loading...
Guac's AI predicts perishable demand down to 0.95 units per SKU – so stores order exactly what they'll sell across thousands of items with different shelf lives.
ENTRY ANGLES
AI demand forecasting platforms with deep category-specific external data integration · Real-time signal tracking for content creators to optimize post performance based on trends and competitor activity · Demand forecasting for physical retail categories beyond groceries and apparel
VERTICALS
CAPABILITIES
Integration of category-specific external data sources, AI demand forecasting modeling, Real-time trend and signal tracking
Guac helps grocery stores forecast demand for perishables – fresh produce, frozen goods, packaged staples – so they order exactly what they can sell, and stop losing money on what ends up in the dumpster.
The platform's AI engine can handle demand prediction across thousands of SKUs with different shelf lives, whether that's loose fruits and vegetables sold by weight, frozen meal kits, or boxed pantry items.
Guac claims its forecasts are remarkably precise: the average error per SKU is 0.95 units – whether measured in kilograms, individual pieces, or packages.
Where does that accuracy come from?
The key insight is that grocery demand doesn't live in a vacuum. A simple example: when a major sporting event is broadcast, beer sales spike because people stock up to watch at home. The same dynamic plays out with ice cream on hot days or frozen ready-meals on rainy ones, when nobody wants to go out for lunch.
So Guac's algorithms don't just look at a store's historical sales data – they incorporate signals from more than 230 types of external events that can affect what sells, where, and when.
An important feature of the platform is that it isn't a black box. Every forecast comes with an explanation of what drove it, so buyers can actually understand and trust the recommendation.
Beyond regional factors like weather or event broadcasts, each physical store is shaped by its immediate surroundings: a nearby school where students buy snacks during the academic year, a local venue that draws weekend crowds, a construction project bringing in an unusual demographic. Guac's onboarding process begins with mapping these local dynamics before any forecasting begins.
The output is ready-to-submit draft purchase orders that factor in projected demand, current inventory levels, and supplier lead times. Procurement managers can review and edit before approving – and once confirmed, orders flow automatically into integrated supplier systems, whether that's SAP or a simple Google Sheet.
Implementation starts with a proof-of-concept on a single store. The store uploads historical sales data, and Guac's team spends two to three weeks layering in external and local factors – then shows the manager exactly how much they could save.
Guac was founded in 2023 and entered Y Combinator that summer, picking up $500K. Since then it has signed its first customers across the US, Europe, and the Middle East, and just closed a new $2.3M round.
Investor interest in Guac starts, as usual, with market size. Global grocery retail was a $12.1 trillion market in 2022, on track to reach $17.1 trillion by 2030.
The problem inside that market is staggering: a third of all food produced in the US gets thrown away, and grocery retailers account for 10% of that waste – simply because they buy more than they can sell before items expire.
In dollar terms, grocers lose an estimated 8% of revenue to poor purchasing decisions. About 5.9% is pure waste from expired unsold inventory; the rest comes from stocking out during unexpected demand spikes.
Get the forecasting right, and you could recover 4–8% of revenue. Across a $17 trillion industry, that's serious money.
Inventory imbalances are part of a broader retail problem. At any given moment, retailers worldwide are sitting on more than $500 billion in unsold stock – and a lot of it moves slowly, if at all.
Online retailers sell just 24.3% of apparel inventory within two months of purchase, 25.4% of cosmetics, and 48.8% of packaged consumer goods. Even over a full year, they clear only 68.7% of apparel and 47.8% of cosmetics.
AI has opened the door to a new generation of demand forecasting platforms that are meaningfully better than what came before – and most of them win by focusing on a specific product category, where the demand signals and data sources are unique.
A [related review](/review/ogromnyj-rynok-bolshaja-problema-no-reshenie-est) from December covered Syrup, which helps fashion retailers forecast demand down to specific colors and sizes – a notoriously hard problem. Syrup has evidently cracked it well enough to raise $24.8M.
The interesting parallel: Syrup also relies on far more than historical sales data. Its AI pulls in weather, holidays, social media trends, fashion editorial coverage, and other signals that can suddenly shift which styles and colorways people want to buy.
In other words, the real competitive advantage of both Guac and Syrup isn't that they use AI to forecast demand – it's that they know which external data sources actually matter for their category, and they've built the infrastructure to find, collect, and correctly interpret those signals.
The broad direction here is building AI demand forecasting platforms for specific product or service categories.
The most critical – and underestimated – piece of these platforms isn't the model itself. It's the integration of a wide range of category-specific external data that can shift demand in a given location. That work is genuinely tedious, and many technically-minded builders skip it because they'd rather solve the modeling problem in the abstract.
That's actually good news for founders willing to go deep on the details. Doing the unglamorous work of tracking down the right data sources is a real moat against technically stronger but less thorough competitors.
Guac and Syrup are both natural templates to learn from. And the groceries and apparel markets are large enough that new entrants can still find room – the window just won't stay open forever.
Beyond those two verticals: the "external signals" principle applies well beyond physical retail. Content creators, for instance, already have tools to help them write posts – but post performance depends heavily on what competitors are publishing, what's trending in the news, which films or songs just dropped, what memes are circulating. An AI writing assistant that tracks and incorporates those signals in real time would be doing exactly what Guac does for grocery buyers.
The window into food retail and apparel is open now – but won't stay that way. The stronger play may be to identify the next category where demand is highly context-sensitive, the data sources are fragmented and underused, and no one has yet built the integration layer to connect them.