top of page

Dashboard Development

As part of our various services, or as a separate activity, we can develop powerful digital dashboards to help you monitor and share key data, derived information and insights.  Such dashboards should be set up and promoted as "single source of truth" for all stakeholders, to enable maximum engagement and good decision making.

 

We are experienced with Microsoft Power BI and Excel, and AVEVA's PI-Vision (and PI-AF).  However, any modern data visualisation platform you may have in use will likely offer the features required and we will readily make use of them.​

Process Performance Analysis
What makes a good Dashboard?

 

A good dashboard, on any platform, is normally the front end to a large and complex data set. It visualises key data and derived results - such as KPIs or other metrics - to tell a coherent story.  Below, we give our views of how such a story is best told.

 

The images shown here give good examples. Click on them for a larger view.

​

 

What is involved?

​

Firstly: a good understanding of the data and the story (or stories!) to be told.

​

Secondly: exposure to - and experience with - lots of purpose-built well designed dashboards. Understand how and why other people did things in certain ways. Be familiar with best practices (see below).

​​

Thirdly: familiarity with the relevant platform features, such as data collection and manipulation, number formatting, interactive controls, etc. Modern versions of Excel, combined with Power Pivot, Power Query, and Power BI, don't leave much to be desired for accessing, visualising and analysing complex data sets.

​​

Last, but not least: a keen eye for detail and the realisation that "less can be more". Know what to leave out. Know what's not required and only leads to clutter.

​

​

So... what are those "best practices"?

​

The “public face” of AnaLOSSyst​ are its Power BI dashboards and these have been very carefully constructed along clear guidelines.  Even if there may occasionally be a need to customise due to local circumstances – perhaps an alternative chart or some slightly different wording – these guidelines need to be followed. We share them below, as they may inform others when constructing dashboards of their own; an ever more common task for engineers in industry.  Tools like Power BI and PI-Vision allow you to go wild with options for fonts, colours, line styles and other embellishments, but dashboards need first and foremost to tell a useful story that is easily recognised.​​​​

1. Make your story easy to find

  • A good single dashboard page may contain two or three “stories”, but not nine or ten!

  • Avoid having to scroll the screen to get to important information.  If it doesn’t fit legibly on one page, consider creating a separate page and group the information sensibly.

  • Don’t include charts or other visuals just because you can. If your audience isn’t served by it, leave it out.

  • Maintain a “single source of truth”.  If information is already tracked and presented successfully in another platform, then don’t repeat it at the risk of inconsistency and confusion.

  • Don’t allow good days to hide bad days. In other words, don’t let positive (i.e., real) losses be cancelled out by “negative” ones. Track main KPIs by day or by shift, if not by hour.

2. Avoid ambiguity

  • Always explicitly state units of measure clearly and be consistent with them.  However, to avoid clutter, don’t repeat them unnecessarily either (see below).

  • Use universally symbolic colours in a meaningful and consistent manner.  Red should only indicate a problem or a dangerous limit.  Orange suggests a warning or alert. Green should represent an “all good”.  Use shades of blue, brown, purple, grey, etc. for more neutral information.

  • Make sure any labels and titles state exactly what is presented in as few words as possible. This is not a trivial matter and may require some careful thinking!

  • Add a clear legend to graphs that contain multiple sets of data.

  • When using multiple vertical axes, make sure they can be easily linked to the right data set; for example via the font colour of he axis labels.

3. Minimise clutter

  • The intended message from a dashboard can become hard to decipher when the display is too crowded. Ensure your “canvas” dedicates enough space to its main story. Use drill-down features to only show detail when specifically requested by the user and create additional “pages” to group information meaningfully and in the space it deserves.

  • Use as few (carefully chosen!) words as possible. Don’t repeat axis titles in chart headers and don’t add axis titles unnecessarily.  For example, if a horizontal axis shows 2021, 2022, 2023, …, it probably won’t be necessary to add “Year” as an axis label, especially if the chart title says “yearly” or “annual” already.  And if the chart title states “% margin”, then don’t repeat that as a label for the vertical axis. It all takes up valuable space... Less is more!

  • Don’t show unnecessary detail. If your dashboard platform allows for “tooltip text” then make use of that to only show certain types of information when the user hovers their pointer over it. A good example of this are data historian tag names.

  • For numerical data, show at most 4 significant digits and (normally) at most 2 decimal places. And be consistent for each data item. Exploit prefixes of the metric system: micro, milli, kilo, mega, … !

4. Maximise visibility

  • Draw the eye of the user to the most important information. Make it stand out.

  • Stick to a colour scheme with good contrast.  Don’t use subtle shades of a single colour to distinguish notably different data sets and only use “symbolic” colours where justified (see above). Also play with line thickness or line style to tell things apart.

  • Don’t expect your audience to have perfect vision. Use consistent fonts and font sizes and keep things easily readable. There is no place for “small print” on a dashboard.

  • If a dashboard may have to be printed as a report, make sure you print it off to check ease of reading.

  • Nowadays, many types of visuals are readily available. Carefully pick the ones that make it easiest to interpret your data. For example: a line chart suggests continuity over time, but should not connect data points for different geographic locations.  A bar chart indicates discrete data points, which should ideally be ordered somehow, and may be unsuitable for long-term trends.

  • allows the user to interact with it wherever relevant Interactive features should be intuitive and robust.

bottom of page