A costly retail problem is becoming a data problem
Fresh food is where grocery stores lose some of their hardest money. Managers have to guess how many strawberries, avocados, cuts of meat, or prepared meals to stock before demand fades and spoilage takes over. Unlike packaged goods, fresh inventory is highly perishable, inconsistently measured, and often poorly tracked once it leaves the back room for the shelf.
Startup Afresh is betting that better forecasting can chip away at that waste. The company has raised $34 million in new funding, co-led by Just Climate and High Sage Ventures, and says its AI tools are already helping retailers cut shrink by 20% to 25% in fresh categories.
The funding round, reported by Fast Company, matters because grocery waste is not a niche inefficiency. The article estimates that U.S. grocery stores waste around four million tons of food annually, with a roughly $27 billion cost. That makes fresh inventory planning one of the more consequential and under-digitized operational problems in retail.
From spreadsheets and guesswork to demand modeling
Afresh’s origin story is almost a study in how analog many food-retail workflows remained until recently. When co-founders Matt Schwartz and Nathan Fenner began studying the problem, they found produce managers relying on printed spreadsheets, rough estimates, and pen-and-paper ordering processes.
That made some sense historically. Fresh foods are much harder to manage than shelf-stable products. Produce sold by weight can lose mass through evaporation. Self-checkout errors can distort what was actually purchased. Spoiled items may be thrown away without being properly recorded. Promotions, temperature, and shipping origin can all affect how quickly a product degrades.
Afresh’s software tries to pull those variables into a forecasting system. According to the source text, the company analyzes data from each grocer, in some cases drawing on hundreds of billions of transactions. Its models take into account pricing, promotions, shipment origin, weather, and even timing related to food-stamp distribution. Demand forecasts are then paired with optimization tools that suggest order quantities for each product.
The premise is simple: if stores can predict demand and perishability more accurately, they can order closer to what will actually sell.
Why fresh categories are different
Retail technology often looks mature from the outside, but fresh departments have remained stubbornly resistant to clean automation. Packaged foods arrive with standardized units, predictable shelf life, and digital supply chain records. Fresh items are noisier. A box of raspberries and a tray of salmon do not behave like cereal or toothpaste.
That is why AI has appeal here. It can absorb more variables than a store manager can juggle manually and can continue learning as new data arrive. The company says those models improve over time, which is especially valuable in a domain where local conditions matter enormously. A neighborhood’s demand patterns, weather shifts, and shopper habits can change what “correct” inventory looks like from week to week.
Afresh reportedly starts with trials in 10 to 20 stores and compares results with a control group operating during the same period. Schwartz said the company typically sees 20% to 25% reductions in shrink when its system goes live.
If those reductions hold at scale, the business case is clear. Even modest improvements in waste rates can produce significant savings when margins are thin and food categories turn quickly.
Operational changes beyond ordering
The technology’s impact is not limited to purchase orders. According to the article, grocers can also use Afresh’s data to redesign displays and improve how they handle items nearing spoilage. In some stores, the software has flagged produce displays that are larger than necessary, allowing managers to resize them or use dummy displays to preserve the appearance of abundance with less actual fruit on hand.
That may sound cosmetic, but display strategy is operationally important. Grocery stores often overstock visible produce because fullness signals freshness and abundance to shoppers. If software can maintain that perception with less physical inventory, it reduces waste without sacrificing merchandising.
The same logic extends to repurposing food. Stores can turn produce approaching the end of its shelf life into prepared products, such as avocados becoming guacamole. Afresh has also launched a separate tool to forecast demand in deli prepared foods, another category where spoilage and forecasting errors can be expensive.
Why the funding matters
AI in retail is often discussed through flashy consumer-facing tools, but some of the more durable uses may be in back-end operating decisions. Fresh food waste is economically painful, environmentally costly, and difficult to solve through labor alone. It is exactly the kind of planning problem where better predictions can compound into measurable gains.
Afresh says its system is now used in more than 12,500 grocery store departments nationwide, including at Safeway and Albertsons. That footprint suggests the company has moved beyond pilot-stage curiosity and into broad operational testing.
The new $34 million round should help it expand further, but the larger significance is sectoral. Grocery waste is becoming legible as a software problem rather than an unavoidable cost of doing business. If that reframing succeeds, it could influence how retailers invest in inventory systems, store operations, and sustainability efforts over the next decade.
For consumers, the change may be invisible. Shelves will still look full, and stores will still restock overnight. But beneath that routine, a growing share of the decision-making may come from systems built to answer a basic but surprisingly hard question: how much fresh food will people actually buy before it goes bad?
This article is based on reporting by Fast Company. Read the original article.
Originally published on fastcompany.com
