Charts speak a thousand words

I probably haven’t had as many burgers in recent months as usual. However, I am of endeavouring to make up for that in the coming months! The thing with burgers is, that it’s not just about the taste. It’s also about how good it looks. Even the tastiest burger in the world isn’t going to appeal quite as much if it doesn’t exactly look the part. Presentation is important with burgers, just as it is with food more broadly. It’s just that in general presentation might seem like a bit of an afterthought, compared with the actual cooking itself.


When it comes to analysing financial markets and data science more broadly, all the buzzwords seem to be about machine learning, artificial intelligence and so on, and for good reason. However, what’s even more important? Being able to understand and communicate your results not just to yourself but others. Obviously, in recent months the coronavirus crisis has really highlighted how important it is to be able to communicate what data is to the general public and not purely to statisticians.


Having tables and tables of numbers isn’t really that optimal for presenting your analysis, particularly if your audience isn’t technical. Even if they are technical, no one really wants to go through pages and pages of tables anyway. Having effective visualisation is key to presenting your results. Of course, we could opt for a simple line chart if we are presenting a time series. Or we can go one further step and use a candlestick chart if we are looking at P&L? Or maybe we can use some box plots? 


It’s also worth noting that using visualisations doesn’t just need to be part of the presentation process, it can also be an important way to explore your dataset and understand it, whilst you’re doing the analysis. For example if we are developing a trading strategy, what does looking at a chart of the returns distribution tell you, that perhaps a line chart of the P&L won’t do? Would a surface plot help to understand the interplay between several variables? If we’re using data corresponding to a geographical areas, what about a choropleth map, say for economic data?


If there is a time dimension, can we create an interactive animated chart, perhaps combining some of the above ideas? One example might be to create an animated volatility surface over time, to display a very large amount of information in an easy to follow way. More broadly, can we create interactive tools to make it easy to explore a dataset, say with a Dash web app or a Jupyter notebook to allow folks without lots of coding knowledge to play around with the data too.


Ok, this is all pretty simple stuff and I don’t think I’ve said anything particularly ground breaking about visualisations, but it’s always worth thinking about! At least in my case, I find it can be easy to settle for the charts I’m most used to (eg. line charts for time series), when perhaps a bit more thought about a more unusual chart could be quite useful.


Even Excel offers a plethora of different chart types, which most of us don’t use all that often. There’s also Tableau which is making big in roads in this area too. If we are using something like Python or R, we have many more possibilities, with tools like Plotly or ggplot, which are open source, unlike Tableau.


If no one can see your analysis, no one will use it. Using visualisation can be a way of ensuring that your analysis reaches a wider audience, and also to help you understand it yourself. Spending time on visualisation is a crucial part of data science.