Updated: Jan 15, 2018
Analytics UX by keeping the user in mind to ensure optimal BI adoption
User experience is all around us, from the most advanced and sophisticated mobile apps we use every day, to the news paper we read, subway stations and our utility bills. Designers, engineers and developers from all industries have put our experience and the way we consume any material as one of the main parts of development and implementation. Same goes for analytics.
User experience is all around us, from the most advanced and sophisticated mobile apps we use every day, to the news paper we read, subway stations and our utility bills.
We must suit the user experience according to several varied parameters, such as the persona at hand, the actual investigation required, the meaning of the derived decision to be made and more. This article will describe several things we can de to ensure highly efficient user experience ultimately leading to optimal BI tool adoption.
Diving into the deep end
Many times when we start working with a plug and play BI tool, we are often tempted to dive right into the deep end. After extensive market research, multiple tool evaluation, long sale process and a thorough POC we tend to get ‘Analytically impatient’ and just start forcing some immediate KPI’s and metrics which may seem obvious to begin with, but can cause clutter and inefficient BI project management. Several things can be done to ensure successful BI project implementation, such as implementing with our agile and iterative BI methodology, here we will discuss how to design our dashboards and reports from a user experience point of view to ensure optimal adoption.
1. Always keep the end user in mind
Before even starting to connect data sources and pulling in calculations, we must understand who the end user will eventually be of this dashboard, what they will need it for, and how exactly they will be utilising it. The first two items are usually pre-determined, but many times, we as implementers can determine, or at least assist in determining how the specific piece of analysis will and should be used. This piece of investigation and requirement gathering should be concluded during pre-planning stages, as seen in our implementation methodology and is crucial for efficient implementation and eventually optimal adoption.
Almost every piece of an analytical tool, dashboard or report should be custom tailored for the end user and their requirements.
To achieve this we should understand the use case, at times better than the actual user. Put yourself in their shoes, what will they need to be able to achieve with this dashboard? How will he make knowledgeable decisions? This too should be a step in the iterative feedback cycle, aiming at improving the user experience with every step.
2. Design Guidelines
Many times, without proper planning or creation of simple dashboard mockups, we can find ourselves with overloaded, clunky and ill-performance dashboards and reports. In our iterative implementation methodology, we should put in some effort into actually planning and creating simple mockups to regard the actually lay out and flow of a dashboard. This is the time to start thinking of dashboard UX, as the better we plan, the higher potential success rate we will achieve. There are a few thumb rules we can always keep in mind:
7 +- 2 - Always have on average 7 visualizations or widgets on each dashboard. Beware of clutter, this will lead to abandonment due to over load and performance issues. Less than 5 widgets don’t really justify a whole dashboard, while more than 9 may overload on the user, and should be branched out to additional functions or dashboards.
Top to bottom - Left to right - The same as when we read a book, our eyes will always immediately focus on the top left hand side of the page, and as we progress with our attention, we’ll move to the bottom right. Keep the importance of your widgets the same. The most important KPI should be the simplest, often times even a single value indicator, or a simple 3 segmentation pie chart. As you progress down into the dashboard, provide more detailed KPI’s and visualizations. If the dashboard starts getting more and more detailed, strongly consider branching out to additional views, avoid clutter!
Keep your colors consistent - Usually no more that 3-4 colors are required in order to get an idea across. If you do decide to go with standard color pallets, make sure that the same colors are reflected for the same purposes across the dashboard. i.e. if you use red as a negative indication and green as positive, make sure that’s the consistent approach across, or if one chart shows US sales with a blue line, then the corresponding pie chart should also state the US as blue.
3. Top to bottom approach - The top of the dashboard is the most important area, holding the most important KPI’s and ultimately where the user will first focus his attention. As we progress down the dashboard we will be able to show more analytical pieces which aim to show higher resolution trends and dimensional break downs. The second, or middle area should focus on important details, with more context to the KPI’s above. Start with immediate KPI’s and values, simple, one figure KPI’s, or 2 tops to show an immediate view of the current state of the analysis piece. If it is a Sales dashboard these values should be total sales this Quarter or month, YTD value and compare to previous periods. As we progress down the dashboard we can show how these values are broken down by salesperson, branch, geographic dimension and so on. Remember, the report as a whole should enable the user to get an immediate view of his operations initially, and then enable further slicing & dicing and drilling down in order to be able to further investigate the item at hand.
The top of the following dashboard will breakdown the viewing levels by importance and relevance. The top section will hold the most immediate and important values, the middle section will show these values in more context, while the bottom section will breakdown these values into higher deeper resolutions.
4. Choose the right visualization - Understand the analysis at hand, trends will always need a line or column chart while grouping and segmentation should need pie charts. Geographic analysis would require maps, make them interactive so the user can zoom in and out of countries and states. Then, enable further segmentation and a capability to dive deeper into higher resolution data by providing tables and lists. There are several tools out there to help us determine which chart to use for each kind of analysis, such as Sisense’s Data Visualization Wizard or the following schema by A. Abela, 2006:
5. Advanced features & capabilities, Add-ons & plugins
Using not-out-of-the-box features such as scripts, add-ons or plugins can extremely improve dashboarding UX and give a look and feel which can be much more natural looking, not just for standard design but also actual functionality. Choosing the right color pallet is one thing, but thinking of additional use cases and functionality can always improve usage. Sisense provides several such plugins which enable loading widgets on demand, taking advantage of valuable dashboard real-estate and being able to show the whole story with only a few immediate visualizations. Also, remember to take advantage of additional features and capabilities such as advanced sharing and email subscriptions, or email alerts enabling one to be notified immediately if a certain threshold has been passed or if an anomaly has been detected.
Don’t push the user to advanced capabilities if there is no immediate need for them. The most powerful and useful dashboards and visualizations tend to be very simple and straightforward. Over complicating things when there is no requirement for advanced usage will simply create cognitive overload and will almost always lead to abandonment, not only of the tool developed, but also may harm the trust your end users have placed in you. Keep it simple to begin with, then if needed, progress to advanced capabilities.
Remember “Not everything that counts can be counted, and not everything that can be counted counts” (A. Einstein). It is our job as BI developers to enable a clear view of complex and at times non-contextual data to our users, to lead them with their data driven decision making tools.