Driving desired behaviour - are you sure of what’s happening? June 15th, 2015

How do customers react when shown graphic visualisations? It’s not always positive and could be costing you in ways you’ve not considered.

My advise, test what you plan to publish for your customers.

Data visualisation, a current zeitgeist and according to many; you have to be doing it, it’s really easy, hugely satisfying and by just following a few simple steps you’ll be a god to all things data.

Or maybe not.

Taking direction form ‘best practice’ without understanding your customers goals, needs and ability to comprehend information may prove unwise and could be costing you in reputation, recommendation or worst still revenue and increased costs.

Our hypothesis: increased clarity reduces the frequency and need for calling customer services.

What did we test: customer data visualised using a number of charting techniques.

How did we test: one to one interviews lasting around 40 minutes.

What test material did we use: html prototype with parsed XML data visualised using the d3 JavaScript library.

Use case and context: For the average customer the information shown would be 100-200 individual items. For a high usage customer this can run to 1000-2000 lines of detail.

Process

Building an html prototype incorporating the d3 JavaScript library allowed us to understand which of the available visualisation methods were most appropriate for our data.

Further benefit of using HTML was realised as we started revising the concepts as we began to understand the data we would end up publishing to the customer. The ability to do so would have been almost impossible outside of code and we’d have been unaware of a number of anomalies that commonly occur (and drive calls) within the data set.

Visualising the data directed our subsequent design decisions whilst at the same time identifying a number of additional functional requirements to support the programme objectives.

Working in code also mitigated the risk of development teams suggesting the planed approach would not be possible later in the delivery phase as we were using the exact same XML schema currently used in the live systems.

We selected five of the available charting types to take into test. These ranged from the simple bar to visually complex bottle to review against a control of the current tabular view.

The reaction: it was polarised.

Customers either ‘like’ graphics or just don’t see the point. Each group show underlying concerns that add further unexpected levels of risk.

Unsurprisingly those participants that ‘liked’ visualisations took time to investigate the detail and became enlightened to the information within, however. The principle of enlightenment in our case added a level of understanding that drove a negative behaviour - increasing the possibility of calling to query an item. Here individuals started asking questions that they had previously no reason for as they became more aware of the detail within the information.

In cases where the inclination to call was not increased a reduced level of satisfaction started to be observed. Satisfaction reduced where customers started to question the original promise they committed too as their understanding in almost all cases is different to reality.

Confirming a number of knowns

With enlightenment being one of the recognised strengths and reasons to present data visually for the client this resulted in the exact opposite behaviour we set out to drive. As detail increased greater uncertainty sets in, this is recognised as a trigger for the next action to be a call.

Also worth noting is that not everyone will be engaged with your service or have the need to do so.

’When serving customers you don’t need to do everything, just meet expectations and deliver on your promise.’

Not all customers are engaged the same way.

‘If you’ve time to look at that you’ve to much time on your hands’

For those with little or no interest in visualisations negative comments were combined with looks of horror when faced with the visualisation for the first time.

Of those that were in this group a common thread was a lack of interest in a visual representation of the familiar tabular view they currently receive.

Those in this group who regularly questioned the content showed increased levels of distrust in graphical views with comments suggesting that detail was being hidden with this ’designer look’!

It wasn’t all bad

Visualisations can prove beneficial, as long as they are simple, clear and require no interaction to discover the detail within. A practice to follow would be to consider the visualisation as a printed artefact where only ink can deliver the detail, as little ink as possible.

Of the options tested the most successful were the simplest. The design that added structure, order and clarity was most appreciated.

The test showed clearly that classified type was most beneficial. Surprisingly were monetary values existed these proved almost secondary alongside the type and reason associated with the detail.

The option that proved most capable of driving the required behaviour was the least visual presented. In reality this added colour to a tabular view, much more design than this suggested that the test hypothesis would not be proven correct and therefore result in an increase in calls not a reduction.

Lessons learnt, understand your audience and the behaviours you potentially drive.