Big Data Dashboard Design Lessons

Choosing the right data to collect, then developing algorithms capable of extracting reliable, analytical insights and predictions, is just half the battle.

To be useful, these results must be conveyed to decision-makers in ways they can quickly comprehend, and from which they can take action.

Enter the analytical data dashboard, which graphically represents key performance indicators (the primary role of so-called operational dashboards) while giving business users -- people typically lacking advanced degrees in statistics or predictive modeling -- a way to interact with the model and its parameters.

The current focus on data dashboards of all stripes is understandable, according to observers who say it reflects the sense that companies that can make faster, data-driven decisions will invariably win the competitive race. This also explains the mini explosion of vendors offering off-the-shelf dashboards that can plug into popular platforms like Salesforce.com's CRM.

[ Find new ways to make big data easier to understand. Read Big Data, 'Breaking Bad' And Orange Juice. ]

For about a year, global audio electronics company Plantronics has been using a dashboard from SalesClic to give a small group of salespeople working on top customer accounts a graphical view of their opportunity pipeline.

"Prior to SalesClic, we used [Salesforce.com's] reports and dashboards," Philippe Van de Voort told InformationWeek in a phone call. Van de Voort is corporate sales manager at Plantronics, which develops, manufactures and distributes audio communications equipment for businesses and consumers.

The problem, Van de Voort said, was that valuable insights might be hidden simply because they were spread across multiple screens and reports in Salesforce.com.

SalesClic, which integrates inside Plantronic's Salesforce.com user interface and pulls its data from the CRM platform, also helps the team's manager by suggesting better allocations of sales resources, based on the status and size of all the opportunities in the pipeline.

Indeed, one trend in the evolution of dashboards is bringing action and analysis together, rather than keeping the two functions segregated, SalesClic CEO Thomas Oriol told InformationWeek in a phone call.

"The goal is to present the most-significant insights in the interface that they use for daily tasks," Oriol said. "Stop that segregation of management and analysis."

One driver for this trend, Oriol said, has been the popularity of mobile devices and apps. Limited screen real estate on these devices constantly forces users to evaluate what information is displayed on the first screen, he said.

The SalesClic application, which claims about 100 users worldwide and won Salesforce.com's Customer Choice Award for Analytics last year, can be used standalone, integrated with dashboards from either Salesforce.com or Highrise, or integrated with the Google Analytics' dashboard.

But just as important as these design technologies is a "respect for the user's mental map," as Oriol described it.

He said there were many interviews with prospective users before SalesClic was launched in 2008, and that that process is ongoing.

Recently, for example, users requested a way to see when trends turn positive. In other words, not simply flag risks in the sales-opportunity pipeline, which are represented as red dots on orange bricks. So green dots were added when an opportunity was moving faster or was deemed bigger than expected.

Beyond design is the basic problem of obtaining the right data, and making sure assumptions in the model are correct.

Keith Cooper, now the CEO of Web data monitoring and extraction vendor Connotate, remembers being a senior executive at a company that helped inaugurate the fixed-price, online backup industry.

A dashboard, designed by consultants, helped Cooper and his management team monitor customer segments by media, as well as customer satisfaction and retention rates. "If we could get the best combo of those four things, we would pour millions into marketing" to those channels, Cooper told InformationWeek in a phone call.

Unfortunately, the model failed to note how much backup data these "$50-per-year" customers were sending. Further research indicated the happiest customers were sending terabytes of data. "So we were spending lots of dough attracting very unprofitable customers," Cooper said.

Cooper offers advice to anyone wanting to develop an analytics dashboard, particularly any dashboard that will sweep together internal and external information and make complex assumptions. "Be really clear and specific on the questions you're trying to answer," he said. Once you have that, and you're satisfied the input data is accurate and sufficiently large to be meaningful, "you want to run [the model] for a period of time, to see if the decisions you'd make from that data would be right."

In other words, check the efficacy of the model first before putting it into operation.

Making decisions based on flashy macro trends while ignoring "little data" fundamentals is a recipe for failure. Also in the new, all-digital Blinded By Big Data issue of InformationWeek: How Coke Bottling's CIO manages mobile strategy. (Free registration required.)