Before start exploring the deep cliffs of the quantitative marketing analysis, I would just like to share my personal set of best practices I always use for an effective dashboard design process. Altough these are focused to a typical marketing use case, these are totally generic as they provide a set of analytical starting points and some cooking recipes on how to approach the task of designing a visual reporting environment for your company.

Tip #1: Goals and Readers are your friends

You probably defined business goals during the data modeling phase (at least) and used them to decide which data to collect. Now it’s time to start over again defining how you and your readers should drive through the data during the analysis. These are what I call analytical paths and will help you for disposing dashboard widgets in the canvas and to make further adjustments over then.

And don’t forget about the readers. Dashboards are made for a purpose: to spread out a message.

But there’re different classes of messages.

Even if the intended reader is just you, think about what’s the main purpose of the dashboard with questions like regarding the readers:

  • Is your dashboard exploratory or explanatory?
  • Does it must be neutral or it could be biased?
  • Are you trying to communicate a message or to convince about a thesis?

Tip #2: Less is more

As Einstein used to say: “Everything should be made as simple as possible, but not simpler”.

Ask yourself which data are really useful for the cause. Data visualization always implies to minimize exposed data, when you’re going to put the smallest amount of data needed to spread out your message and no more. Remember than more data you put on the canvas, more time the reader needs to decode it. And this means more probability she will get bored before being able to catch the important message.

Don’t increase the entropy of the Universe with too many widgets. If you cannot figure out a possible outcome from a widget, it’s not worth the pain to put it on to the dashboard.

Think about your dashboard as an entry point, the tip of the iceberg, and don’t oversize it. This often means you will end up stripping off some dimensions of data so carefully crafted during data modeling. Sad but true, it’s definitively the norm.

Tip #3: Stuck things together

I usually prefer to not spread the payload across tons of different tabs. When performing a marketing analysis, you probably want several degrees of freedom. You may want to traverse data top-down (such as, starting from the market space, then drill down to subcommunities). Or you may prefer to have parallel swimlanes (such as, jumping between attributes and products). This is typical to freeform analysis scenarios. And it’s cool, if you can control it.

Keep your tab hierarchy simple. As a rule of thumb, if you have three main axis in your data cube and you end up with more than three tabs, perhaps you’re oversizing the probe.

Draw your tabs structure to let you navigate only a subset of the possible data axis, and never too much broadly (ie, to drive top-down or swimlanes, but not both). Parameters will do the rest.

Tip #4: Parameters are your friends

Use in-tab interactive elements to filter, slice and project your data, so you can fine-tune your analysis without leaving your eyes from the screen. Tableau is very smart, here, as it offers a complete set of elements and parameters types to embed into dashboards. You can add tons of interactive stuffs, and actions, and macros.

But don’t abuse. They are here for tuning. If you end up using parameters as complete context-switching buttons, you probably need to blueprint your analytical paths and your tab structure again.

Tip #5: pen-and-paper mock-ups first

This probably sounds weird. There are tons of cool web and desktop applications for making digital mock-ups. So what, do I really need to pen-and-paper the mockup, first?

The short answer is: yes.

The long answer is: hell, yes!

You’re not graphic designer, although dashboard are often cool to see. Your intent is to discover hidden patterns on data and you need empathy. Hand-brain coordination helps a lot, here. If you don’t need a collaborative environment , don’t waste time with drawing software (choosing, buying, downloading, installing, learning, mastering…), use colored pencils and paper and start drawing.

You still would probably make a vector mock-up of the dashboard and build a good-looking kiddo. But not now, not really at design stage.

Bonus Tip: Be lean, iterate and innovate!

Designing a data visualization concept is an iterative and never-ending task. Be simple at first, design an experiment, then collect feedback (from your experience, from your users, from your colleagues) and go blueprinting again, over and over. Iterating is the key. Innovation here is crucial, as you won’t spot important topological patterns without. So don’t be afraid to experiment and be creative! You’ll get results…sooner or later!

Have a good design and see you next!

Share This