It’s astonishing, but 30% of all produced food products end up in landfills! Among them, 10% are discarded by grocery stores due to expiration, resulting in financial losses. Consequently, stores are willing to invest in platforms capable of predicting the demand for products. However, the catch is that these platforms must also be able to consider additional external factors, like these.
Project Essence
Guac helps grocery stores forecast demand for fruits, vegetables, and other perishable goods to procure them in the quantities they can sell — avoiding losses from disposing of expired unsold items.
The startup’s AI engine can handle sales forecasting for thousands of product SKUs with varying expiration dates, whether it’s fresh fruits and vegetables sold by weight, packages of frozen products, or boxes of groceries.
Guac claims that their forecasts are highly accurate, with an average forecasting error of 0.95 units of product measurement — kilograms, pieces, packs, cans — for each SKU.
How does it achieve such precision?
The key is that the demand for products depends on a myriad of external parameters. For instance, a humorous example: if an important sports match is expected to be broadcasted on TV, it sharply increases beer sales because people stock up on beer to watch the match. Similar stories happen with ice cream sales on hot days or frozen semi-finished products during rainy weather when people prefer not to go out to cafes for lunch.
Therefore, Guac’s algorithms, in building forecasts, rely not only on historical sales data in the store but also leverage information on more than 230 external types of events that can influence the sales of specific types of products in a particular store.
An additional advantage of the Guac platform is that it is not a “black box” that outputs forecasts from unknown sources — each forecast comes with an explanation of what it was based on.
In each offline store, demand depends not only on the general regional situation like weather or sports broadcasts but is also temporarily or long-term influenced by events in its vicinity: events happening nearby, students from a neighboring school regularly buying small items, but only during the school year, and so on.
For this reason, the implementation of Guac begins with analyzing the surroundings and identifying local factors that may additionally influence demand in a specific store.
The platform’s output is the issuance of ready-made draft orders to the purchasing departments of stores, taking into account future demand, the current state of the warehouse, and the delivery time of ordered goods. Purchasing managers can edit these orders based on their considerations.
Confirmed orders automatically enter integrated systems for interacting with suppliers through Guac — whether it’s complex systems like SAP or simple tables in Google Cloud.
The implementation of Guac starts with demonstrating the platform’s capabilities immediately using a specific store as an example. The store manager uploads historical sales data to the platform, and Guac specialists, within 2–3 weeks, add general and local factors that may affect sales — after which they show how much the store can save on purchases.
Guac was founded in 2023, and in the summer of the same year, it joined Y Combinator, receiving $500,000 in investments. Since then, it has found its first clients in the USA, Europe, and the Middle East, and now it has raised new investments amounting to $2.3 million.
What’s interesting
As usual, investors’ interest in Guac is explained by the size of the potential market. The food sales market was valued at $12.1 trillion in 2022 and is projected to grow to $17.1 trillion by 2030.
The problem lies in the fact that in the United States alone, one-third of the produced products are discarded. Grocery stores contribute to 10% of this waste because they purchase more products than they can sell before their expiration date.
In monetary terms, grocery stores lose 8% of revenue due to poor procurement planning. This includes a loss of 5.9% of revenue because of discarding expired unsold items. The rest is lost because they purchase too little during sudden spikes in demand for specific products.
Therefore, effective procurement algorithms can increase the revenue of grocery stores by 4–8%. And 4–8% of $17 trillion is a significant amount of money.
Unbalanced inventory in warehouses and stores is part of the overall problem of unbalanced retail inventory. At any given moment, there is over $500 billion worth of unsold goods in the warehouses of retail sellers.
These goods sell very slowly, if at all 🙁
Looking at online stores, within two months, they sell only 24.3% of purchased clothing, 25.4% of cosmetics, and 48.8% of Consumer Packaged Goods (CPG), including groceries and beverages. Even within a year, they manage to sell only 68.7% of clothing, 47.8% of cosmetics, and 87.2% of packaged goods.
The development of AI technologies has led to the emergence of a new generation of platforms that better address the procurement and inventory problems by building more accurate demand forecasts. Each of these platforms usually focuses on specific product categories, as forecasting demand for each category has its own specifics.
For example, in December, I wrote about the startup Syrup, which helps stores forecast demand and place orders for clothing purchases. Syrup’s strength lies in its platform’s ability to predict demand down to the colors and sizes of clothing — a very challenging task. Syrup seems to be successful in this aspect, as it raised $24.8 million in investments.
The trick in forecasting clothing sales with Syrup is that they also rely not only on historical sales data. Their AI engine uses information about weather, holidays, trends on social media, articles from fashion magazines, and other data that can influence the sudden popularity of certain types of products and even their colors.
In other words, the value of Guac and Syrup lies not just in using AI technologies to forecast demand but in understanding which external data, such as fashion trends or events, can impact sales, and knowing how to find, collect, and interpret them appropriately.
Where to Run
The general direction of movement is the creation of AI platforms for forecasting demand for specific categories of goods or services.
Arguably the most crucial part of such platforms should be the integration of a large amount of specific external data that can influence sales in a particular category and location.
This is a separate and often challenging task that many tech enthusiasts simply cannot handle, as they are more focused on finding a solution to the problem “in principle,” and they are not willing to delve into the intricacies of external details. This, in fact, is quite beneficial, as it gives founders who are ready to delve into details the opportunity to outperform many tech-savvy competitors 😉
Examples that can be emulated for this purpose are precisely the ones mentioned today – Guac and Syrup. Implementation can even target the same products and clothing, as these markets are so vast that there is still room for many players. That is, if you don’t miss the opportunity, of course 😉
In which other areas would you like to venture with similar forecasting platforms?
By the way, the use of additional parameters influencing demand in the broad sense of the word is important for AI platforms in other unexpected areas as well.
For instance, there are many tools now helping bloggers write posts. However, the popularity of these posts will depend not only on what is written in them but also on various external factors – what competitors are blogging about, what’s in the news, what new movies and songs have been released, what topics are currently trending, what fashionable memes are circulating, and so on. Good AI assistants for bloggers should be able to skillfully and timely use such external data.
What other areas, besides blogging, can you name for applying demand forecasting platforms in a similar broader sense?
About the Company
Guac
Website: tryguac.co
Latest funding round: $2.3M, December 19, 2023
Total investments: $2.8M, rounds: 2