Data Visualisation

A visualisation is a translation - from the language of numbers, which humans read slowly and badly, into the language of visual perception, which humans read instantly. Like any translation it can be faithful or it can lie, and the difference is a matter of craft, not taste. The human eye judges some things fairly accurately, such as relative positions along a scale, and other things poorly, like the precise intensity of a colour or angle in a pie chart. A good visualisation encodes the most important quantity in the most accurately read channel; a careless one buries the key number in the channel the eye reads worst. We know which is which, and we know the distortions that mislead even when no one intends them to, such as axes that exaggerate a trivial change by starting far from zero. A visualisation that misleads is worse than none at all, because it can lead to confident decisions on false premises.

There is no single best chart, only the best chart for a particular question asked of a particular dataset. Different representations are needed even for some of the most common questions about data: how a quantity is distributed, how two variables relate, how something changes over time, how a whole divides into parts, or how groups compare. We choose the representation to fit the question, by writing custom code with tools like d3.js. We pay particular attention to one thing most visualisations simply omit: uncertainty. A finding that is tentative and a finding that is rock-solid look identical as two bars on a chart, and drawing them that way is a quiet lie. Where a result carries uncertainty - and, as our Data Science service will attest, real findings usually do - we make that uncertainty visible, so that the confidence of the picture matches the confidence of the evidence.

A visualisation is communication, and communication has an audience. The same finding shown to a board, to an engineering team, and to a customer is three different visualisations, because each needs to see a different thing and act on it differently. We design for the specific people who will use the picture and the specific decision it must inform, rather than producing a generic chart and hoping it lands. This is craft we have practised where it counts: our team has built data visualisations in professional contexts handling data worth millions to the clients who relied on them, where a misread chart was not an academic matter but an operational and financial one. That is the lens we bring - not visualisation as illustration, but visualisation as an instrument for making better decisions.

The deliverable of a data visualisation service is fairly self-explanatory: data visualisations. You receive visualisations built for clarity, designed around specific questions and audiences, each representing the data faithfully and its uncertainty truthfully, and each delivered in a form you can use, whether that is a static figure for a report, an interactive graphic for exploration, or a component ready to drop into a larger system. Where the finding behind the picture needs analysis to produce it, that is the work of our Data Science service; where you need not a one-off picture but a live, continuously updating view that people watch while they work, that is our Data Dashboards service; and where a visualisation must become a maintained part of your software, our Bespoke Software Development service can build it in. In every case the aim is the same: to turn data into something that improves what your organisation actually does.