A Case of Analytical Improvements

Page 2 of 2

      Bookmark and Share

THE PROBLEM IN DEPTH
Every day, the small editorial staff of the B2B publisher SmartBrief is responsible for churning out more than 60 email newsletters. The newsletters, themselves called SmartBriefs, are sent out to organizations that span a wide swath of industries, such as healthcare, retail, consumer packaged goods, telecom, and finance. Needless to say, producing the daily newsletters is a time-intensive process that involves combing news sources across the web, assessing which of the news stories of the day are germane to each individual newsletter, and aggregating the stories into a coherent publication to be sent out to subscribers’ inboxes. Such an endeavor requires diligence, patience, sound editorial judgment … and a really good text analytics tool.

"We have a web scraper that finds stories online," says SmartBrief’s VP of technology, Chris McNeilly. "But we needed a way to classify them in a meaningful way so that our team of editors is able to find and select the best stories."

SmartBrief spent some time making the rounds of text analytics vendors, but it never quite found a solution that was a good fit. In particular, SmartBrief wanted something easy to use. In the past, its solutions were somewhat cumbersome, and they required a high level of expertise in order to function properly. According to McNeilly, "The products that we’ve used in the past, to get anything out of them you needed user experience or a degree in computational linguistics."

Ultimately, though, it’s really speed and efficiency that matter most to SmartBrief and its bottom line. McNeilly continues, "The end result is that we want faster story selection. We want to shorten the amount of time it takes to create a SmartBrief, and we want to improve the quality of the Briefs. That means being able to tone nodes quickly and being able to add nodes based on feedback from our editors. We don’t have the time to send all of our information out to consultants."

THE SOLUTION
Late last year, SmartBrief caught wind of a company called Lexalytics, which, at first blush, sounded like a company that had the right kind of text analytics solutions for SmartBrief. Lexalytics offered a rapid installation time for its product, comprehensive support for configuration and customization, as well as all of the text analytics essentials—entity extraction, entity relationships, document summarization, sentiment extraction, and classification toolkits.

Lexalytics CEO Jeff Catlin says that his company understood SmartBrief’s particular needs and approached the project with those needs in mind. "They do document classification of a wide range of content off the web," he says. "Is the information medically oriented? Is it about banking or finance? Our product has the ability to classify the content to improve the quality of the publications they deliver—because it’s all about the quality of the information they deliver. They wanted to make their automated system as good as possible in order to save money and time, because they review all of the content themselves."

Their willingness to "help us develop a workable solution in a fast time frame" was what finally led SmartBrief to select Lextalytics, McNeilly says. SmartBrief then initiated the process of integrating Lexalytics’ text analytics engine at the API level. "It will be a direct replacement for our users. In addition to the results being better, they don’t have to learn a whole new interface. We’re just swapping out the engine," McNeilly says.

"One of the big things about Lexalytics is that its software embeds right into the web platform," says SmartBrief’s McNeilly. "It’s not a client-server solution that gets bogged down—it scales much better with our web platform."

Lexalytics has played an active role in the installation of its product into SmartBrief’s system. It has worked with the publisher to ensure that the transition goes smoothly. McNeilly says that he has "seen a lot of emails going back and forth as they’ve been helping us with the installation. They’ve been very helpful getting us up-to-speed. We’ve got our old system and our new one running in parallel, and we’re making sure that the performance is what we want it to be. They’ve been making changes for us, and they’ve been helping us figure out what’s going on."

Once the installation is complete and the new system is up and running, SmartBrief plans to hire its own technician to handle the maintenance of the taxonomies and to do training on the system. "Usability was something we’ll be able to manage in-house, as opposed to outsourcing it," McNeilly says.

THE OUTCOME

The Lexalytics implementation has been a smooth one, and the company’s editors have already seen a marked decline in the amount of time it takes to produce a SmartBrief. To create the publication, a SmartBrief editor goes into a content management system and opens up a particular taxonomy. At that point, he or she will see topic areas, and within those areas are the hundreds of news stories that have been aggregated, available for viewing at a glance. This saves the editor the time of having to visit hundreds of trade publication or news sites individually to collect the stories. "What has been taking editors an hour and a half to do can now be done in less than an hour," McNeilly says.

SmartBrief has nothing but praise for Lexalytics and the benefits its has reaped from its new text analytics software. The improved efficiency of its workflow will allow the company to expand its email newsletter offerings into smaller niches. Whereas in the past, SmartBrief only offered newsletters for large industries, the company can now forge into more nuanced markets because it has the time and the tools to do so.

The editorial staff at SmartBrief will be able to work with the company’s in-house technician to refine and hone the taxonomy as time goes by and the company’s newsletters evolve. "Our editorial staff decides what the major topic areas are, and they will let the development team know. Once we have our full-time taxonomist on, she will go into the tool and bring in sample stories, extract keywords, and build up particular nodes. We hope it will be a productive back and forth between the domain experts—the editors—and the engineer," McNeilly says.

SmartBrief’s success with Lexalytics has led the publisher to consider putting Lexalytics technology to work in other ways in its operation. In particular, Lexalytics’ sentiment analysis capabilities—which analyze the tone of a document on the entity level and offer confidence scores that rate how well the technology believes it has evaluated the tone of a document—could be of use to SmartBrief. "We might be able to rank news stories according to sentiment for all of the companies that we track," McNeilly says.

For now, however, SmartBrief is basking in the glow of its newfound productivity and is grateful to have found such a responsive vendor in Lexalytics. "I really can’t stress enough how great their customer service has been," McNeilly says. "They’ve really tried to understand our needs, and they’ve worked hard to help us come up with workable solutions."

Page 2 of 2