In the midst of post-election second-guessing over why so many polls and pundits failed to predict Donald Trump’s win, there was one clear success story: OpenText™ Election Tracker.
ElectionTracker, the web app that analyzed news coverage of the Presidential race from over 200 media outlets worldwide for topics and sentiment, was a great showcase for the speed, robustness, and scalability of the OpenText™ Information Hub (iHub) technical platform it was built on. With demands for more than 54,000 graphic visualizations an hour on Election Day, it ramped up quickly with no downtime, performance you’d expect from OpenText™ Analytics.
Moreover, the tracker’s value in revealing patterns in the tone and extent of campaign news content provided valuable extra insight into voter concerns that pre-election polls didn’t uncover, and that insight didn’t just end after Election Day.
It’s just one in the series of proofs-of-concept on how our unstructured data analytics solutions shine at analyzing text and other unstructured data. They bring to the surface previously hard-to-see patterns in any kind of content stream – social media, customer comments, healthcare service ratings, and much more. OpenText Analytics solutions analyze these patterns and bring them to life in attractive, easy-to-understand, interactive visualizations.
Also if some unforeseen event ends up generating millions of unexpected clicks, Tweets, or comments that you need to sift through quickly, iHub offers the power and reliability to handle billions of data points on the fly.
Hello, Surprise Visitors!
Speaking of unforeseen events: Some of the Election Tracker traffic was due to mistaken identity. On Election Day, so many people were searching online for sites with live tracking of state-by-state election results that electiontracker.us became one of the top results on Google that day. At peak demand, the site was getting nearly 8,000 hits an hour, more than 100 times the usual traffic.
Senior Director of Technical Marketing Mark Gamble, an Election Tracker evangelist, was the site administrator that day. “On November 8 at around 6 a.m. I was about to get on a flight when I started getting e-mail alerts from our cloud platform provider that the Election Tracker infrastructure was getting hammered from all those Google searches. I’d resolve that alert, and another one would pop up.”
“We had it running at just two nodes of our four-node cluster, to keep day-to-day operating costs down. Our technical team said, ‘Let’s spin up the other two nodes.’ That worked while I was changing planes in Detroit. But when I got off, my phone lit up again: Demand was still climbing. It was just unprecedented traffic.”
“So we had our cloud provider double the number of cores, or CPUs, that run on each node. And that kept up with demand. The site took a bit longer to load, but it never once crashed. That’s the advantage of running in the cloud – you can turn up the volume on the fly.”
“Of course, the flexibility of our iHub-based platform is unique. All the cloud resources in the world won’t help you if you can’t quickly and efficiently take advantage of them.”
Demand on the site was heightened by the Election Tracker’s live, interactive interface. That’s intentional, because OpenText Analytics solutions encourage users to take a self-service approach to exploring their data.
“It’s not just a series of static pages,” explains Clement Wong, Director of Analytics On-Demand Operations. “The infographics are live and change as the viewer adjusts the parameters. With each page hit, a visitor was asking for an average of seven visualizations. That means the interface is constantly issuing additional calls back and forth to the database and the analytic engine. iHub has the robustness to support that.” (In fact, at peak demand the Tracker was creating more than 15 new visualizations every second.)”
“Some of the reporters who wrote about Election Tracker told us how much they enjoyed being able to go in and do comparisons on their own,” Gamble says. “For example, look at how much coverage each candidate got over the past 90 days, compared to the last 7 days, then filter for only non-U.S. news sources, or drill down to specific topics like healthcare or foreign policy. That way they didn’t have to look at static figures and then contact us to interpret for them; the application granted the autonomy to draw their own conclusions.”
Great Fit for Embedding
“The self-service aspect is one reason that iHub and other OpenText Analytics solutions are a great fit for embedding into other web sites (use cases such as bank statements or utility usage)”, Gamble adds.
“First of all, an effective embedded analytic application has to be highly interactive and informative, so people want to use it – not just look at ready-made pages, but feel comfortable exploring on their own.”
“Embedded analytics also requires seamless integration with the underlying data sources so the visuals are integral and indistinguishable from the rest of the site, and it needs high scalability to keep up with growing usage.”
The iHub/InfoFusion integration underlying the Election Tracker is already being used in other proofs-of-concept. One is helping consumer goods manufacturers analyze customers’ social media streams for their sentiments about the product and needs or concerns.
“If you think of Election Tracker as the Voice of the Media, the logical next step is Voice of the Customer,” Gamble says.
The Election Tracker is headlining the OpenText Innovation Tour, which just wrapped up in Asia and resumes in spring 2017.