As more companies look for easier ways to solve the Big Data puzzle, new platforms like Quill — from Narrative Science, Inc. — offer innovative solutions. The company’s primary mission is to “give people a fast and simple way to understand data.” That means democratizing data with a fresh and grounded perspective on analytics, and solving for specific use cases to take Big Data solutions that last mile.

SiliconANGLE recently spoke to Stuart Frankel, CEO of Chicago-based Narrative Science, about how the company helps businesses go beyond just reporting the numbers and find narratives that effectively communicate insights buried in Big Data, and how his company can provide a realistic set of expectations beyond the hype. Prior to co-founding Narrative Science, Frankel was CEO of the Performics division of DoubleClick, as well as an SVP at DoubleClick.

Configurable tech that’s highly flexible

Q: Is asking Quill to answer a question as easy as receiving a descriptive narrative answer?

Stuart Frankel of Narrative Science

Frankel: Quill is not a Q&A system. The user doesn’t necessarily ask a question of Quill to get a report. The tech is configured for a specific use case or report type. If there’s a sales report, Quill is configured to generate on a weekly basis. The user experience is accessing Quill to build that report for the first time, and quill generates the report weekly on its own.

Quill is not a needle-in-the-haystack tool — such tools tend to create more structured data, and we love that because Quill can use that data. But you have to know what you want to use Quill for before you use it.

Much of the efforts in the industry have been around getting analytics under one roof, but these solutions haven’t been sufficient. Quill is that last mile.

Q: What aspects of Quill are manual? Establishing business rules? Solving for outliers?

Frankel: Quill’s a configurable tech and highly flexible. With that flexibility comes the need to configure the tech to meet the needs of both the general use case and within specific client relationships. That deployment period can take a couple of weeks or a couple of months, depending on several things.

Is it a product we’ve developed (i.e., customized)? People at our company can manage those configurations, or we can work with the client to do so. We have a quality control process.

Q: How did you go about creating an AI that’s different from other big data solutions through narratives?

Frankel: It incubated at Northwestern University from years of research by my two co-founders, Kris Hammond and Larry Birnbaum. Their background is in AI, and there was a real rooting of domain knowledge in building these types of systems.

Quill is a system driven by the communication goal — looking at the types of analysis you’d do that’s driven by the person doing the reporting. If you’re reporting on a customer’s investment portfolio, the communication you’ll want to develop is around portfolio performance against a goal, benchmarks, peer groups and based on that report. You then drive backwards to determine the analytics and data sources. That’s something really unique — we have lots of IP associated with Narrative Science (patents), and many deal with this concept of establishing a communication goal.

Built to scale

Q: How does Quill compare to IBM Watson?

Frankel: They’re very different technologies. Watson is a discovery tool that finds correlations. Watson ultimately gives an answer, but not a description.

Q: How does Quill scale? What were the challenges here? Other language support?

Frankel: Today our customers use Quill for English-only reports, but the system is easily configurable to translate. While that’s not a focus right now, we expect to have built-in options by Q1 or Q2 of next year. The bulk of Quill’s narrative is understanding what to say, not how to say it. Whether that’s American or British English, those tweaks are relatively trivial. When you move to other languages, there will be aspects of Quill that will need to be modified.

In terms of scaling the tech, it was built to scale. If you’ve got five reports to generate in a month, you don’t need Quill. You can do that. If you have 50,000 reports per month, that requires scale — not as an afterthought or atop infrastructure. There must be a focus on scale from the very beginning. From day one, our first customer was a media co. generating baseball stories based on stats. In 2010, we could run an unlimited number of games data in near real time, and that was by design.

Q: On the one hand, you’re disillusioning Big Data’s hype with this grounded perspective on analytics. How can you provide a realistic vision of the future where automated services may take over human jobs?

Frankel: I want to talk about democratizing data. We see this in government all the time. Look at the data that’s available from cities and municipalities — big trends around data transparencies. The more data you give to regular citizens without the time to analyze it, you’re not democratizing it at all. It’s a meritocracy. Whether that’s government or in the enterprise organization, more data doesn’t necessarily mean you’re being more transparent.

Quill exposes data coming from data, but only gives them the info they need. It gives them a report of what they need to know. How is the salesforce performing, communicating to individual customers, etc.? The idea is that you’re taking away activities or automating activities that do not add value — taking time away from highly valuable and intelligent people. That’s a positive trend. Going back to my financial portfolio example, you have managers who should be thinking about fund investments, understanding how macro and micro trends impact portfolios, instead of spending so much time creating reports.

Easing data governance concerns

Q: How does Quill handle data governance?

Frankel: One powerful thing about Quill is it puts the customer in control of data output. One concern with data governance is who has access to data, and that varies by organization. And, more importantly, what are users doing with that data and how are they communicating around it? We normalize communication.

Imagine a brokerage with several financial advisers. They’re communicating individually with customers and have access to lots of data. Macro-data across customer base, individual customer data and from other sources. How do they use that data to communicate with the customer? They’ll bring their own style and opinions to that, which can create compliance risk. You’ve got thousands of people communicating in disparate ways. Quill can normalize that communication where an organization is able to communicate in a normalized way, even if it’s multiple users communicating that data.

photo credit: kevin dooley via photopin cc