By Henry H. Eckerson and Wayne W. Eckerson
It is often said that a picture is worth a thousand words. But in the era of big data, a paragraph from a natural language generation (NLG) tool might be worth a thousand pictures.
NLG tools automatically analyze data, interpret it, identify the most significant parts, and generate written reports in plain English. In essence, NLG brings artificial intelligence to business intelligence (BI), automating routine analysis, saving business users time and money.
Although BI products generate visualizations, reports, and dashboards, business users still have to analyze and interpret data. That’s where NLG comes in. It automatically performs the analysis and generates an English language translation of what is significant and meaningful in the data. Business users no longer have to study the data to interpret its meaning; NLG tools do that for them.
Moreover, NLG tools bypass the need to create visualizations, charts, and reports in the first place. The tools can sift through large volumes of data and generate reports automatically. This is particularly valuable in the age of big data where huge amounts of data can overwhelm business users and IT departments alike. With NLG tools, data analysts can spend 80% of their time analyzing data rather than 80% of their time preparing data. In other words, NLG tools augment the job of business users so they can focus more on high-value tasks and less on menial work.
For example, Dominion Dealer Solutions use a NLG tool from Narrative Science to generate unique vehicle descriptions based on data gathered from a variety of sources, including automotive reviews, Kelley Blue Book, and, CARFAX. A study revealed that cars with NLG-generated descriptions sell an average of twenty days earlier than those without one.
Since NLG tools automate analysis and have human-like capabilities, some people worry that NLG and other artificial intelligence products will undermine job security. Although this is a legitimate concern, officials at all three companies I interviewed emphasize that NLG tools augment the job of business users, and don’t replace them. They say NLG tools allow business users to focus on higher value tasks. Although NLG tools generate written analysis, business users still have to read, understand, and act on the reports. Rather than replacing analysts, these tools allow for more analysis to happen more often.
However, this explanation is incomplete. What if an NLG tool automates reports that usually take a dozen people a month to write? In this case, the people who are paid to prepare those reports will need to be shifted to another role in the business, if one is available, or laid off. For example, Forbes uses Narrative Science as a cost-effective way to generate content and expand market coverage. Instead of hiring people, Forbes chose Narrative Science to generate the content.
Although NLG technology is quite new, there are already a number of vendors selling NLG products, including Yseop (pronounced “easy-op”), Narrative Science, and Arria. (Another early player, BeyondCore, was recently purchased by Salesforce.com for an undisclosed amount.) Each offers an NLG tool that automatically creates reports and analysis from data sets and existing reports. They use inference and natural language generation engines to identify the most significant parts of the data, draw insights, and recommend actions that get embedded in automatically generated written reports.
Because they automate analysis, most NLG vendors are forging partnerships with business intelligence vendors who want to embed NLG capabilities into their products. The partnerships provide BI vendors with NLG capabilities they don’t have and likely won’t build, and NLG providers get new channels and revenue streams for their technology. Yseop’s Savvy, for example, has plug ins for Excel and Qlik.
Here is an overview of three major NLG products today:
Based in Chicago, Narrative Science sells Quill, an NLG product favored by financial services companies, although the product can be used by companies in any industry. “If there is structured data and there needs to be something communicated from that data, it’s a use case for Quill”, says Mary Grace Glascott, Director of Product Marketing at Narrative Science.
For example, a large financial services company uses Quill to provide automated feedback to call center representatives. Quill analyzes call detail data and automatically generates a personalized performance report written in English for thousands of customer service representatives, something that is difficult to do with traditional BI tools.
Initially, Quill was only available in the cloud, requiring users to send data out and receive analysis back. However, the product now also runs on-premises. Narrative Science also sells extensions that enable Quill to run within major BI platforms, including Qlik Sense, Tableau, SAP, BusinessObjects, Lumura, and Microsoft PowerBI..
Based in Dallas, Texas, Yseop sells Savvy, an NLG tool which plugs into any data warehouse or database. Savvy specializes in helping companies leverage their CRM data to generate reports and narratives that make it easy for sales people to cross-sell and upsell products.
When using Savvy with a CRM application, such as Salesforce.com, business users select the account they want to analyze, and then click on an Yseop tab to see a written report. Matt Rauscher, vice president of Yseop, says, “Savvy takes data from a CRM application and its rules engine automatically decides, based on the data, what products a salesperson should sell to which customers, and then the NLG tool writes what they need to do and why”.
Unlike other tools, Savvy can currently write in multiple languages, which is a requirement for global enterprise companies. Also, Yseop positions Savvy as a development platform for building NLG-based applications.
Arria’s NLG Platform and Recount
Arria offers NLG Platform, which works with almost any data source or application. NLG Platform runs on the cloud or on-premises and works in any industry, including financial services, healthcare, and marketing. Soon, Arria will release two cloud versions of the product: an enterprise version called Articulator Pro and Articulator Lite geared to non-NLG programmers that will compete with Yseop Savvy.
Arria also sells Recount, an accounting solution that works with accounting software from Xero. Recount automatically generates reports that identify key trends and issues in their accounting data that they need to pay attention to. The tool helps small and medium-sized business owners focus on the business rather than administrative duties and bookkeeping.
According to Jeff Zie, CMXO and Head of Recount, “Xero is a fantastic accounting platform, but it doesn’t tell owners what’s going on and whether it’s a good thing or a bad thing. Using Recount is like picking up a phone to a financial analyst and asking ‘what’s going on with my business?’”
So, What’s The Takeaway?
NLG tools automate analysis, taking the capabilities of BI tools to the next level. Rather than generate charts and tables, NLG tools interpret the data and generate analysis in written form that tells business users what’s significant to know. The tools perform routine analysis of predefined data sets, eliminating both the manual labor required to generate reports and the skilled labor required to analyze and interpret the results.
Although NLG vendors say their tools augment, not replace, jobs of report writers and analysts, in some cases, NLG tools will reduce the number of people required to generate and analyze data. For business owners, this is great news: they can reassign staff to other jobs, increase their productivity, or lay them off.
We expect a multitude of NLG solutions to hit the market in 2017. We also expect leading BI vendors to integrate NLG functionality into their products.
Bio: Henry H. Eckerson is Research Analyst at Eckerson Group and Wayne W. Eckerson is founder and Principal Consultant at Eckerson Group.
- Putting Alexa to Work: Moving Conversational UI from Hype to Reality
- Introduction to Natural Language Processing, Part 1: Lexical Units
- Natural Language Processing Key Terms, Explained