cornball photo of man smelling sour milk and overreacting
PHOTO: shutterstock

When was the last time you updated your analytics dashboard? Too often, we think a dashboard needs very little adjustment over the course of its use. The truth is an analytic dashboard's value can expire over time like milk, with bad data being that sour milk. 

But without an expiration date or an off-smell, what are the indicators a dashboard needs replacement?    

5 Signs Its Time to Refresh Your Analytics Dashboard

Here are a few signs a dashboard is reaching its sell-by date, along with some suggestions on how to address those issues.

1. Is the dashboard too busy for the screen?

Screen size significantly affects the space available for graphs and navigation bar. The most important visuals should be front and center, not buried behind menus. Changing device usage, spurred by the bring your own device trend, can render a dashboard untouched by stakeholders. It can be especially hard to tell if remote workers share the same access frustrations.

2. Are comparison and differences among data visually relevant?

A change in metrics values can reflect a desired activity — such as a rapid increase of conversion traffic during the early days of a marketing campaign — and then just as suddenly show no significant changes. Standard issues like ad fatigue can be the culprit, but they can be hard to spot if the visuals skew comparisons or are imprecise in displaying changes.  

The image is an example of a low-precision correlation. It is a correlations of regression variables in a matrix. One side shows colored circles, while the other half shows more precise numbers corresponding to the circles. The circles can give a visual impression of which variables may be positively or negatively correlated. The numbers give further precision to understand the degree of correlation. Both kinds of visuals are valid, but the information they can convey differs.

poor visualization representations

3. Does the dashboard visuals still spark ideas?

In some instances a simple table is sufficient, but tables can hide relationship details. The right dashboard visuals should reveal a deeper relationship amongst the data within the first few seconds someone looks at the graph. A graph that was selected for a set of data can become the wrong graph if the data has changed over time.

4. Is there a mismatch between user roles and visualizations?

Are the dashboard graphs covering too many metrics without a clear action? That may be why no one is responding to it. Having the same dashboard used by everyone in your organization may seem like a noble collaborative objective, but likely, every manager who is interested in the metrics can have different agenda, with different metrics providing value to their department.

5. Are there unverified technical changes to data sources?

Programming architecture can create technical inaccuracies in data that supports the metrics being visualized. Web analytics installed with a bad tag installed on a website can double-count session visits, time, and other metrics. An API delivering data can be altered, changing the graph as a result. Sustaining these errors can erode manager trust in a dashboard, encouraging those ignoring the graphs to continue to do so.

Related Article: Telling Stories Through Data Visualization

4 Steps to Improved Dashboards

The steps to take to improve or replace a dashboard can include the following:

1. Reevaluate the outcomes desired with the metrics

The real value of a given metric comes in the form of the questions it raises or answers it provides. It may seem like basic advice to change them, but metrics should initiate good discussions, or conjure the risks being overlooked if they are examined periodically.

Metrics should be tailored to the audience who will best respond to the outcomes from the analysis. Solicit user feedback to improve the metrics and visualization being selected.

2. Remove graphs and charts that obscure report precision

The rule of thumb for choosing charts is based on how specific you want to explain data. A histogram is useful if you are not interested in explaining outliers, but a boxplot would be best if identifying outliers among observations is necessary for stakeholders. When you create a graph consider how readable the data and its composition is: does the visualization displays accurate indicators that relate to what the user should immediately see?

3. Evaluate the martech used for your dashboard

Be prepared to evaluate capabilities from multiple vendors to select the best dashboard functionality. Keeping pace with the innovation and new capabilities available can help your organization avoid technical debt. A key test is being able to segment users and deploy corresponding reports based on each segment's use cases and required capabilities. 

Related Article: How Data Visualization Tools Are Making Self-Service Analytics Easier

4. Set a trial period to evaluate changes, addressing one feature at a time  

Periodic reviews of how well dashboards are impacting data can help you focus on where further refinements are needed in code and APIs changes, as well as dashboard refinements. Starting with one change at a time makes debugging dashboard technical features easier, as well as managing progress to raise stakeholder interest. Use the evaluations as an opportunity to also ensure data security protocols are being followed.

The point of a dashboard is to provide a snapshot of digital marketing campaign performance for each user. The viewpoints of that data are dynamic, influenced by changing business and technical demands. Evaluating those dynamics against the intended actions from dashboard reports can keep a dashboard and its metrics from going stale.