Pizza Hut Will Use AI... to Recommend What Pizza to Get... Based on... the Weather - 4 minutes read

Pizza Hut has posited a strange question: What weather is pizza weather? Perhaps it’s a riddle. Or perhaps a machine-learning salesman in a brimmed straw hat waltzed into a Pizza Hut marketing meeting with a whistle and a tune:

Pizza when it’s fair

Pizza in a squall

Pizza pizza all the time

Pizza wall-to-wall!

Invest in machine-learning technology that’ll trick customers into eating pizza and chugging blue cheese dip depending on the weather la la la!

Yes, Pizza Hut is using AI technology to meteorologically predict what foods you might be most interested in ordering depending on cloud cover and humidity.

The news comes from a VentureBeat interview with Tristan Burns, global head of analytics at Pizza Hut, under the arm of Pizza Hut Digital Ventures. Its custom AI platform will analyze, among other things, the weather to “surface relevant product recommendations.”

It’s tempting to run with the weather angle, but that obfuscates the sketchy outline of a nascent machine learning strategy to probe your innermost pizza-related desires. Pizza Hut Digital Ventures will also “ingest customer behavior and a little bit about who customers are.” Unclear what behavior it shall ingest and whether this applies to when people order delivery from their browser or an app or Pizza Hut’s brick and mortar store, or everywhere.

It’s probably safe to assume everywhere. Pizza Hut’s parent company Yum!, which also owns KFC, Taco Bell, and WingStreet, announced earlier this year that it’s acquiring the AI and algorithmic analytics platform Kvantum. That company hoovers and consolidates data from social media and retailers, among other sources.

And so, Pizza Hut forges ahead alongside machine learning-thirsty fast-food competitors like McDonald’s and Sonic, which seem to be obsessively reducing friction using consumer patterns at every pain point. In 2019, we learned that McDonald’s spent $300 million on a startup Dynamic Yield Ltd, which tailored the digital menu feed tailored to time and location-related patterns, despite the fact that McDonald’s basically serves two types of meat in different shapes. (Fight me). That ostensibly applied to tailoring its drive-thru menus, which also factored in weather patterns and timing. Although determining that people should be shown an egg McMuffin and coffee in the morning doesn’t seem like rocket science.

Of course, McDonald’s is no fool: a quick look at the presser shows that it will also tailor “all of its digital customer experience touchpoints, such as self-order kiosks and McDonald’s Global Mobile App.” More like the same old adtech stuff. That year, McDonald’s also acquired a speech recognition start-up Apprente, in order to, McDonald’s said, “allow for faster, simpler and more accurate order taking at the Drive Thru with future potential to incorporate into mobile ordering and kiosks.”

Another fast food machine-learning company 5thru has touted a more invasive tactic: scanning license plates to pull up customer profiles, with prepayments and tailored menu boards to their liking. Er: “leverages technology for a zero-friction drive-thru experience.”

I don’t want license plate scanners anywhere for a variety of reasons laid out for years by civil rights organizations, but let’s really dive into how Pizza Hut’s menu offerings might dovetail with atmospheric conditions and their effect on the human body. Pizza, a hot food, falls under the FDA recommended category of optimal serving temperature at 140 degrees Fahrenheit. This also goes for Pizza Hut’s wings, cheese sticks, fries, cookies, and cinnamon sticks. From this data, we can extrapolate that pizza is warmer than body temperature and likely to remain warmish as it makes its way to your belly, therefore causing you to feel warm. You could perhaps chug a blue cheese dip on a cooler day, but does a person who chugs blue cheese dip obey logic, much less the comings and goings of the sun? Seasonal calorie intake patterns and airport locations might be a better metric.


Powered by