Building Better Dashboards

I've been meaning to post about building dashboards for a while now, as a follow-on from the Beautiful Data post. Data are increasingly a part of our everyday working life — yet far too often they're presented as a basic table: hard to read and difficult to interpret.

In this post, I want to get the ball rolling again on how to make data presentation more insightful. There are five principles I come back to on every dashboard project: relevance, context, colour, story, and aesthetic.

1Relevance

The example above shows three Key Performance Indicators: Visits, Orders, and Conversion Rate. These are very typical metrics used by many sites, as they detail traffic volume (footfall), sales figures (throughput), and effectiveness (efficiency).

People can easily get lost, or simply lose focus, when looking at data — so ensure you're presenting the right data to the right person at the right time. A dashboard built for everyone ends up useful to no one.

2Context

A critical element of data presentation is context. Without it, you're simply stating numbers, which on their own are meaningless. As such, it's vital to always present a trend for a given metric, as this instantly gives temporal context — how has the metric varied over time?

In my dashboard designs, I present a trend line — either as a dedicated chart or as a sparkline, as shown in the tiles above. I also compare the metric against comparable time periods: last week (Week-on-Week), last year (Year-on-Year), a 3-month running average, a target, or a predicted/forecast value. For the comparators, I find it best to present both the literal numerical delta and the percentage (or percentage-point) change side by side.

3Colour

In the example above, I've used colour to alert the user to a negative change in Conversion Rate. Colour can play a great part in communicating the story behind the data — but you need to use it sparingly. You want the user to focus on impactful change, not be distracted by other elements. Think of a flashing red warning light in an aircraft cockpit: it works precisely because it's the exception, not the rule.

You also need to take into account that around 5% of the male population has a red/green colour deficiency, so don't rely on colour alone — pair it with directional icons, bold text, or position to reinforce the message.

You can also try 'The Drunk User' approach (credit to drunkusertesting.com) — a bookmarklet that distorts and skews a webpage, then asks you to complete a task. It's a deliberately exaggerated way to test whether your dashboard can genuinely stand on its own feet under impaired attention and focus.

4Story

One concept I've started introducing into my dashboard designs is a plain-English description of the data. I've found that people often add a "commentary" to their dashboards when circulating them to stakeholders — but this is typically just a description of the data, not genuine insight.

As such, I've tried to remove the need for this basic commentary by adding an automatically generated description, e.g. how a metric compares to the last 7 days, or to the highest and lowest points in the trend. The intention is to free up time for the business impact, rather than the basic logistics of "what happened".

5Aesthetic

Once you have all the core elements of the dashboard in place, you then need to spend time on the look and feel — the aesthetic. You want to make the dashboard engaging for your target audience; don't put them off with pointless or confusing design.

If you apply all of the above, your dashboard should sing. Lastly, think about making the dashboard interactive — letting users drill into a trend, filter by segment, or change the comparison period — so it becomes a tool for exploration, not just a static snapshot.