System for diners orders up Azure big data platform in the cloud

Ziosk is joining the tabletop tablet computers it places in restaurants with Microsoft's Azure cloud platform and big data tools to try to improve the business of casual dining through machine learning applications.

Predictive analytics applications driven by big data have become the lifeblood of many businesses. From mobile network operators and financial services firms to retailers and online auto brokers, big data platform technologies and analytics tools -- sometimes running in the cloud -- are being used to cull information on customer activity and predict who is likely to buy what.

But some types of businesses have been left on the outside looking in. For example, dining experiences at restaurants typically aren't tracked digitally and analyzed either as quickly or to the same extent website visits are. This may not seem intuitive in an age when negative tweets from smartphones can torpedo a restaurant's reputation; but store managers and data analysts at restaurant chains are often left to dig through reports that are delivered well after the events they detail.

Hospitality technology and services provider Ziosk LLC is trying to address the issue by adding more automated and real-time analytics capabilities to its tabletop tablet system for casual-dining restaurants. The effort, which utilizes elements of Microsoft's Azure cloud computing portfolio, is aimed at enabling restaurants to automatically target promotional offers at, and meal recommendations to, diners.

As of January, Dallas-based Ziosk had placed its dedicated tablet computers on tables at 1,400 restaurants in the U.S. to streamline diner services and provide customers with online games, social media links, built-in cameras and other entertainment options. The tablets can be used to place orders and pay checks, said Kevin Mowry, chief software architect at Ziosk, whose clients include the Chili's and Red Robin chains.

Order up on real-time analytics

Ziosk previously offered reporting services for restaurant managers and data analytics teams. To support that, the company built a data warehouse based on Microsoft’s SQL Server, as well as menu management and survey systems. But analyzing real-time information about customer interactions with the tablets can provide big benefits for dining establishments, according to Mowry. 

One pixelColin White on managing big data
in the cloud

The first objective for Ziosk was tracking promotions in order to understand how and when people are likely to respond to them, so as to better predict which offers will appeal to particular customers. But "there is a lot more we can do," Mowry said. For example, data on what people viewed, clicked on and ordered can be collected from the tablets. That and other data can then be analyzed to make more informed guesses about what other diners might like.

"Today, if you're in the restaurant business, you have a printed menu, and you have an electronic check, but you don't know a lot more about what is going on with the guest experience," Mowry noted. Recommendations may not go far beyond a list of the day's specials; but on the other hand, as Mowry asserted, many retail websites use analytics algorithms and fast processing engines to offer fairly sophisticated product recommendations to visitors. "We think we can do that with food menus, too," he added. 

Turning to machine learning

For Ziosk, moving into the realm of real-time analytics requires the application of machine learning, a form of advanced analytics that works on large data sets and automatically adapts its predictions as the information being analyzed changes. And Mowry and his team are looking to a big data platform in the Azure cloud to make the leap into machine learning happen more immediately. 

Mowry said Ziosk followed in-house precedent in choosing Microsoft's technology. It also tapped Richardson, Texas-based Artis Consulting L.P. to help create a proof-of-concept deployment including Microsoft's  Hadoop platform, Azure HDInsight, and a predictive analytics application called Azure Machine Learning that the vendor released in February.

"We've been using Microsoft technologies for a lot of purposes for quite a while," Mowry said, describing Ziosk as a midsize company that needs to conserve its use of developer resources.

"There are a lot of complexities in machine learning models. They can require a certain set of skills that aren't easy to obtain," he added, mentioning R and Python programming as examples. What Microsoft has done with Azure Machine Learning is bring "a sophisticated form of predictive modeling to companies like ours. You could call it machine learning for the masses."

Teach your models well

Mowry's team interacts with the cloud application services via Microsoft's Azure platform as a service (PaaS) interfaces. He said that the machine learning models must be "trained," using known good data sets, and that doing so enables users to find out how closely new models adhere to reality.

"When you're training the model, you try different algorithms. That isn't difficult, because Azure Machine Learning has a lot of built-in algorithms," he said. "You link to a data source, run a test, then swap in other algorithms until you get to the best answer possible."

Microsoft claims it has used its experience building Xbox gaming networks and Bing search services to build some of the Azure machine learning libraries. Mowry found favor with the algorithms in the libraries -- those developed by Microsoft and ones created by third parties.

Machine learning is also an ongoing process, he said. As time goes by and more data is gathered from Ziosk's tablets, both the data sets and the machine learning algorithms can be refined. In turn, restaurant managers can finetune their menus and promotional offerings, and if all goes well, diners will get everything while it's hot.

Jack Vaughan is SearchDataManagement's news and site editor. Email him at [email protected], and follow us on Twitter: @sDataManagement.

Next Steps

Hear a startup's tale of data in the cloud

See what happens when big data meets the cloud

Meet two users dipping into machine learning

Dig Deeper on Big data management