Poor forecasting of food demand results in more waste than expected.

According to someone sauce, U.S. grocery stores throw away 10% of the approximately 44 billion pounds of food the country produces annually.It’s not just bad for the environment – food waste is a major source carbon emissions — but expensive for grocery stores. around Retail Insights Food and grocery retailers lose up to 8% of revenue due to inventory shortages.

Entrepreneurs Euro One and Jack Solomon say they have experienced first-hand the micro-level impact of prediction problems, as their local supermarkets often run out of their favorite guacamole.

“We found that even the largest retailers are having trouble predicting future demand and are frequently experiencing overstocks and understocks,” Wang told TechCrunch in an email interview. Told. “Recent extreme weather events have exacerbated fresh produce shortages, making it even more important to allocate limited supplies efficiently. Added to this is inflationary pressures and rising labor costs. , grocery store profits are increasingly threatened.”

Wang and Solomon co-founded the company with the idea of ​​using technology to tackle problems. Guac, a platform that uses AI to predict how many items a grocer will sell per item at a given store location each day. Guac recently raised $2.3 million in a seed round led by 1984 Ventures with participation from Y Combinator and Collaborative Fund.

“Food waste and food security are issues that Jack and I care deeply about, and we were very excited about the opportunity to actually solve food waste at the source,” Wang said.

Previously, Wang worked at Boston Consulting Group and Solomon researched AI for grocery logistics. We both graduated from Oxford University, where we met.

At Guac, engineers Wang, Solomon, and Guac have developed a custom algorithm that predicts grocery order quantities by taking into account variables such as weather, sporting events, betting odds, and even Spotify listening data. We are trying to understand consumer purchasing behavior by building a. Guac customers receive recommendations such as expiration dates, minimum order quantities, promotions, and supplier lead times that are integrated into their existing inventory ordering software and workflows.

“Traditionally, forecasting was done using Excel formulas or simple regression models,” Wang says. “But for fresh produce that expires quickly, you need something better. Because we use so many external variables, we can identify the real-world variables that cause changes in demand.”

Guac is certainly not the only startup in the food demand forecasting game. Crisp, which provides an open data platform for each link in the grocery supply chain, and Freshflow, which is building AI-powered predictive tools to help retailers optimize fresh food inventory replenishment.

But Wang says Guac is differentiated by both its commitment to transparency and its thorough tweaking of its predictive models.

“Rather than a black box that magically predicts a 20% increase in demand, our machine learning model tells our customers: “This 20% increase is due to conferences being held nearby,” Wang said. “Even if a retailer is already using machine learning, we can improve our predictions by having access to more external data sets. Including only specific datasets (such as weather or holidays) actually doubles the prediction error.”

Some early customers seem confident that Guac can add value. The company partners with retailers including grocery delivery companies in North America, Europe and the Middle East, including an unnamed supermarket chain with about 300 locations. Guac is also already profitable and expects to expand its engineering team next year.

“The grocery industry is quite resilient to economic downturns,” Wang said. “Everyone has to eat, but when the economy slows down, fewer people eat out and more people actually buy groceries. The pandemic has also accelerated the digitalization of grocery stores, making predictions We can now integrate more seamlessly with our customers’ systems. Speaking of the pandemic, shopper behavior has been very different during the pandemic, as grocers only have access to historical sales data from the past three years. This means that it is very difficult to rely on and predict future demand. Our algorithm allows us to adjust for how the pandemic biased sales data in 2020 and 2021. “We can also adjust for the residual effects of the pandemic afterwards.”