Forecasting Fresh: Why Every Grocery Store Needs AI


Avocados have become a staple for many shoppers’ grocery lists over the past few years. With millions of recipes and health tips focused on the avocado, it’s no wonder sales have more than doubled since 2010! And for the produce teams that run fresh departments, keeping this crucial fruit in stock is no easy task.

Holy guacamole! Forecasting the right amount of avocados is tough

Avocados present a plethora of variables and challenges that simply don’t exist in the cereal aisle, especially when it comes to freshness and quality. Customers want a variety to choose from. They might need a small one for lunch today and three large ones for a party Friday night. That means stores need to build a perfect display of ripe, unripe, and soon-to-be-ripe avocados. And yet, most grocery shoppers in the United States are left having to choose from full displays of rock-hard avocados or piles of mushy, near-guacamole. Why? Because supermarkets simply aren’t using the right technology.

Demand planning for cereal is much easier, but even that goes wrong

Cereal management runs on safety stocks and perpetual inventory. When the inventory count dips below its safety stock, the mainframe reorders a set amount. Even though the math behind ordering cereal is fairly trivial, the data is nearly impossible to get right. Today’s technology solutions determine a safety stock level based on historical demand and the system does not recompute its safety stock daily because it doesn’t need to. 

In general, ordering cereal is easy, but stockouts still happen. And since that’s the case, the tech built for center store (aka shelf-stable foods) definitely won’t work for fresh departments. What store teams need is artificial intelligence that’s built for the dynamics of the department it serves.

Grocers need tech that puts fresh first instead of on the backburner

So when it comes to ordering avocados, the ideal scenario would be to know exactly how much of each type your customers want to buy, and then to stock those exactly. But the factors that influence who buys avocados, how much they buy, and how ripe they want them are almost endless. 

Consumer demand for avocados is highly influenced by price, promotions, weather, holidays, the price of tomatoes, the price of organic avocados vs. conventional, and seasonality. External factors have a big impact too, like ever-changing food trends and how COVID-19 has dramatically changed when and where people shop for groceries in the U.S. 

Together, these factors beg for highly adaptable demand forecasting that’s responsive to changes in consumer demand, rather than a “set it and forget it” system. What’s the solution? Easy-to-use technology that leverages artificial intelligence (AI) to streamline those complex operations inherent to fresh.

Leveraging AI makes placing perfect orders easy

Afresh’s AI solves the complex problem of ordering avocados with state-of-the-art neural networks that take into account dozens of fresh variables for each forecast in a way no other existing solution does. Our networks predict demand trajectories over long ordering horizons to forecast how demand will evolve over the next few orders. No matter what, a perfect point demand forecast is impossible for a grocery store to achieve. Because all forecasting is built from historical data, there will always be factors that cannot be captured by AI. So at Afresh, we train our models to predict distributions of demand rather than on point forecasts. We rigorously test, validate, and monitor for forecasting quality to ensure fresh teams can build truly the best orders possible for every item in every store, every day of the year. 

Take a deep dive into fresh-focused AI in our whitepaper: Putting AI in Grocery Aisles

Uncaptured events are inevitable, making predictions tricky to get right. Maybe the local high school is having a tailgate or the town newspaper published a new recipe for fruit salad. But through rigorous validation of data, sophisticated technology, and an approach that’s focused on fresh, store teams can rest easy knowing Afresh’s AI refines itself constantly to provide the most accurate recommendations possible. (And that’s led to 97% adherence to those order suggestions across all our partners!) 

By crafting beautiful, bountiful displays of a shopping list staple, grocery retailers can double their sales, gain deep customer loyalty, and represent the pinnacle of quality in fresh among competitors. It’s one of the most valuable puzzles a produce manager can solve. Equipped with highly accurate probabilistic forecasts, grocery retailers can now tackle the real problem: How do we build that perfect shelf, day after day?

Ready to make fresh your biggest competitIve edge? Request a demo!