New Looks for Intranets and Dashboards
Initially, the motion toward personalization may seem opposed to the universal accessibility that dashboards deliver to users of any type of technical orientation—their quintessential use case. Crawford acknowledged the binary nature of these two concerns, which requires careful navigation. “On the one hand, there’s this expectation that these mechanisms become universally intuitive and easy to understand,” Crawford revealed. “But also, I want it to be personalized. I want DoorDash to show me the restaurants that I might like, not the restaurants that all the people in my area might like.”
Such recommendations are a classic machine learning problem, which today is being addressed with a variety of traditional and advanced statistical models. A relatively recent new development for dashboards is their incorporation of measures to evaluate the content, visualizations, or other dimensions of what’s displayed as a feedback mechanism for the dynamic algorithms furnishing suggestions.
This form of AI is also improved depending on which dashboard characteristics a particular user favors. The result is a mutable dashboard able to learn personal preferences and “break the rules as well,” Crawford said. “A good dashboard is great at that. It will persist and be reliable for what it’s serving up to you, based on what it knows you need to see, the context, and you as a user. But it also knows when to say, ‘Here’s some other things that might be of interest to you, even though you haven’t asked for it before.’” With this approach, the same dashboard would yield different results to employees in the same department trying to accomplish the same task. Those differences would be predicated on their respective ways of working and personal preferences.
Data Quality
The inclusion of personalized, dynamic statistical algorithms for recommendations is an interesting facet of contemporary dashboards, because it brings attention to the notion of accuracy. Holistically, the underlying purpose of dashboards (whether or not they’re contained in intranets) has not, and likely will not, change. They’re designed “to serve as a true single pane of glass,” Crawford observed. “That’s what makes them so functional: the aggregation of multiple datapoints into a single, one-stop shop.” Consequently, dashboards place a premium on data quality so that the information they illustrate is timely, correct, and beneficial to end users. Poor data quality has the potential to undermine any dashboard endeavor, regardless of how sumptuous the visualizations, or how much information they show. As with most things related to data, the foundation for successful data quality requires prudent metadata management.
“You’ve got to revisit your metadata quality to ensure you’ve got good data hygiene,” Crawford advised. “Good data management and data hygiene are part of the ongoing maintenance and care of your data transformation initiatives.” Since dashboards provide a visual overview of any type of business metric, the data quality dimension of completeness is particularly germane to proper implementation of these tools.
Factors that may exacerbate this aspect of data quality include anything, from the incorporation of new source data to major changes impacting an organization. During these transitional moments, a good dashboard “needs to account for the full information state,” Crawford said. “It needs to be relevant during times of transformation, such as how can we ensure that all our current data is considered in this database, along with system migrations, or office consolidations, or mergers and acquisitions.”