Analytics teams using Microsoft Power BI can choose a variety of data visualisation techniques… Are some better than others though?
Microsoft Power BI offers users a wide variety of different data visualisation options to help look for meaning in their data.
Given the large amounts of data Power BI often deals with however, accurately, and more importantly usefully, portraying it has been a large driver for Microsoft over the last few years.
If a user can’t tell at a glance what their data’s saying then are they, in fact, any better off than looking at it in its raw form?
Accurate and efficient data visualisations have become core to modern analytic and business intelligence processes, with many organisations experimenting with the swathe of new options that an improved CRM data structure and Power BI have made possible.
Of course, the guiding principles behind data visualisation haven’t really changed in decades, but as technology has offered more and more visualisation options to developers and analysts the tendency has been to try and cram in as much information as possible, with the end result that data graphics have become more and more complex and inscrutable over time.
Microsoft Power BI cuts through that, allowing data analysts to create their own, custom, data visualisations to help display their data in reports that are a lot more intuitive to grasp.
The problem is that with the range of options available, your data can sometimes get displayed in a way that’s either not the best format or worse, actually misleading to someone reading it.
That’s why we’re discussing some of the best techniques to display different data sets, alongside their associated pros and cons as we know a lot of people really struggle with selecting the right data visualisation technique for the right task.
Most people using Power BI for reporting won’t be looking for a ‘deep dive’ into their data, but rather a high-level understanding of a key metric in as succinct way as possible.
That’s what the KPI visualisation is there for.
It will let you highlight a single key datapoint and show it rising and falling over time.
Starting with the easiest then, line charts are one of the most instantly recognisable ways to display data, allowing users to identify trends with just a glance.
They can display data measured over two axes, with different categories of data being displayed by differently coloured lines.
Line charts are most effective when used to display data over time, for example, rising and falling profits, plotted monthly over the course of a year. In fact, Microsoft Power BI will actually allow you to create time series charts from your data, allowing you to drill down by flipping one axis between yearly, monthly, weekly or even daily instances.
They simplify data into an easily understood graph that can be understood at a glance… however that simplification can sometimes be their weakness as complicated, underlying trends can sometimes get missed.
Need to show something as a percentage breakdown? The humble pie chart is your friend!
Pie charts look great and are helpful in displaying ‘top line’ information but… many data visualisation experts agree that people can struggle to take away detailed information from them, often struggling to process close differences in the different sizes of the ‘slices’.
For high level information at a glance, they’re excellent but if you need more detail, particularly with large data sets, you might want to consider a different data visualisation technique.
Perhaps even more so than line charts, bar charts are one of the easiest ways to display simple data sets. Sans all the different colours, curves, gradients, angles or shapes of other display methods, a simple bar chart is unparalleled for showing off relative sizes of categorical data at a glance (e.g. sales by district).
The bonus to a bar chart is that they can be understood by virtually anyone without any specialist understanding of your data or the need for a key at the side of the graph explaining what everything means. On top of that, Power BI allows you to implement variations on the bar chart for stacked and grouped options, allowing you to visualise the makeup of subcategories whilst still maintain the detail of the overall categories.
Scatter plots, also known as scatterplots, scatter graphs or scattergrams, are used to display the general density of data in a two-dimensional format or the relationship between two values (i.e., outdoor temperatures and ice cream sales).
That means they’re great for displaying relative densities of data as well as overall trends (that you might not have been aware of before) and outliers in an easy-to-read manner.
For instance, along one axis you could plot a sample groups IQ, along the other you could measure the time taken to solve a complex problem. The general density of ‘scatter plots’ would then let you infer trends based on that information.
Bubble charts are used when someone needs to accurately display three different dimensions of data by plotting them along an X/Y axis with varying sizes of bubbles, for instance, using the example above, outdoor temperature vs ice cream sales vs ice cream truck size.
A bubble chart can, with a minimal of explaining, provide quite complex information in a condensed and visual manner.
Some ‘advanced tips’ for bubble charts include labelling the separate bubbles for clarity and adjusting the layout/sizes so bubbles aren’t overlapping each other.
Although not native to Power BI, network diagrams are often used in conjunction with other data visualisation techniques to show how the different data sets being represented elsewhere are connected to each other in reality.
They’re particularly good at letting users grasp the inter-connectivity of data that might have been difficult to imagine without visualisation.
Most data visualisation experts agree however that if not put together carefully, most network diagrams quickly become indecipherable. The trick is to show a high level of detail and then, if needed, let the user drill down on successive network diagrams for greater, expanded information.
Sankey diagrams can be used to plot out process flows via lines and arrows that can vary in width to show relative sizes of the data in individual flows.
Depending on the amount of data present, some Sankey diagrams can, initially, be quite overwhelming at a casual glance but there really aren’t any better data visualisation techniques for the representation of process flows, however, Power BI offers some extra functionality with Sankeys that make them a lot more interactive than normal, allowing users to explore truly complex process flows.
When using one it’s often worth including an annotation to explain to anyone viewing it how the diagram is set up and what it’s demonstrating.
Treemaps are used to display data in a hierarchical fashion in which nested blocks are used to represent different data sets.
Those nested blocks are particularly good at identifying trends but if they’re not sixed correctly or ordered in a non-intuitive way then they can be difficult to read.
Finally, the circle packing chart is a variation on the treemap, using, as the name suggests, circles rather than blocks to represent the relationships in your data.
By plotting different sized circles and placing them within other circles, you can easily display data ‘three levels deep’.
Getting a little (a lot) more technical now, t-SNE or, to give it it’s full name, t-Distributed Neighbour Embedding, actually uses machine learning to model dimensional data sets as two or three dimensional data points for display in a scatter plot, using colours, shapes or other visual elements to represent the third dimension.
Data visualisation analysts developed it to combat some limitations of the more traditional scatter plot.
Much like a line chart, stacked area charts show the breakdown of a trend into subcategories.
A line chart tells you if something’s going up or down but an area chart can tell you what’s responsible for that trend (is it one district having a great month, or are all districts growing equally?).
If you’ve geographic data, Power BI has a range of ways to plot this out.
Without any technical set up you can throw some geographical data at Power BI and it’ll recognise the columns corresponding to areas (could be latitude/longitude, could be postcodes, could be zip codes… not to worry, Power BI will work it out) and plot it on a great looking, interactive map, which also handles grouping things up automatically.
Finally, there’s a vast library of pre-built custom visualisations available online, covering a myriad of uses but if you’re still not satisfied Power BI will let you can build your own, allowing you to display your own data in any way you find easiest!
As we finish up it’s also worth mentioning the interactivity available in Microsoft Power BI.
All Power BI visualisations allow you to drill down into your data at the touch of a button.
For instance, say your dashboard has a map of sales by county, and a line chart of total sales. You might notice on the map that Yorkshire is way below the rest, so you click through, and hey presto, your line chart will automatically show the sales in just that area.