We are thrilled to announce that Unilytics has joined Optimus SBR
Learn more

Posted by Eric

November 3, 2015


Leave Reply

5 Must Dos BEFORE Creating Dashboards

In the world of data visualization, a lot of focus is paid to the visualization tools themselves. Should I use a time-series visual, a scatter plot, or a heat map? What is the best way to visualize survey data? Can I export my dashboard to a PDF? These are perfectly valid questions, but as much, if not more, attention should be spent on what needs to occur BEFORE you even open up your visualization tool. Here are five of the most important pre-visualization considerations we recommend based on our extensive experience with data and visualization.

Identify Core Question

One of the single most important steps you should make for any dashboard is to define what question it needs to answer. What is the core question it is to address? If you don’t know the question, how can you succinctly answer it? Every decision you make in your dashboard should be directly attributable back to that core question. Otherwise, you risk creating a “Frankenstein dashboard.” This is a dashboard that tries to answer a bunch of different questions from a variety of viewpoints and as a result really doesn’t answer any question directly or easily. It’s a Frankenstein that is stitched and bolted together, but really isn’t very functional to any particular group of end users.

By the way… “the question” shouldn’t be a large, endless diatribe consisting of many sub-questions and corollaries for investigation. Refine the question down to a single sentence and you’ll be doing well.

In addition to knowing the core question you’re going to answer, you should know who your dashboard is intended for. If it’s for the executive team, you need to keep it simple and not overwhelm them with unnecessary sliders, widgets and moving things. As useful as a box-and-whisker plot is, will a busy executive have the time for you to explain how to interpret it? If your audience is analysts that might be OK.

If your dashboard audience is analysts it’s OK to provide additional refinement or corollary investigative elements such as drill downs, drill-throughs, or cross references. They have the time and skill set to follow your investigative interactions. Keep in mind, it’s sometimes more difficult to create a single, succinct dashboard that will be effective for your audience than it is to create a Frankenstein dashboard attempting to be everything to everybody. With your audience, you need to be aware of how much time they have available to interpret your dashboard, their technical capacity for self-service investigation within your dashboard, their familiarity with statistics, etc.

Verify Data Integrity

Make sure the data you’re using is properly prepared for visualization. You only have so many chances to “restate” your findings before people lose trust in your ability to produce dashboards. A correction here or there isn’t usually a problem, but after a few of these, your audience will go elsewhere for their answers. If you aren’t sure about your source data’s accuracy, find the subject matter expert (SME) for the data you’re using and verify your assumptions up front.

Remember… garbage in, garbage out. Part of this means validating your calculations against a source system. Too many people skip this step because they believe the calculations are simple, but there are few things that are so simple they can’t be messed up when you filter, sort, or aggregate your data in a dashboard.

Not only should you make sure your source data is accurate, but it should be appropriately structured. Many people spend huge amounts of unnecessary time working around incorrectly structured data. Have you tried to visualize pivot tables or crosstab data? You end up with really long formulas that require way more typing than is necessary. A quick reshaping and you can cut your effort by days.

One of the most common examples of this is survey data. Every response is a row, and every question is a column with individual responses in each intersection. If you don’t reshape your data and pivot the survey questions into rows you’ll end up writing horribly long formulas to calculate average response scores, and you’ll incur considerable long-term maintenance costs when survey questions change. A simple reshaping of your data can save hundreds or thousands of dollars in ongoing maintenance of your dashboard.

Automate Data Acquisition

Every human interaction, manual process, or manual correction in your reporting process introduces the possibility of erroneous data and incorrect results. Every time you change or transform your data you need to validate that the change resulted in consistent and predictable changes in your data. Any manual steps are even more risky and require even further vetting.

When you automate the process of gathering data for your dashboard, you have more time available to analyze the resulting information. Our clients commonly tell us that their report analysts spend 90% of their time building reports and only 10% of their time analyzing the results. Is this really the best use of a qualified analyst’s time? Will a high producing report writer benefit your organization more than an analyst who can spot trends before they become problematic? An analyst providing critical input into your corporate strategic decisions is likely much better to the organization’s bottom line.

Automate… automate… automate! A dependable, repeatable, timely reporting process will go a long way toward alleviating organizational reporting pains all the way from analysts to executives.

Establish Consistency

Use a structured, standardized process to take your dashboard from conception through release. Make it a dependable, repeatable, timely process… this applies to the construction of your dashboard as well as the acquisition and restructuring of your data. If you don’t follow a specific process to guide your dashboard building, it’s more likely important steps or information will be missed.

We follow our exclusive D³ – Dashboard Design & Development methodology. This guides all of our consultants in a similar direction, regardless of the data, the client, or the visuals. It steps them all the way from conception, through visualization, and on to finalization… with specific requirements gathered and explicit deliverables at each step along the way.

Many of the points presented in this document are covered in our D3 methodology, including identifying the audience, how to ascertain the main question, and the interactive theme. It also covers defining the primary and supporting visuals and their layout. It defines the interactive elements and, very importantly, saving the “perfecting” tweaks (e.g., colour palettes, alignment, tooltips, etc.) until late in your process. Whether you use a methodology like ours, or create one of your own, document it, learn it like the back of your hand, and adapt it as your needs change and your level of sophistication increases.

Apply Visual Cognitive Principles

Plan ahead to ensure your dashboard design takes advantage of visual best practices. Did you know that there is more to dashboards than pretty colours, shapes, and data? Decades of empirical science have resulted in a well-defined, well-understood hierarchy of visualization principles. Proper use of these principles facilitates rapid interpretation of the information in your reports and dashboards. The faster your users interpret your visuals the more productive they can be and the more self-servicing they can do; an extra perk here is you can spend less time explaining the results!

Visual cognitive principles include such things as alignment, containment, position, size, colour, continuity, proximity, etc. These principles are not based on some product manager’s personal preference for small circles or blue lines, they are based on the science of magnetic resonance imaging (MRI) and blood flow in the brain. There are real numbers behind these principles.

For example:

  • When comparing two measures, the brain can more easily understand the similarity or contrast of the measures using position
  • Did you know that the human brain interprets lines as continuity? This means that using a line chart to represent non-continuous values such as products or other categorical data is not a good idea.
  • Did you know that pie charts are one of the worst possible visuals because the human brain is terrible at estimating the area of circular objects? Do you remember A = π r2 from high school geometry? Do you remember how to convert that formula for just one sector of a circle? Neither will your end users.
  • Pie charts also make it difficult to estimate the difference between slices of pie that aren’t next to each other. This visual will only slow down your end user’s productivity.
  • Proper use of visualization principles will help you make sure you don’t make these type of mistakes. We have a blog on The Principles of Data Visualization… check it out to learn much more detail on this topic.


There they are… five MUST DOS to consider before you even load up data in your dashboard tool. If you wait and try to do these after you’ve started your dashboard, you’re more likely to generate incorrect information, hinder your end user’s ability to understand the data, or spend far more time than necessary building and maintaining your dashboard.

You don’t have to get it right the first time, but you should get better every time you deliver a dashboard, especially if you consider everything that should happen before you even sit down and start. These five things aren’t a complete list of everything to consider, but they are very important and a great start.


We offer Tableau Training sessions built on the D3 methodology to incorporate best practices for performance, visualization, and dissemination of Tableau dashboards.

Leave a Reply

Your email address will not be published. Required fields are marked *

Explore Posts By Category



Want to know more?

Contact us