As Joe Vargas worked on the marketing desk of a major tree fruit grower, he noticed a problem: Sales associates relied on printouts of spreadsheets and their “gut” instead of forecasting. He knew there had to be a better way.
Now, as the CEO of Huckleberry Signals, he says he and his CTO — artificial intelligence veteran and data architect Amanda Kuelker — hope to help grower-packer-shippers not only forecast but also improve outcomes in their produce businesses.
While Vargas says he helped his company build a modern data stack incorporating data from the Washington State Tree Fruit Association, Nielsen data and more to better understand the business, he says many companies are still analog.
“People don’t want to build dashboards,” Vargas says. “They don’t want to build out Excel workbooks.”
Beyond the Spreadsheet
What produce desks need is an analyst, but what Vargas says happens to companies that do employ an analyst is that they get bogged down with requests for information when the entire company — from CEO to marketing — could benefit from having access to that data.
He says this is where AI and large language models can truly help find, process and contextualize data coming from all sorts of sources, including enterprise resource planning systems, warehouse systems, business intelligence tools and even “institutional knowledge.”
“We’re a full infrastructure,” Vargas says. “This is where we can come into any organization that has any sort of POS system or QC (quality control) system or any part and bring all that data in and model it in a way that makes sense to the person that’s consuming the information.”
As a produce business onboards Huckleberry Signals, it deploys an agent called Huck to learn more from the key stakeholders using the platform. Huck will learn what questions the stakeholders need answers to, organizational questions about how pieces of data are referred to (i.e., liquidation vs. grow or profits vs. returns) so that Huck better understands how people will interact with it and respond with better information.
Vargas says a good example of what Huck can bring to the table is if a salesperson is on a call with a major retailer that asks about last year’s sales figures. Instead of that salesperson having to hang up, ask an analyst who will have to look up the information and then get back to the retailer, Huck can provide that information quickly and then offer some other information that might be of interest to the retailer.
Leveling the Playing Field
“It flips the script,” Vargas says. “I’ve watched growers, packers, shippers be just so outmatched by retailers … because they have the data. … If you can come back with answers that are backed by facts, now you’re in a whole new game. What I’m passionate about is really trying to get the grower returns. If there’s a fifth-generation grower and wants to be a sixth-generation grower, we want to give them that opportunity.”
Vargas says his goal is to help organizations take in any kind of data — whether that be QC data, inventory, sales, Nielsen panel or weather — and make it accessible and helpful. Vargas emphasizes that because produce is a timing-sensitive business, the goal of Huckleberry is to bridge the gap between a company’s disconnected data silos before the clock runs out on a perishable shipment.
“That’s really what Huck is about: getting down, giving each person on your team the ability to ask questions, get answers,” he says. “You’re really working with Huck like he’s your analyst.”
Vargas also says that Huck will continue to improve over time, as with all large language models. He says Huck will be able to provide descriptive analytics, which he calls “the mean, the median, what happened last year versus this year,” to predictive analytics, where Huck can determine what is going to happen based on historical data to prescriptive analytics, where Huck will be able to offer recommendations.
“And that takes some time, because you have to understand there’s a feedback loop with our program, but it’ll get there,” he says. “That’s six months to a year working with an organization. We start building out machine learning models. We start doing some of these other data science methodologies. We can get there and really work with these companies, and Huck becomes less of an analyst and more of a colleague. This is somebody who knows your business, and they can iterate on the data so much faster on such a bigger data set than people can do in an Excel workbook.”
Vargas says that while spreadsheets might have a place on a sales desk, it’s the lack of information available at an organization’s fingertips that is most costly.
“I think that a lot of companies are just drowning in data but starving for answers,” he says. “There’s data everywhere. They just don’t know how to use it, and they don’t have the tools, and they don’t have the know-how.”
He says this is exactly what Huckleberry Signals hopes to accomplish.
“We meet you where you are,” he says. “You bring your data. We go and investigate. We have architects on the team. We can set all these things up for you at a fraction of the cost.”
Built for Security
Vargas says the data in Huckleberry from each organization has protections and silos so, for example, a competitor also using Huckleberry Signals doesn’t have access to it. He says his team has also built significant guardrails to prevent hallucinations that many publicly available large language models are known for.
“All the data and analytics get brought into a structured data warehouse for us,” he says. “So, anytime that Huck is going to get information, he’s going and grabbing from a structured data set to report back on actual numbers, and we can go through and validate that very quickly.”
Huckleberry Signals is also undergoing SOC 2 certification to ensure the data is protected and secure and, of course, accurate.
“When it comes to analytics, you have to be very careful, because that decision you’re making could be a million-dollar decision, and you’re basing it off of data,” Vargas says. “So, that’s where our focus is, really. We have a lot of fail-safes, a lot of stuff in place.”
Vargas says Huckleberry Signals will be sold as a subscription model to ensure all companies using the system will be running on the most up-to-date version. While many in the fresh produce industry are reluctant to incorporate AI, those who take a wait-and-see approach will be left behind, he adds.
“A year from now, you’re going to have to start at the same location,” he says. “It’s not going to be the benefit that maybe we’ve seen with other SaaS (software-as-a-service) products, like the quicker you get in, the quicker you can get started, the quicker you’re going to have an edge on that competition.”
Why Huckleberry?
And as for why the name Huckleberry Signals?
“Huckleberry hasn’t been propagated, so it’s a very unique commodity,” Vargas says. “It’s close to home for me. I’m in Montana. I go huckleberry picking every year. It’s really a staple in my world.”
He says Kuelker also thought about the movie “Tombstone” and Doc Holliday’s popular line referencing huckleberry, which meant the name got her vote.
“Huckleberry and Signals is really a nod to getting answers faster,” he says. “Understanding the signals, understanding the market changes, trends — and so, putting those together is very unique, and it’s going to be something that gives you a competitive advantage.”
