Most people think about data dashboard design in purely visual terms, but how you lay out a dashboard can also...
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
have major functional implications.
That was a lesson Coca-Cola Bottling Co. recently learned. The independent beverage bottler had built a single dashboard that gave everyone from upper-level executives down to front-line sales workers a single view of company performance. Executives wanted everyone on the same page, but what they got were slow load times and crashing reports.
"We pretty much had a nightmare," said Kevin King, director of reporting and analytics at Coca-Cola Bottling, based in Charlotte, N.C., in a presentation at Tableau Conference 2016.
The company originally brought in Tableau in 2008 and had some initial success. But, by 2014, the dashboard was taking between 30 seconds and three minutes to load, making it unusable by sales teams and executives alike. King said he and his team initially investigated server performance, but found they had more than enough CPU and memory to execute their dashboard. The problem wasn't technical. It all came back to the data dashboard design.
The primary dashboard was pulling in upward of eight metrics, each built on independent queries that accessed data from separate data stores, including an enterprise data warehouse, ERP and customer relationship management systems. The design of the dashboard basically caused it to replicate these data stores in the Tableau server every day. It was choking on too much data.
Not only was there too much data, but there were too many extraneous elements in the dashboard -- things like boarders, headers and logos, each of which took time to load. It was time to clean out the garbage.
"The first piece of trash we had to get rid of was the irrelevant data," said Alejandra Ospina, senior business analyst at Coca-Cola Bottling.
Ospina led the effort to restructure dashboard data queries so they pulled in only the data they needed. This brought the volume down from 650,000 rows of data to 15,000, a 96% reduction in data overall and a significant improvement in dashboard load times. She also pulled out many of the unnecessary visual elements that were slowing down performance.
King said the experience shows it's important for analytics teams to lead data dashboard design projects, rather than building whatever business teams want. It's always important to build support for analytics projects at the executive level, and teams need to deliver functionality that executives require, but that doesn't have to mean running afoul of design best practices.
Part of the redesign process included conducting focus groups with different end users to learn more about how they would like to use the tool and building custom dashboards that are more tailored for each job role.
King and his team also had to explain to executives why the single dashboard they asked for initially wasn't an option, a potentially fraught situation. But under the circumstances, the discussion wasn't as hard it might have been. "The fact that we had a failure meant it was easier to explain why we needed to make changes," he said.
Everything you need to know about dashboard design best practices
Eliminate the clutter to improve your data dashboard designs
Effective data dashboards don't have to be too fancy