#MakeoverMonday this week looks at the relationship between the price of gold and the price of oil. It has been a great dataset to learn from, as I realised that sometimes gut instinct isn’t always best. As soon as I saw the data, I decided that I wanted to try something new, and visualise the Measures using a connected scatterplot. I hadn’t experimented with this chart type before, so I referred to resources from the link above and banged out a really quick and (so I thought) simple view of the data:

Screen Shot 2017-04-09 at 16.16.13

I was comfortable with the chart type and it at least “worked” for me, but a real benefit of #MakeoverMonday is the feedback system that’s in place. At work, I constantly urge colleagues to seek peer review of their work to ensure that we deliver the best end product, and that is exactly the same situation as I found myself in. A couple of justified early comments were:

Screen Shot 2017-04-09 at 20.08.42

Screen Shot 2017-04-09 at 20.08.54

Neil and Andy were asking the same question, and if I step back from what I produced, what do I see? Well – I look at the two significant horizontal shifts from right-to-left (drops in oil price). Until then, a clear correlation is in place with both Measures on the up.

When those horizontal adjustments occurred, it can be seen that vertical variance is minimal – the price of gold holds up in spite of the drop in oil price. In general, it could be argued that a previously clear correlation between these Measures has been compromised since 2008.

However, the points remain – I hadn’t sufficiently made this clear and a connected scatterplot isn’t a bog-standard chart that people can see and immediately understand. So, what could I do? I decided to iterate. I love the new recap blog posts that Andy and Eva write about #MakeoverMonday, and they have both touched on the way people have taken feedback on board to (hopefully) create a better second/third/fourth stab at the task.

I decided to plot the trends for oil and gold prices on separate charts, and to annotate key points in time affecting the trajectory of those trends:

Screen Shot 2017-04-09 at 20.18.13.png

The axis titles and annotations should make clear what commodity is being referred to. I chose to retain the reference bands from the oil chart on the gold chart, just to show that the marked decline in oil prices was not having a proportionately equivalent impact on the price of gold.

Furthermore, I elected to add a green band to the gold chart to highlight the year (2013) in which the largest adjustment to the price of that commodity was experienced. Why? I wanted to explain the “noise” that affected the path of the original connected scatterplot.

Finally, I tried to tie these separate elements together:

Screen Shot 2017-04-09 at 19.37.42.png

The thinking was that having the underlying trends on the left could aid the viewer to understand the volatility of the connected scatterplot. By drawing attention to the 2008 and 2014 trend shifts on the rightmost chart, I hoped that it would cue the viewer to recall the underlying trends, and I double-encoded that by adding big arrows to denote the decline in the price of oil.

With hindsight, the addition of those arrows is probably superfluous, which is backed up by Colin’s observation:

Screen Shot 2017-04-09 at 20.28.55

A few issues here:

  1. The conclusion that I intended the arrows to draw attention to the leftmost charts, rather than to indicate the decline in oil prices
  2. The confusion about how time is depicted on a connected scatterplot

Colin followed that up with a further sensible comment:

Screen Shot 2017-04-09 at 20.31.12

Given the chart choice and subsequent confusion caused, it seemed logical to iterate once more with the final addition of a “how to read this chart” info button:

Screen Shot 2017-04-09 at 20.40.56

The removal of arrows and addition of the info icon resulted in this final submission:

Screen Shot 2017-04-09 at 20.41.49.png