The forecasting model worked. The food bank was not in it.

Grocery retailers have adopted demand-forecasting software widely enough that the buffer stock they used to donate is thinning, and food banks say they are now purchasing food that arrived free a few years ago ^[1][9][11]. This is the cleanest current example of an optimization that performed exactly as sold and pushed a cost onto a party who never appeared in the business case.

The tooling is not speculative. Afresh says it is in use in more than 12,000 grocery departments and has prevented more than 200 million pounds of food waste ^[3]. Guac, another AI ordering tool, says customers have cut food waste by as much as 38% ^[4]. Crisp forecasts demand to place orders and then estimates shelf life so retailers can reprice perishables before they spoil; one study puts the waste reduction from dynamic pricing at 21% ^[5][6]. The models take inputs including the weather forecast and the timing of food stamp disbursements, and in some cases place orders directly ^[2][7]. Fast Company reports that such tools can cut a store's waste by more than a third ^[1].

The mechanism that hits food banks is specific and worth naming, because it is not spoilage. Robbin Peterson, development director at West Seattle Food Bank, who worked in corporate grocery in the Seattle area in 2017 and 2018, says stores used to hold roughly 10% extra stock so shelves looked full, and that they no longer do: "we're not getting the fluff that they used to invest in" ^[12][13][14]. Resale apps such as Too Good to Go take another slice, letting stores clear near-expiry items before a food bank ever sees them ^[8][15]. West Seattle Food Bank reports a 10% decline in donations from the largest grocery stores it works with over the past six years, more than 15,000 pounds ^[11].

Supermarkets have historically been the largest source of donated food for most food banks, so the substitution shows up in budgets ^[10]. Joseph Slater, chief operations officer at Gleaners Food Bank of Indiana, says the food supply is tightening while demand for charitable food rises, forcing food banks to buy ^[9]. In 2018, most of Gleaners' spending went to infrastructure such as warehousing, and 16% of charitable donations went directly to buying food; this year 50% of the budget will ^[16]. As the source states them, those two figures are not strictly the same measure, but the direction is a roughly threefold shift toward purchasing ^[17].

Retail is only the visible end. Slater says meat producers that once donated slowed production after pandemic surpluses and found new markets, with low-demand cuts such as chicken drumsticks now exported rather than given away, and farmers are using software to predict demand at planting ^[18][19].

Two honest caveats. The evidence is anecdotal, and waste reduction does not have to come out of donations: sensors that move produce from storage to the sales floor before it turns destroy no gift ^[20][21]. Donation volume also still swings on things no model controls, including a recent surge of lettuce and bagged salads that shoppers avoided over cyclospora fears, and supermarkets still reject truckloads of imperfect produce that lands at food banks ^[22][23].

What to watch: whether any retailer or vendor publishes donation volume alongside waste reduction, and whether food bank purchase lines keep climbing. Right now the win is measured and the cost is not.