Artificial intelligence is reshaping how pet food companies in Latin America set prices, forecast demand and manage their product portfolios, Iván Franco said during an education session at Petfood Forum 2026.
Franco, founder of Triplethree International and an independent advisor to pet food companies on pricing, market entry and category management, said the industry is collecting more information than ever without the systems to act on it.
"Today, you have more data than ever, but also less clarity than ever. And that's not a coincidence," he said.
Companies have dashboards, tracking systems and market research from many sources, but that volume is generating noise rather than direction, Franco said. "We have a lot of data, but we don't have a system that turns all that data into actionable answers and decisions," he said, adding that teams often default to repeating past tactics and hoping for results.
Weekly meetings tend to describe past performance without answering forward-looking questions, Franco said, such as how a 5% or 10% price increase would affect revenue, whether a product will run out of stock within weeks, or which SKUs to protect.
What AI adds: speed, scale and recalibration
Franco said AI delivers three advantages over traditional analysis: speed, scale and continuous recalibration. Building a price elasticity model once took four to six weeks and often required outside help at significant cost, he said, while such a model can now be set up in a few hours.
Models that were once limited to a single SKU, market or segment can now span hundreds of SKUs across multiple markets and channels at once, and they can be recalibrated as new data arrives, he said.
Segmentation uncovers a hidden high-value buyer
Franco described several Latin American case studies. In one, a company had relied for five years on three buyer segments built in a spreadsheet using two variables: average spend and price.
Using behavioral data and a K-means clustering algorithm, Franco's team identified six segments. One overlooked group bought premium products in smaller packaging and stayed loyal regardless of price or promotions, he said, and the company had classified those buyers as basic when they were a high-value segment.
An elasticity model supports a price increase
In Mexico, a company held its prices flat for 12 months during a period of high inflation even as production costs rose about 20%, Franco said, because its commercial team feared losing volume despite no data supporting that assumption.
"After an elasticity model indicated that premium buyers were relatively inelastic to price, the company raised premium-line prices by 8% to 12%," he said. "Blended margin improved by about four points and volume fell only 1.5%, in line with the model's prediction."
Measuring whether promotions pay off
A third company ran promotions three months at a time, Franco said. Analyzing what happened after each promotion, his team found that only six of 16 SKUs generated real incremental demand, while the rest pulled forward future sales.
"You run a promotion and you pull forward demand from the future," he said. "You borrow demand, and you pay for it with margin. The company cut its promotional calendar, running one-month promotions on only the six SKUs that showed incremental demand, and recovered about six points of gross margin."
Rethinking a portfolio delisting
In another case, a major U.S.-owned retailer in Latin America considered delisting the premium item in a three-SKU family because its sales trailed the other two, Franco said.
"The company assumed buyers would downgrade to a lower-priced option, but the model predicted they would instead switch to a competitor or trade up," he said. Franco advised against the delisting, and the company has held off on the decision.
From data to decisions
Franco urged attendees to start by identifying the decision costing them the most, then use the data they already have — two years of history is enough, he said — to build models such as price elasticity, distributed lag promotion analysis, demand forecasting and consumer segmentation.
The goal is not another report, he said: "You don't need more bots. You don't need more market research reports and slides. You need decisions."
The future of the market will be defined not by who offers the most but by who makes better, faster and more accurate decisions, Franco said.
