The Next Phase of Grocery AI is Predicting Waste Before it Happens

Food waste sits at the intersection of some of the biggest challenges facing retailers today. As grocers face tighter margins and growing pressure to improve sustainability, how they manage waste in-store has become one of the primary levers they can pull.

The scale of the issue is hard to ignore. Global food waste costs have risen the past three years, and today it costs over $540 billion a year.

It's not an easy challenge to solve because it's rarely the result of a single decision. Stores often have thoughtful forecasts and head office plans, but in-store execution is very difficult to keep consistent. Store associates are juggling a huge number of competing priorities, and addressing waste often gets deprioritized in favor of a lever that feels more urgent in the moment, such as product replenishment. It doesn't help that the markdown process itself has historically been manual and slow, requiring associates to walk the aisles and audit expiry dates by hand.

Retail Insight works across nearly 70,000 stores globally. What we've seen firsthand is that solving food waste requires better timing and better signals. It needs to be underpinned by modern technology that puts retailers on the front foot, turning food waste from a cost center into a genuine opportunity.

Waste Starts Earlier Than Most Retailers Think

There are endless examples of stores having too much inventory when the markdown process should have started days earlier. Take a fresh NPD promotion the retailer invested heavily in that simply undersold. The retailer is left with a surplus of stock on-hand, but because nobody flags it until the manual check the next day, discounting doesn't start until then either, a full day later than it should. As a one-off, this may seem insignificant, but retail is a game of big numbers. These minor issues, repeated across thousands of SKUs and stores, add up to a meaningful hit to waste and margin by year end.

This is why the traditional approach to managing waste is too reactive and too rigid. Associates typically identify products nearing expiration the night before the final day they can be sold, by which point most of the margin opportunity is already gone, particularly on lines that were already overstocked or underselling. Connected store technology changes the timeline. Instead of waiting for a manual check the night before expiry, retailers can flag at-risk stock as soon as sales and inventory signals suggest a product will not sell through in time, and adjust price accordingly, days earlier than the manual process would have caught it.

From Reactive Decisions to Dynamic Markdowns

One of the clearest examples of artificial intelligence already creating measurable value in grocery retail today is dynamic markdown optimization.

Traditionally, markdown decisions have relied on fixed schedules, manual reviews, and static discounting rules, with associates setting markdown levels with limited visibility into demand patterns, inventory levels, or likely sell-through.

Modern technology changes that calculus. By drawing on hundreds of signals, including store demographics and time of year, dynamic markdown models determine not only which products need reducing, but how aggressive the discount needs to be to drive maximum sell-through without sacrificing margin. This approach has been shown to increase profit on fresh waste by more than 10 percent. Rather than treating every product the same, retailers can base each markdown decision on how that specific item is actually performing in that specific store.

Retailers then pair those insights with clear, in-app prompts that tell associates when to prioritize a markdown and by how much. However, the final call stays with the person on the floor, who can see things the model cannot, such as a damaged product.

The most advanced retailers are going further still, connecting their markdown process to electronic shelf-edge labels and automating it end to end. That increases the reliability with which markdowns happen, improves profit, reduces waste, and cuts the in-store labor needed to run the process manually.

The Move Toward Predictive Waste

While dynamic markdown technology has helped retailers respond far more effectively once a product has become a waste risk, there has traditionally been a gap around exactly when that markdown should be applied.

Most retailers run on fixed markdown schedules, applying the same number of markdowns, one-touch or two-touch, to the same categories at the same point before expiry, regardless of actual selling conditions.

New machine learning techniques, combined with more granular enterprise data, now make it possible to identify which products are likely to become waste well before they reach that point. Store associates can receive predictive markdown prompts up to a week out from expiration, catching sell-through risk early enough to avoid the overly aggressive, last-minute discounting the traditional process forces.

The emerging approach to predictive waste looks beyond inventory and delivery timelines to a wider set of forward-looking signals. We're working with several retailers on which data points genuinely move the needle on food waste and weaving them into our own markdown technology, including weather conditions, regional and local demand patterns, and customer basket data.

The end goal is a supply chain feedback loop where sell-through signals feed back into buying decisions and teams can see assortment optimization happening in near real time, driven directly from the shelf.

This is a rapidly evolving space. As RFID and 2D barcodes become more widely integrated, retailers will be able to track individual items by unique identifier at scale, sharpening predictive waste further and giving teams a degree of item-level visibility the industry has not had before.

The Next Evolution of Retail Operations

The technology now exists to manage the full lifecycle of fresh products with a precision that was not possible even a few years ago, helping retailers answer four questions in close to real time:

• Which item needs a markdown?
• When should that markdown begin?
• What discount will maximize sell-through without giving away unnecessary margin?
• How do we prevent the same waste next time?

The next competitive advantage in grocery belongs to whomever moves from managing waste to preventing it before it starts.

Tom Coe is vice president of growth at Retail Insight, a provider of advanced retail optimization software.