EXCLUSIVE: How the Produce Aisle Inspired a Grocery Tech Startup

Proving the notion that need begets inventiveness, a grocery tech company that started because of personal grocery shopping frustrations is taking its operations to the next level. Recently, the Delaware-based young tech firm Guac announced that it had completed a $2.3 million round of seed funding to help its mission of using artificial intelligence (AI) to better predict future demand for grocers while reducing food waste. 

The fresh funding includes backing from Y Combinator, 1984 Ventures and Collaborative Fund, along with support from Instacart, Roblox, Jane Street and Citadel Securities. The investors are banking on Guac’s algorithm, which incorporates hundreds of external variables, ranging from Spotify data to weather, to project shopper behavior. 

[Read more: “2024 Grocery Innovation Outlook”]

Progressive Grocer recently caught up with Guac CEO Euro Wang, who founded the business with Jack Solomon. 

Progressive Grocer: Tell us a little more about how your love of actual guac led you to innovate in this space. What other grocery industry experience did you have, including at Boston Consulting Group? 

Euro Wang: Funnily enough, we discovered the problem of bad grocery demand forecasting, because our favorite guacamole kept running out of stock at our local supermarket.

My co-founder, Jack Solomon, and I both also had experience in the grocery industry. At BCG, I witnessed just how expensive bad demand forecasting was even for major grocery retailers. Jack worked on machine learning research in grocery at Imperial College London – he focused on operations optimization for one of the largest grocery retailers in the UK. 

We were excited by the prospect of bringing our experiences to solve the problem of food waste, something we both care deeply about — I’m personally passionate about solving global food insecurity, and Jack is particularly passionate about preventing meat from being thrown away — he’s been vegan for 12 years. 

PG: What makes Guac different than other technologies or solutions available to grocers right now? 

EW: At Guac, we’re different because we really focus on hyper-accurate demand forecasting. 

There are a few startups and some legacy players in the grocery inventory space, but none that specialize in forecasting: They offer demand forecasting as a small component of a larger end-to-end supply chain solution. Instead of building lots of different features, we really focus on forecasting, which allows us to be the best at it.

Existing solutions that use machine learning for grocery demand forecasting use very simple machine learning models and only use historical sales data to train the model. At Guac, we use historical sales data, but we also incorporate hundreds of external data points — like weather, hyper-local events, sports games, and school term dates — to actually capture shoppers’ buying patterns. That’s our secret sauce that allows us to take forecasting to a whole new level of accuracy.

PG: How does Spotify data, in particular, play into more accurate forecasting?

EW: Grocery demand drastically fluctuates every day based on what’s happening in the real world — things like whether it’s raining, or if it’s the Super Bowl, or if there’s a parent-teacher conference at the local school. So, accurate forecasting is not just about using the best machine learning models, but it’s about how to capture the real world things that affect demand. If the model doesn’t know that there’s a Taylor Swift concert in the area next Tuesday, it won’t be able to capture the demand that’s affected by it. 

Spotify data is one of the unique datasets we use to understand the impact that events will have on local stores. We use Spotify listening data as a proxy for the number of people likely to attend a concert — when we combine this information with other variables like stadium and event venue capacity, we’re able to capture the impact of events with very high accuracy.

Another example of a unique dataset we use is sports betting odds — it turns out that when a sports match is predicted to be a close game, more people watch the game, which drives up beer sales!

PG: How is this solution scalable for grocers that may have smaller operations and for larger chains?

EW: For every customer, we do custom integrations to ensure our demand forecasting can fit smoothly with their existing ordering system. We also spend 2-3 weeks tailoring our external variables for each customer we work with, collecting external data points that are particularly relevant for the specific retailer. Essentially, our solutions are custom-built for each customer, enabling us to work with customers of all sizes. 

We’ve also built up a huge data library across the U.S. and the world, so that we can quickly access hyper-local data points relevant to each store location. For example, our events data library covers the entire U.S. — when we put in a store’s location anywhere in the U.S., our data library can give us events data for that specific zip code. This allows us to deliver the same level of hyper-granular external datasets for a 2,000 store retailer as well as a 10-store chain.