. This loss is not limited to the fields but extends to our grocery stores and households. In the United States alone, about 30% of surplus food in grocery stores, which amounts to a loss of $16 billion annually, is discarded. This pattern of waste translates to Americans throwing away approximately 80 billion pounds of food each year —a value of about $161 billion. The waste by American consumers alone totals to around $218 billion annually.

However, FLW isn't solely about the monetary value of wasted food, it's also about the missed opportunities to feed the hungry. Astonishingly, while an estimated 30% to 40% of the world’s food is lost or wasted, roughly 11% of the world’s population struggles with hunger. This stark contrast highlights the inefficiency and inequity within our food systems.

Understanding FLW requires differentiating between 'food loss' and 'food waste.' The FAO defines food waste as the discard of edible foods at retail and consumer levels, and food loss as the decrease in food mass at the production, post-harvest, and processing stages. The contribution to FLW is widespread, with consumers, farms, manufacturers, food services, and retailers all playing a part. In the American context, while consumers are responsible for wasting as much as 37% of food, farms contribute 21%, manufacturing 14%, food service 16%, and food retail 12%.

The McKinsey report sheds light on the different rates of FLW depending on the product and its distribution in the supply chain. For example, in the case of field-grown tomatoes, out of 100 tomatoes ready for harvest, only 59 to 72 make it to the retailer, with losses occurring at every stage from harvesting to retail.

Before we look at the suggested solutions and the role of technology, let’s understand why the problem is more complex than it seems. 

Why is the Food Supply Chain Complex?

Former students of mine may wonder why this isn't a newsvendor problem (or any other version of a simple inventory problem), and the answer is that even from a theoretical standpoint, finding an optimal policy is challenging due to several inherent complexities. Perishable inventory management often involves a multi-tiered supply chain, from suppliers to warehouses to retail outlets. Optimizing inventory across these multiple stages, each with its own perishability and demand dynamics, adds to the complexity.

Complex Dynamics of Perishability: The perishability of products introduces a temporal dimension into inventory models. Unlike durable goods, perishable goods have a finite lifespan, after which they become unsellable or lose significant value. This adds a layer of complexity in determining how much and when to order, as the utility of the inventory decreases over time.

Unpredictable Demand and Supply Patterns: Perishable goods exhibit highly variable and unpredictable demand and supply patterns. Seasonality, trends, and random fluctuations in demand are common, and make it difficult to make accurate predictions. I will add that demand may also depend on how food looks —demand for bananas, for instance, has a peak between the moment bananas are ripe and until they become too ripe. Similarly, supply can be uncertain due to factors such as agricultural yields, transportation issues, or supplier reliability. Incorporating these uncertainties into a theoretical model is challenging. The randomness in demand and supply, along with the stochastic nature of product life and quality deterioration, makes it difficult to create a deterministic model. Theoretical models often rely on stochastic processes and probabilistic approaches, which are inherently more complex.

Incorporating Shelf-Life into Decision Making: Traditional inventory models primarily focus on balancing ordering costs with holding costs (remember the EOQ formula?). However, for perishables, the shelf life imposes an additional constraint, where holding inventory too long leads to spoilage. The theoretical model must, therefore, integrate shelf-life constraints into the decision-making process.

Cost Structures: The cost structures in perishable inventory management are often non-linear. Spoilage costs, for instance, do not accrue linearly over time but may increase sharply as products approach their expiration dates. Similarly, the cost of lost sales due to stockouts can be significant and non-linear, especially if customers switch to competitors. Theoretical models must balance various costs, including ordering costs, holding costs, spoilage costs, and costs associated with lost sales due to stockouts. Finding an optimal balance that minimizes the total cost is a complex task, especially when these costs are interdependent and affected by perishability.

Data Requirements and Model Complexity: Accurate perishable inventory management requires detailed data regarding demand patterns, shelf life, spoilage rates, etc. The complexity of models that can handle this level of detail is significantly higher, making it challenging to find and implement optimal policies.

In summary, the theoretical difficulty in finding an optimal perishable inventory policy stems from the need to simultaneously consider multiple, often conflicting factors such as the ones mentioned above. These factors make the models more complex and computationally challenging.

Can Technology and AI Save Us?

As the article from the Atlantic mentions, initiatives to mitigate FLW are emerging, reflecting an increased awareness. These range from innovative approaches like vertical farming, which shortens the supply chain for perishable produce, to companies like Misfits Market in Philadelphia, which redistributes misshaped or near-expiration produce to lower-income consumers at reduced prices. 

Let’s take a closer look at some of these initiatives:

Leveraging AI for Better Forecasting

Artificial Intelligence is like a ray of hope for changing how the food supply chain works, mainly by helping predictions become more accurate. AI excels in situations where you have a vast amount of data and non-linear patterns, so by analyzing historical sales, weather patterns, consumer behavior, and market trends, AI can provide better predictions for demand. This ensures that producers and retailers produce and stock precisely what is needed, reducing surplus that often leads to waste. For instance, AI algorithms can predict the ebb and flow of demand for perishable goods, allowing grocery stores to order optimal quantities and reduce the likelihood of spoilage on shelves. While the inventory problem remains hard, any problem is easier with better predictions. 

Extending Expiration Dates

This is the point analyzed in the article. For example, the undesirable flavor from a bag of stale nuts is the result of oxidation, a natural process exacerbated by exposure to air, heat, or UV light, which particularly affects lipids —fats and oils commonly found in many pantry staples.

Antioxidants, both natural and synthetic, are the food industry’s primary defense against this process as it neutralizes oxidative rancidity and extends the shelf life of food products. However, finding the optimal mix of antioxidants is a complex and costly endeavor, which involves extensive experimentation to avoid antagonism —incorrect combinations that diminish their protective effect. This is exactly where AI can be extremely helpful. 

But technology can also prove useful through better preservation technologies and packaging methods, which can also extend food products’ shelf life. This not only simplifies inventory management but also provides consumers with a longer window to consume the product, and ultimately, decreases the amount of food discarded. Innovations such as modified atmosphere packaging and smarter packaging that includes time-temperature indicators can help keep food fresh for longer periods and provide more accurate information on food safety.

Improving Data for Enhanced Traceability

AI algorithms need data, and better data can lead to superior traceability in the food supply chain, allowing for more precise tracking of food items from farm to fork. With enhanced traceability, stakeholders can quickly identify and address inefficiencies that lead to waste. For example, if a batch of produce is found to be spoiling prematurely, data can pinpoint whether the issue occurred at the farm level, during transport, or at the retail stage. 

The big picture is that incorporating food waste technology into the supply chain is costly, with the current alternative—landfills—being much cheaper. The siloed nature of the supply chain poses additional challenges in scaling new technologies. Regulations, both federal and local, help and accelerate the adoption of these innovations. 

The Role of Regulation 

State and local policies, such as those in California and New York, which require food waste separation and management, play a crucial role in driving innovation in the supply chain. 

At the national level, the Food and Drug Administration’s Food Traceability rule mandates companies to maintain and provide detailed records of food products, which is expected to drive technological adoption. This facilitates the FDA’s ability to swiftly pinpoint and extract potentially contaminated food from the market when aiming to reduce the incidence of foodborne illnesses and deaths. The regulation reflects the interconnected nature of the food supply chain, emphasizing the necessity for collective action to ensure successful implementation and to bolster the overall safety and reliability of the food system. 

Moreover, data-driven insights can inform better resource allocation, reduce redundancies, and ensure that food reaches its intended destination in the best condition possible. It also enables more effective recalls and reduces the risk of contamination, which can lead to widespread waste. In addition, the data can be used to connect excess food with potential users. For example, retail outlets can partner with food banks and charities to donate food that is nearing its sell-by date but is still safe to consume, ensuring that it feeds people rather than landfills.

As I mentioned, not all regulations are Federal. Policies like California’s food waste separation law and New York’s bill for food donation and recycling are examples of how legislation can stimulate progress. ​​Incentivizing technological scaling in this siloed industry is complex, and regulatory frameworks play a significant role. 

More Systemic Issues

While advancements in technology and regulatory measures are pivotal in combating food waste, they may not suffice in addressing the more systemic behavioral issues associated with food waste, both on the consumer side and the supply chain. 

For example, consumers often prefer aesthetically pleasing produce, leading to the rejection of perfectly edible but “ugly” items. This preference contributes significantly to waste within the food supply chain.

The “ugly produce” problem highlights the behavioral tendencies that result in the discard of perfectly nutritious, albeit cosmetically imperfect, produce. To address this, the concept of an imperfect food supply chain has emerged, offering a market for these goods that do not meet conventional beauty standards but are otherwise consumable. This initiative reduces waste and provides consumers with more affordable options.

Beyond consumer behavior, a more severe issue lies within supply chain coordination. A research paper written by Elizabeth Paulson, and her advisors at MIT, Retsef Levi and Georgia Perakis, sheds light on the complications that arise from supply uncertainty in produce supply chains​​. Uncertainty regarding demand, but also yield. The core problem is the frequent disruptions in delivery times, quantities, and quality due to supply shortages, which is a significant challenge for retailers dealing with perishable goods. 

The difficulty lies in retailers’ dual-sourcing strategy, where they source goods from two different suppliers to mitigate risk. However, without transparency between retailers and suppliers, the effectiveness of this strategy is compromised. Retailers, acting under the impression of scarcity, may over-order to ensure adequate supply, leading to excess inventory and increased waste —particularly problematic for perishable items that cannot be stored indefinitely.

Another layer of complexity is the competitive environment among retailers, which often leads to a lack of information sharing and collaboration. This competitive mindset fuels a cycle of over-ordering and waste as each retailer independently strives to secure sufficient stock.

The study advocates for a reverse information-sharing approach, where suppliers share inventory and order information downstream with retailers. This exchange of information helps to counteract the over-ordering behavior stemming from perceived scarcity, ultimately reducing costs and waste across the supply chain​​. By sharing information, retailers have a clearer picture of available supply and can make more informed ordering decisions.

How can this information be shared credibly? 

Blockchain! 

I’ve written more than 100 newsletter articles and managed not to use this term once. But finally, blockchain is the solution to something.

Now, of course, implementing reverse information sharing comes with its own set of challenges: Supplier resistance, system integration challenges, concerns about data privacy and security, and the need to align incentives are all significant barriers to the implementation of reverse information sharing in supply chains. Suppliers may resist in fear that it could weaken their bargaining position or expose their operations to competitors. Additionally, developing and integrating systems capable of handling such information is a complex and costly endeavor, requiring substantial investment and ongoing maintenance. The increased data sharing also raises serious concerns about data privacy and security, necessitating stringent protective measures. Finally, the success of reverse information sharing depends on aligning the incentives of all parties involved. While actions to increase supply chain surplus exist in theory, there is often a reluctance to adopt them in the short term.

Conclusion

The cost of integrating food waste technology into the FSC might be high, but the return on investment becomes evident when considering the reduction in waste and its associated costs. Better technology, both when it comes to traceability as well as making better predictions and elongating the expiration dates of products, can increase efficiency, reduce waste, and save money. 

But while the allure of artificial intelligence as a panacea for the complex issues of the food supply chain is compelling, it’s clear that AI alone cannot resolve all the issues. Behavioral nuances, competitive dynamics, and the need for transparency and collaboration present hurdles that technology, in isolation, cannot overcome. However, AI’s ability to process vast amounts of data and provide actionable insights can undoubtedly assist in tackling these systemic issues. It can facilitate better forecasting, encourage more sustainable consumer behaviors, and support the implementation of strategies like reverse information sharing. Ultimately, while AI is not a standalone solution, it is an invaluable tool in the multifaceted effort to reduce food waste and enhance the efficiency and resilience of the food supply chain.

And I’m sure humankind will benefit in many other ways from AI, before it takes over the world that is…