Rawpixel.com - Fotolia

Wi-Fi provider uses big data analytics to drive customer experience

The self-styled “Uber of Wi-Fi”, iPass is using Databricks’ Unified Analytics Platform and machine learning to monitor its worldwide network and help its customers get the best experience

Wireless network and connectivity aggregator iPass has improved its customer service proposition, gained new insights into its business and even introduced new product lines, after deploying Databricks’ Unifed Analytics Platform and machine learning capabilities to monitor millions of Wi-Fi hotspots in real-time.

Just as Uber owns no cars itself, iPass owns no actual Wi-Fi network access points (APs), but acts instead as an aggregator of publicly available hotspots in locations such as airports, hotels and shopping malls, selling its single sign-on service to both consumers and enterprises with highly mobile workforces.

It counts airlines such as Emirates and SAS – which offers pilots access to the iPass network in their electronic flight bags (EFBs) – among its key customers, as well as large tech enterprises such as Ericsson and Hitachi, and incidentally, Uber. Its network partners include big names such as AT&T, BT, China Telecom, Orange and Swisscom.

Around three and a half years ago, the organisation decided to pivot from simply reselling connectivity services to a software company offering connectivity as a service powered by big data, analytics and machine learning.

Tomasz Magdanski, iPass director of big data and analytics, who transferred from London to Silicon Valley to lead the project, explained that this pivot came about because while iPass had successfully established itself as a connectivity layer for its customers, it effectively had to also maintain the whole service despite the physical layer being owned by someone else entirely.

“The challenge we used to have was how to monitor and manage the user experience, and steer customers on where to connect, where not to connect and why, based on things like reliability and security,” Magdanski told Computer Weekly.

“We wanted a platform capable of collecting all the data our applications around the world are generating and merging that into a recommendation engine to serve our clients.”

Hadoop cluster

To start out, Magdanski explained, he implemented an on-premise Hadoop cluster in iPass’ datacentre, but this turned out to be a less than optimal decision – the build and implementation grew so complex that the process ended up taking iPass 18 months.

“It was brand-new technology and we didn’t have the expertise in-house,” he said. “By the time it was done there were Spark upgrades and other parts of the infrastructure needed replacing, so the team had to start all over with new versions, new cycles, to just stay current with the technology. It was very hard to stay on top of it with a small team and a small budget.

“The next thing to hit us was scale – we got Hadoop up and running but the data and compute started to scale to the point where we started evaluating whether it even made sense to run big data, it was so overwhelming.”

As the crisis at iPass’ datacentre deepened, Magdanski knew the only solution was to move from on-premise Hadoop up into the cloud with additional Apache Spark to better manage and scale the organisation’s data, and meet customer expectations of 24/7 network access.

“We looked at cloud providers and ended up comparing Amazon EMR (Elastic MapReduce) to Databricks. Based on the experience of the previous 18 months, we knew we needed expertise more than just a cloud environment, but Databricks was actually able to do both,” he said.

“They gave us the ability to launch clusters with the touch of a button, reconfigure on the fly, and access to their engineers to bring us up to speed.

“It took 18 months to set up the first iteration of our big data initiative, and six weeks to set up, build and release the second.”

Read more about big data analytics

  • Big data and predictive analytics may seem synonymous, but understanding the constraints of each discipline is the key to extracting business value from projects that combine them.
  • Before purchasing big data analytics software, companies must first identify their specific needs and then evaluate how the product features address those needs.
  • Companies that leverage data to optimise business processes will not only survive, but thrive during market downturns, and dominate the industry on the next upswing, this report from Quorum argues.

Databricks’ Unified Analytics Platform is specifically designed to eliminate the need for disparate tools, reduce deployment times, and improve the productivity of data engineering teams, something Magdanski says has been a noticeable plus for iPass. His team is now also using Databricks to build exact, transform, load (ETL) pipelines, streaming, analytics, machine learning, data lake processing and reporting.

“Databricks was the obvious solution as it’s equipped to collect and scale massive amounts of Wi-Fi data. We have been using Databricks in development through to production, which has allowed our data engineers to focus less on building infrastructure and more on creating new data products and providing the best Wi-Fi recommendations to mobile devices,” said Magdanski.

Beyond giving iPass’ IT department some much-needed breathing space, the implementation of Databricks has helped the organisation deliver more relevant and timely information to its customers using data-driven algorithms.

Helping users’ connectivity decisions

It can now help users make far better connectivity decisions by providing them with data on white- or blacklisted APs, steering connections to Wi-Fi or 4G mobile networks based on performance, or from unprotected hotspots to protected ones, for example.

“We now provide reports and dashboards to customers to help them understand their network usage and how our platform fits into their broader connectivity strategy, keeping them secure and productive as they travel,” said Magdanski.

The implementation has also enabled iPass to introduce new product lines. In late 2017, it announced Veri-Fi, a suite of analytics services that effectively productises and monetises the network data itself.

The aggregator is now able to sell data on aspects of performance such as AP functionality, customer experience, signal interference and more back into its service provider partners – the actual network owners – to help them monitor, manage and improve their own services.

For iPass, this is creating more meaningful, deeper relationships with its business partners, and most importantly for its shareholders, new revenue streams.

Read more on Wireless networking

Data Center
Data Management