Lots Of Data, One Analyst, Many Reports -- Narrative Science

Narrative Science can make an equity research analyst 20 to 100 times more productive by using computers to turn data into narrative that sounds like a person wrote it, according to the company CEO. The Chicago-based company, which grew out of some journalism research at Northwestern, has developed software called Quill, that can create news stories, operational reports, customer statements and investment research.

“We can do it at incredible speed and scale and sounds like a human wrote it,” said Stuart Frankel, the CEO. “It is a horizontal platform, it can take just about any data from any industry and create any kind of narrative content.”

With a number of industries to choose from, the company decided to work on financial services as one of its core verticals with an initial focus on a few areas. Investment research is one — helping companies expand the universe of companies, stocks and  publicly traded debt issues that they can cover.

“A traditional research analyst has only so many hours in a day and can cover only so many stocks,” Frankel explained, “but if we can take what the analyst does, codify it, map it to our system and use it to leverage the analyst’s knowledge, we can increase his productivity 20-100x.”

Another area for Narrative Science in financial services is in customer communications where wealth management firms tend to inundate customers with a clunky amount of data oddly presented in lists and pie charts without much useful insight on how a customer’s portfolio is performing.

“We can take the data and deliver a personalized statement showing a customer’s portfolio, how it is doing in absolute and relative terms and against benchmarks and show the advice or moves a financial advisor has made. We are also doing things to help financial advisors prepare for meetings with clients,” Frankel added.

“Now they will probably pull your account and try to refresh their memory and look at their notes. We we can assemble a document that prepares that financial advisor for that meeting, giving them talking points and topics they should focus on.”

The narratives are individually tailored, he added.

“One of the powerful aspects of our technology is that it is not template-driven. Each report is built from the ground up. It starts with the data and whatever the system gleans from the data it will generate in a document for the audience, even an audience of one.”

The company was selected for the 2013 FinTech Innovation Lab in New York, providing exposure to the leading money center banks in the city.

“It was great; it is great,” said Frankel of the intensive 12-week program whose impact tends to continue for months of interactions with banks after the formal program ends.

“The Lab gave us new ideas around product development. It gave us a lab to stress test a number of ideas very quickly in front of decision-makers and people who knew what they were talking about.”

Like other fairly new companies, he said, Narrative Science can face difficulty in finding the right people in a large distributed organization like a global investment bank.

“The Lab helped us identify the right people to give us feedback and potentially work with us. It opened relationships for us and accelerated discussions which put us in position to do pilots.” He hopes some large commercial deals will come out of it, he added.

Mutual fund companies could use Narrative Science for their reports and marketing materials that summarize the latest performance information. Insurance companies that want to improve their reporting around agent sales performance might also find it useful.

Frankel sees great opportunity beyond financial services as well, because every industry is faced with exponentially growing amounts of data.

“Imagine the amount of time and money companies are spending taking data, trying to get a couple of interesting things out of that data, and then putting it into PowerPoint or Word. It takes an extraordinary amount of effort.

“The demand and breadth of the opportunity is bigger than we had anticipated,” Frankel added. “We knew we were onto something, knew people people were struggling with the amount of data that they had, but as we talked to these firms, especially large organizations, we realized there is an unbelievable amount of opportunity, even with functional solutions like helping them report on HR activities or customer activity, or  creating marketing communications and compliance documents.”

The company’s technology has been around for three years and was incubated at Northwestern’s schools of engineering computer science and Medill School of Journalism, he said.

“The idea was to automate the creation of long-tail content such as coverage of high school sports or Little League. You have millions of these events with tiny rabid audiences who are interested in the stories but the economics don’t work.”

Much the same applies to small companies in earnings season, he added.

“We do earnings previews for Forbes. A few days before a public company is going to announce, we generate a story. We have taken concepts from journalism, such as the concept of angles, to find out what the story is. If it is a financial news story on a publicly traded firm, we might say it is going to beat earnings or fall short of expectations.” The company has two  patents — issued in just two years– and another 10 patents pending on the way it uses data to create a news story, he added.

For financial services, Narrative Science worked with content architects including writers, editors and financial analysts to configure the system.

“As we get into new domains and new reporting types, there is configuration work up front, but once we have done that, it is a relatively painless process to launch a new customer or new report.”

Narrative Science can tune its text to fit  particular domain, type of report, type of content and it can even be customized to fit an individual’s voice and style.

資料來源：Forbes