Search
Homepage This page's url is: -crn- Rankings and Research Companies Channelcast Marketing Matters CRNtv Events WOTC Jobs Dell EMC Newsroom Hitachi Vantara Newsroom HP Reinvent Newsroom IBM PartnerWorld Newsroom HPE Zone Tech Provider Zone

Sweet IoT Journey: How One Solution Provider Helped Implement Microsoft Azure Machine Learning At Hershey

The strategy involved several challenges but being a first-mover in the Internet of Things was key, says New Signature's Luis Morinigo in a session at the Midsize Enterprise Summit.

An early dive into the Internet of Things landscape yielded sweet rewards for The Hershey Company once the chocolate manufacturer began tracking the weight of Twizzlers during production.

That was the story Luis Morinigo, IoT and advanced analytics practice lead for Washington, D.C.-based solution provider New Signature, shared with a audience of midmarket IT leaders Monday evening at the Midsize Enterprise Summit. He and George Lenhart, the former senior manager of IS disruptive solutions at Hershey, explained that while the journey involved several challenges, their ability to leverage machine-learning-fueled analytics ultimately paid meaningful financial dividends.

Morinigo said the advent of scalable cloud-based analytics – Microsoft Azure Machine Learning, in this case – became a catalyst for the Twizzlers project because it reduced technical barriers involved in implementing data science capabilities. As other companies began applying IoT to maintenance and operations, Hershey and New Signature felt a clear need to move forward.

[Related: 5 Key Technology Trends For Midsize Enterprise Companies ]

"George and our team realized it was time to act," he said at the event, being hosted by The Channel Company this week in San Antonio. "Because if we could be first-movers in the market, we could certainly open up many opportunities as it related to building out these proactive capabilities."

Morinigo and Lenhart met resistance within their respective organizations, which did not initially consider IoT part of their "critical road map." Lenhart said Hershey rejected the idea four times before he finally got the green light from his CIO. At New Signature, a Microsoft systems integrator, Morinigo said he was able to build an IoT and advanced analytics team because he felt the scalability of the cloud finally enabled business intelligence initiatives for significant success.

Once approved, the pair moved to demonstrate value as quickly as possible during an allotted six-week window. They didn't aim for perfection, knowing they could improve and retrain the model once more data had been collected.

Completed in just two weeks, the first model built by Hershey and New Signature achieved 60 percent accuracy – not perfect, but a step in the right direction when it came to implementing IoT.

"I said, 'Wow, 60 percent is pretty good considering today we have nothing,'" Lenhart said.

Tracking the net weight of Twizzlers during the production process proved difficult, they said, because cooking temperatures fluctuated within a wide range. Hershey had developed a way to manage the weight variability of the candy, but this was a reactive process built on historical data collected in 15-minute intervals.

The pair saw machine learning as the prescriptive solution. More than 20 sensors were installed around the manufacturing plant that allowed both teams to capture hard-to-track variables. By isolating these, the machine-learning model could predict the weight fluctuations of Twizzlers in real time. The more often Hershey could pull that off, the more savings realized, Lenhart said.


So Hershey and New Signature had each of those sensors track the cooking temperature on a second-by-second basis for two months, generating some 60 million data points. The data collection process itself actually uncovered issues in the production plant, Lenhart said – anomalies such as a daily, 320-degree temperature spike that happened to occur during the machine cleaning process.

As they refined the model and gave plant engineers the latitude to make decisions based on what was predicted, IoT began to prove its value. Lenhart said the plant floor operators, who were only making 12 cooking adjustments per day before the implementation of machine learning, were soon making about 240 adjustments per day, dramatically reducing the candy's weight variability.

"The organization realized there were significant benefits that could be realized across multiple production lines," Morinigo said. "There's very much a human element involved with working projects like this – transformative projects – and several steps in the journey. That's important to realize as you're thinking about IoT and these opportunities."

Read more: Superstar Microsoft SI New Signature Gets $35M, Names New CEO

Back to Top

Video

 

sponsored resources