Like most retailers, Walmart has been basing decisions on weather data for years in obvious ways, such as putting up umbrella or snow-shovel displays in advance of rain or snow.

But now, in the second year of an extensive partnership with the Weather Co., the Earth's largest retailer is delving far deeper into sometimes unlikely correlations between weather and store sales on a Zip Code level. Even when those correlations make no obvious sense, Walmart has been acting on them with store-level merchandising and hyper-local digital advertising -- and achieving big results, according to U.S. Chief Marketing Officer Stephen Quinn.

In recent talks at the University of Arkansas' Center for Retail Excellence fall conference, at the Association of National Advertisers Masters of Marketing Conference in Orlando, and with Ad Age, Mr. Quinn revealed some of the unusual weather-purchase relationships his team has discovered.

"We didn't know, for example, that when it's low wind, that has some impact on whether or not people will eat berries," Mr. Quinn said at the University of Arkansas Oct. 9. Ideal berry weather turns out to be low wind with temperatures below 80 degrees. So, Walmart has begun serving up merchandising displays and digital ads for berries in Zip Codes where such weather exists, and as much as tripling berry sales when it does, he said.

Walmart also has found people are more likely to eat steak when it's warm out with higher winds but no rain, but not if it gets too hot. On the other hand, ground beef does better with higher temperature, low wind, and mostly sunny conditions. Salads sell better when the temperature tops 80 but winds are low. "You don't necessarily have to know why," Mr. Quinn said. "Just serve up the hamburger ads in those conditions," which he said has led to an 18% improvement in sales.

"We're now able to do this at scale," he said. "So we now can say to a partner, such as Gatorade, that we'll only serve up your ads when it's 95 degrees out."

Walmart has found thousands of such correlations that it's now trying to harness, he said.

Among other ways Walmart is trying to use data to improve its merchandising or marketing is by tracking trends on Pinterest. "We can see stuff trending on Pinterest, like all these inexpensive crafts people can make with Mason jars," Mr. Quinn said at the ANA conference Oct. 17. "And we'll put that Pinterest post right over an end cap and all the things you need to do that craft for $5 or $7, and people are just delighted. It makes Walmart seem more relevant."

Within two years, Mr. Quinn said he expects Walmart will be taking data from shopping lists saved on its app and using an algorithm to show people the most effective path through the store to find everything. Store navigation is an area where he acknowledged rival Target is ahead right now with its Cartwheel app, but he said he believes most retailers soon will have similar capabilities.

Matt Kistler, senior VP-global customer insights and analytics at Walmart, said at the Arkansas conference that while the retailer has plenty of data, it's still short on insights. For instance, it would like to know more about millennials.

"We have a lot of information about dollars spent in certain categories, certain likes or dislikes, but it's no real insight about our business," he said. Walmart buyers are hearing from some suppliers that its market share is flat with millennials, from others that its share is rising and from at least one that its share is falling.