Okay, so today Iām going to be talking about data visualisation and why I believe weāre doing it all wrong.
Are you doing data visualisation wrong?
Now, you think thatās probably quite a bold statement Iām making, but some of the reasons that I think weāre doing data visualisation wrong, is as follows:
- Firstly, I think the visuals that we use lack impact
- One of the reasons they lack impact is our visualisations just have too much data in them
- As a result, they become overly complex and overly confusing
- And the combination of all these factors means that the insight that the visualisations are trying to provide just get lost, or worse still, become misleading.
Now, I really think we need to do something about this and start turning poor visualisations into great visualisations, and I believe it starts with delivering impact with the right visual.
How to improve your data visuals
Here, we have an example of a data visual with two data sets plotted across time. If I draw your attention to the title, itās mentioning about correlation, so I would expect to see something similar across the two data sets. However, our visualisation, clearly, isnāt delivering that.
We can see, for example, the blue line has a seasonal peak, but the orange line looks relatively flat. Now, one of the main issues for that could be to do with the scale. Because we plotted both data sets on the same scale, we may not be seeing the true trends of the orange line at the bottom. Furthermore, in terms of our legend, thatās not overly clear either. For example, blue line says ice cream, but we donāt know whether thatās revenue or consumption or anything else.
Thereās definitely a few things in this visualisation that can be improved, and we can convert it into something that looks a little bit like this. Iāll just talk through each of the key areas weāve done here to try and improve the visualisation to really deliver on the insight, but also to make it clear and simple to understand.
In terms of the title, we havenāt changed too much there. We just added a small subheading, which relates to both the data sets so that theyāre there and clear from the outset. Weāve done a lot of work on the actual visualisation itself. One of the key things is on the axes, so now weāve split the data between each of the two axes. By doing that, itās really brought out of trend of the shark attack data, which we put in a bar just for a nice bit of sort of alternative. From there, that really links back to our title as well, so now weāve got a visual thatās really befitting of the insight that weāre trying to put across.
In terms of the legends, weāve moved that from the bottom up to the top, but now, importantly, we actually have context with our data as well, so we know that weāre looking at number of shark attacks and ice cream consumption.
Furthermore, weāve used effective use of colouring here as well. If you take note of the axes and the legend descriptions, theyāre all coloured in the same way as the data sets themselves, so every time youāre seeing something orange, you now know it has to do with the shark attack data set, and everything blue has to do with the ice cream consumption data set. Itās just little bits there which really help make the visualisation simple for the audience but, importantly, really delivers on the insight as well.
How to reduce the information in your visualisation
Weāve covered visuals and how important that is. Next really important step is to cut the clutter.
Hereās an example of another visual, which is plotting sessions by day over the last 13 months. Now, if we take a step back and look at this visual, what weāre asking you, the audience, to do here is to digest around 400 points of data in a single visualisation. Now, thatās going to really, really difficult to achieve, and not only that, the chart itself looks very, very cluttered and very, very messy.
Furthermore, the headline that weāve got for this chart is that sessions peaked by 643% in early March 2017. Now, realistically, that is a tiny proportion of the graph. The rest of the graph isnāt delivering in terms of the headline that we want to take across.
One of the simple things we can to, to start with, is to narrow the date frame, so letās just say we look at March. Here, the visualisation brings it out a lot clearer. We can clearly see the peak. Actually, itās a little bit easier to understand and digest. However, we can take this a step even further. Having had a look at the data, we could understand that all of that increase was coming from one channel. It was coming from direct traffic, and it was a result of some potential bot traffic coming into Google Analytics.
With that in mind, letās make it even simpler. Letās just focus on the direct traffic because we know thatās where itās coming from. Weāll shorten the time period even more to make it easier to digest. Not only that, because we did that little bit of extra work, weāve got a stronger insight come out, as well, so rather than just saying that sessions have increased 643%, we can now say that potential bot traffic caused a 643% increase in sessions. Not only have we simplified the visualisation drastically, to make it easier to digest and understand, but weāve delivered a stronger insight off the back of it as well, which I feel is really, really important.
Simplicity adds value
Iām also a very, very keen believer that simplicity in data visualisations add significant value, and it doesnāt have to be on chart-based visualisations like weāve seen today. For example, quite often on a presentation, weāll see something like this, which is promoting either a product, a service, or introducing a person.
Now, the format of this would be great if it was on a physical handout. However, in a presentation situation, this visualisation just doesnāt work well at all. One of the main reasons is itās very, very text heavy, so you, the audience, are going to spend your time reading the text, and itās diverting attention away from the speaker itself. Furthermore, in this particular instance, the visual just doesnāt need to be there. Itās cluttering. Itās taking up space.
What Iām going to show on the next slide is that I can drastically reduce whatās on the actual slide but give you more information than whatās on there at the moment. If we move on to this next slide, if we just focus on the title for the minute, just something really simple there: āMeet Neil, Your Digital Analystā, so you now know who I am but, importantly, you know what I do, which you wouldnāt have done before.
In terms of all that text that was on the previous slide, it was broken down into three key pillars. One pillar is around academic, so I can put that down as a pillar, and we can mention Surrey University, mathematics and business studies, but Iāve also added an extra line of when that study took place, so between 1998 and 2002, so thereās an additional piece of information there.
The second piece was talking about experience; so again, letās treat that as a separate pillar and put the key elements there, so GA certification, client and agency side experience, and been in industry since 2008. Finally, it was talking a little bit at the end of the text about interests outside of work. It focused around sporting interests. It mentioned about running, but Iāve split that out into more information, in terms of liking park runs and having an ambition to run a marathon, but also liking football and following Farnborough Football Club as well. So a nice example of that where less really is more if we use the visualisation in the right way.
Donāt lose site of the insight!
Another important thing to consider with visualisations is to not lose sight of the insight. If we take this visualisation as an example, draw your attention to the headline: āMobile Engagement is High for Those Under 35ā. However, the visualisation is really, really difficult to understand. Weāve got three individual buckets of data: for desktop, mobile, and tablet. Then, for each of those, weāve got a bar for each age range. Because of the way itās been coloured, and youāve got the data there in the buckets, itās actually really, really difficult to understand that, for the insight, which piece of data am I trying to point you towards?
In this case, Iām trying to point you towards these two pieces of data to demonstrate the point. However, again, all of these pieces just arenāt adding value in terms of the headline thatās trying to be put across.
Letās drastically simplify this whilst delivering some great insight. The first thing we could do is we could focus just on the under-35 audience, but split it between mobile and non-mobile engagement. Doing this changes things drastically. All Iāve done is just rearranged that data set, put it in a simpler format, and bang, weāve got a really, really strong piece of insight now. We can see that mobile engagement is three times larger for under-35s than non-mobile, so the insight packs a punch, and the visualisation is so much cleaner and so much simpler to understand.
Similarly, we could just look at mobile engagement and compare that between under-35s and over-35s. Again, nice stat there. Mobile engagement is one-and-a-half times larger compared to 35-plus. Now weāre in a situation where we could use either one of those two previous visuals to demonstrate our point much better than the very first visual, but also, weāve got a much stronger insight with some really impressive numbers behind it.
How to achieve great data visualisation?
Iāve talked a bit about various examples of how we can improve data visualisations, but how do we achieve great data visualisation? Now, I think itās a combination of all of these points here.
- Great data visualisation really starts with strong data analysis. You absolutely need that, but thatās only half the story. We canāt just have great data visualisation just off the back of strong analyses.
- Weāve got to have impactful and relevant visuals that are clear and simple to understand for the audience being presented to.
- The relevant data needs to be displayed as well. We need to cut the clutter, make it as simple as possible.
- I think if we combine all of that together, weāre going to deliver strong insights, but importantly, the visual is relating directly back to the insight that youāre trying to put across.
In summary, if I could just collate that together into a couple of sentences, I believe that great data visualisations enable your audience to fully understand the data being presented, and it showcases the insight in the simplest manner.
I believe if we can think about this ethos when weāre putting together our data visualisations, the world will be a better place, and weāll have many more great visualisations to look at. Thank you.