The Problem with Dashboards

Eric Sandosham, Ph.D.
4 min readSep 16, 2023
Photo by Chris Liverani on Unsplash

Background

This is the 4th article on my new series on calling out bad thinking and bad practices in data analytics / data science. In my last article, I wrote about the problem with Data Visualisation (see here) as practiced currently, calling out inconsiderations in eye-tracking and cognitive load, and it resonated with many. This article focuses on a topic that is a near-neighbour to data visualisation — (management) Dashboards.

Over the last couple of years, my consulting partner (Sally) and I have been called upon to redesign the management dashboards in a variety of industries and functions. Most of the time, the client thinks it’s just an aesthetic make-over exploiting the functionality of their latest visualisation tool (usually Power BI or Tableau). But I soon realise that the issues run deeper. While many organisations clearly struggle to create visually relevant charts for their data (i.e. poor visual vocabulary), the bigger gap was that they struggle to formulate the right metrics to begin with.

Dashboard Design Principles

There are essentially 2 large areas of considerations when it comes to designing dashboards. The first is about getting the metrics right, while the second is the storytelling aspect that is manifested through the visual design and aesthetics. Many dashboards lack a unifying ‘frame’ (for lack of a better word) to give meaning to the metrics shown. Many organisations simply group similar or logically-related metrics together. Many organisations simply show the standard line charts and column charts to convey the information signals in their metrics.

I have found that it helps to apply the following methodology to dashboard design. It is vital to recognise that dashboards are monitoring tools. Dashboards monitor (a) deviations from targets and baselines, (b) relative differences to peer groups, © trend changes over time, and (d) existing and emerging correlation between metrics. Knowing these 4 objectives of dashboards enable us to anchor it with the use of the right visual vocabulary to ensure that information signals are unambiguous and easily digestible. See my previous article on Data Visualisation.

Getting the Right Metrics

To start off a dashboard design, you need to first determine what you are monitoring; it could be a business, a process, or even a population segment. We then determine from whose perspective you are monitoring, i.e. the specific target viewing audience of your dashboard; e.g. management, operations or finance. Next comes the metrics. Here is where it gets interesting. I have found it particularly useful to ensure that we design metrics that capture the continuum of Input -> Activity -> Output -> Outcome -> Impact. This ‘metrics continuum’ was popularised in the 2011 United Nations (UN) Results-Based Management handbook to drive country-level development.

Here’s an example of how to apply this metrics continuum for the population of banking relationship managers:

  1. Input metrics: number of customers, asset under management (AUM), number of relationship managers.
  2. Activity metrics: number of customer appointments, number of campaign offers.
  3. Output metrics: volume of sales generated across different products, number of cross-sales.
  4. Outcome metrics: revenue growth, AUM growth.
  5. Impact metrics: market share, customer share-of-wallet.

These metrics need to be configured so that we address the monitoring objectives stated in the earlier paragraph — deviation, peer group difference, change over time, and correlations.

Here’s another example of metrics for managing the Learning & Development (L&D) function in an organisation:

  1. Input metrics: number of high-quality learning assets (courses and materials), number of learning partners / vendors, learning budget per employee.
  2. Activity metrics: learning hours, utilisation of learning modality (i.e. classroom instruction, asynchronous online, etc.).
  3. Output metrics: learning roadmap completion, learning satisfaction score.
  4. Outcome metrics: cost per learning, learning gains / learner’s performance improvements.
  5. Impact metrics: talent mobility rate, attrition due to skills redundancy.

Getting the Right Design

Let us now consider the aesthetic visual elements that are deemed as best practice for modern dashboards design:

  1. ‘F’ shape arrangement of information to minimise eye tracking — because of social media attention priming over the decades, your audience will read from the top left-hand corner of the page and scan across to the right. They will then go the next line starting again at the left-hand corner and scanning right. And then they will scan top-to-bottom starting at the top left-hand corner. This is the quintessential ‘F’ shape. Therefore, you need to arrange your metrics / information based on this eye-tracking pathway in order of importance.
  2. Use the right visual vocabulary based on your metric objectives — we already landed on dashboards being used for monitoring.
  3. Maintain consistency in terms of visual vocabulary, colour and visual cues to minimise cognitive load — e.g. the same colour must mean the same thing across the charts in the dashboard.
  4. Place comparison metrics placed side-by-side to accelerate information scanning — e.g. if you want your audience to compare the metrics across 2 products, then the charts depicting each product’s metrics should be placed side-by-side rather than top-to-bottom.

Conclusion

Like data visualisation, designing good dashboards is a learnable competency. A major flaw in many organisations is that they treat dashboard design as a literal design project and engage with visual designers for the tasks. But they miss out the fundamental aspects of metric construction and information conveyance. It’s important that we treat dashboard design as a holistic endeavour, as an important addition to an organisation’s capability assets.

--

--

Eric Sandosham, Ph.D.

Founder & Partner of Red & White Consulting Partners LLP. A passionate and seasoned veteran of business analytics. Former CAO of Citibank APAC.