The Blog

Improving Data Visualisation: More Effective and Accurate Heatmaps

Posted by /// 16.11.16 /// Tags: No Tags

By Johannes Lehmann

Organic Response have developed the Portal, a cloud-based platform that can, through our Sensor Nodes, deliver rich insights into building utilization, energy consumption, lighting behaviour and more. If you are unfamiliar with the Portal, our website is a great place to learn about it.

In developing this software there are several aspects that have received special attention, to deliver an experience that is not only a continuation of the simplicity laid out by the lighting control platform, but also presents a compelling value proposition in its own right. One of those more recent developments that we’re particularly proud is the heatmap.

The Portal typically collects sensor data from hundreds of Sensor Nodes per floor. In order to allow meaningful insights to be derived from such a wealth of information, it is imperative that it be visualised in a way that is both accurate and easy to interpret. That is, we need to be able to, at a glance, extract knowledge and that knowledge must be based on accurate and undistorted facts.

Heatmaps are an intuitive visualisation that have received a lot of attention recently. Using occupancy as an example, in a heatmap we would like to show areas with higher occupancy as ‘hot’ (i.e. red) and areas with less occupancy as ‘cold’ (i.e. blue). The way this is commonly drawn is by picturing each sensor as being a source of heat the magnitude of which is greatest at the centre and decreases as we move further away from it. The sum of the heat from all sensors then becomes the heatmap, resulting in something that looks like this:


Although common, this summing has a major flaw – comparing two areas with equal occupancy, the one with more sensors will appear hotter than the one with fewer sensors.

Existing products in the market deal with this problem in different ways. The simplest is to adjust the colour scale such that the outcome is relatively binary, i.e. occupied or unoccupied. This sidesteps the problem by way of inaccuracy by simply showing less.


This is now easier to interpret, but comes at a heavy price. A lot of the information in the previous image is lost. Another workaround is to vary the size of the circles to trade-off the smoothness of the image against accuracy, but it is exactly the smoothness that makes the heatmap so well suited to our human capabilities in the first place!

We weren’t satisfied with any of those approaches and so set out to develop a proprietary approach to drawing heatmaps. Looking at the same data as before, here is what we can show:


The result is an image that is extremely accurate, intuitive and beautiful. The algorithm can almost ‘understand’ and adapt to the floor layout, which becomes particularly apparent when looking at corridors or meeting rooms.

Try to find an interesting insight on this picture and then look at the preceding representations. Is it visible at all? Would you have been able to find it as easily?

Here is another example showing the energy consumption on a particular day:


Seeing which areas have used the most energy becomes absolutely intuitive, requiring no additional interpretation or accounting for inaccuracies. In fact, the effects of a better visualization multiply across all timescales, all types of data as well as all current and future analytics capabilities.

The end goal is to allow the users to interpret data easily and accurately. This has been a step towards making the Portal interface not just more aesthetic but highly impactful towards getting the most out of your building and your Organic Response installation.

No Comments

* Required Fields

Make a comment