See below four best practices for organizations interested in adding NLG to their dashboards.
One of the biggest selling points for ongoing big data and analytics efforts has been the dashboard. Dashboards are great ways to summarize information in an eyeshot for time-strapped executives and managers.
The caveat, of course, is that dashboard users must fully understand what their dashboards tell them.
Sometimes, understanding a dashboard isn't so simple. A high-level summary of quarterly sales by region might make sense to a sales manager, but a sales lead in charge of a district within a region might prefer a dashboard that displays results within the context of a single region and its districts. That manager might want to know how his district compares to other districts in his region before he wants to view region versus region performance in a broader picture.
SEE: Hiring kit: Microsoft Power BI developer (Tech Pro Research)
These types of differences in information needs have driven companies to do two things with their analytics dashboards: Summarize dashboard visuals with accompanying narrative; tailor each dashboard to the job responsibilities and information needs of the individuals using the dashboard.
"The goal is to make information understandable," said Mac Zionts, CEO of Automated Insights, which provides natural language generation solutions. Natural language generation (NLG) is a technology that transforms data into clear, human-sounding narratives. "In the big data and analytics movement, companies have tried to take raw data and turn it into actionable and meaningful dashboard insights. Many users can understand and act on this data when it is presented in a visual manner. The problem is, not everybody can.
"When users don't understand the dashboards that they use, the next step is usually to call in a business analyst to help them work through the dashboard so its meaning can be unlocked," Zionts continued. "This means that the business analyst must take the time to explain the dashboard."
SEE: Quick glossary: Business intelligence and analytics (Tech Pro Research)
How NLG works with dashboards
An alternative solution is using NLG alongside a dashboard. That way a narrative can be posted alongside a dashboard visual to summarize the information presented. This helps the user become more self-sufficient in dashboard interpretation--and saves the business analyst's time.
When a business analyst designs a dashboard, the analyst creates a visual representation and written narrative that posts alongside the visual to explain to the user what the visual summarizes.
These narratives (and dashboards) can be modified by role. For instance, you might have a worldwide sales executive who wants to know everything about everything but a regional manager who only is interested in his or her sales region. The analyst can re-purpose dashboards and narratives to fit the information needs of various roles within the company.
"When you post data in a specific contextual narrative that goes along with the visual information, dashboard understanding is improved among users," said Zionts.
The data can also be understood on several different levels:
- Descriptive (describes what you're looking at);
- Diagnostic (presents conclusions or actions about the data);
- Predictive (projects future trends and events based on the data); and
- Prescriptive (recommends actions that should be taken).
SEE: Hiring kit: Data architect (Tech Pro Research)
NLG best practices
Below are four best practices for organizations interested in adding NLG to their dashboards.
1. Understand present goals and pain points
Users might not always be forthcoming when they are confused by dashboards--but a business analyst and or IT can tell pretty quickly which dashboards are being used regularly, and which are not. The goal in this exercise is to identify dashboards that aren't being used and find out why. If the problem is based on usefulness or understanding, the addition of a written narrative that displays alongside the dashboard visual might solve the issue.
2. Start with your most widely used dashboards
The sooner you can use natural language to supplement your most popular dashboards, the sooner that many different users and departments can use these dashboards more productively.
3. Consider sales and marketing as one of your first users
For many companies, deployments of NLG begin in marketing and sales. Sales and marketing users are natural fits for dashboards since it can instantly inform those users about revenues--and written narrative summaries of the dashboards will help even more.
4. Look for an NLG tool that easily integrates with existing analytics software base
A near "plug in" or an easily integratable solution will work best. This way focus can remain on expanding dashboards with narrative summaries--not on system integration.
- How to select dashboards that improve KPI tracking and increase project success rates (TechRepublic)
- How to create your first Microsoft Power BI dashboard (TechRepublic)
- Conversational AI goes mainstream at Capital One bank (ZDNet)
- MicroStrategy 2019 introduces "HyperIntelligence" contextual BI (ZDNet)
- Big data: More must-read coverage (TechRepublic on Flipboard)
- Why dashboard design is critical to analytics success (TechRepublic)