Visualising your data: How do you learn from what you have?
By Shaheen Ahmed-Chowdhury, part-time Data Visualisation Specialist at ERIKS Digital and Mathematical Sciences student at Utrecht University.
A lot of companies have a lot of data. Data on customers, on orders, on website clicks, on employees, on finances, on invoices, on anything you could imagine. A common way of storing this data is within what’s known as ‘relational databases’, such as those can be implemented in SQL. Different tables have links through what are labelled as ‘keys’, and it’s an art form in itself in structuring a new database so that it’s ‘primary’ and ‘foreign’ keys are planned correctly. Fail at this step and you have a clunky, unmaintainable pain in the rear which will only make your and your colleagues’ lives harder than they need to be.
However, let’s say that you’ve done a good job in this step. You’ve perhaps managed to translate a huge and cumbersome Excel spreadsheet into something that follows relational database design ideas well, and you’ve managed to set up a data feed to this database, whether this be through some rudimentary application which users click and fill in fields or some API connection which pulls data in from an external website such as Google Analytics.
Your database is getting fed data, and all of that data has been separated out into a sensible table structure. Now how do you look at it?
Much of the insight that can be gleamed from this data will remain locked away within your new fancy relational database setup, if you don’t find a way to look at it!
Well, this is where PowerBI comes in.
I’d never heard of PowerBI before I came to ERIKS Digital, but in my time here, I’ve learnt to work with it, and I am pleasantly surprised.
PowerBI makes the task of ‘seeing’ your data very simple. You connect to your database within Power BI, import the data you want, and play around with a visualisation interface that actually, is probably easier than Excel’s!
You can drag and drop different columns together to form figures, in a way that perhaps takes some getting used to (some of the axis and field names for figures aren’t the most intuitive), but once you get used to this, the process becomes very quick to producing visualisations.
You don’t have all the flexibility you do with visualisation packages in Python such as Matplotlib or Seaborn, but most of the time you don’t need it. You just need to throw some bar charts or line graphs together and visualise some trends over time. Visualisations for business don’t need to be overly complicated, they just need to be clear, concise and easy to process.
This is what some of us do for the marketing team here at ERIKS Digital. We produce clear, easy to process visualisations of messier data, and Power BI allows me to do this quickly, and react to the marketing team’s changing requirements. I don’t have 150 lines of visualisation code which I need to pore over, re-importing a different dataset with different columns and different calculations of aggregate statistics. Power BI takes all of that away from me and lets me just adjust the things that I want.
It allows me to get to the core of my responsibility as a Data Visualisation Specialist here at ERIKS Digital, and that is to produce actionable insights, which our marketing team can use in their day-to-day decision-making processes. They work heavily with multiple sources of data, and with a level of technicality that perhaps hasn’t commonly been associated with the sphere of marketing in the past.
These actionable insights we aim to produce need to be clearly displayed, and this idea of clear, coherent data visualisation is one that has become somewhat of an art form within the data science community. Competitions from data science websites such as Kaggle often push this exploratory, clever yet instantly processable form of data visualisation, and value it highly.
https://www.kaggle.com/nahimsouza/data-vis-grammar-of-graphics-with-plotnine
This approach is often labelled as a ‘grammar of graphics’, and more and more visualisation packages seem to be being produced with it in mind (such as in the article linked above). R’s ggplot2 package is another quintessential example of this, and PowerBI is its user-friendly, easy to implement little brother. Complexity is stripped away from the process, so that the user can focus on what they need to, which in our case is to produce clear and informative data visualisations.
However, a cost comes with abstracting away this complexity, and that is the introduction of a level of inflexibility in visualisation. If you need to provide insights over several different variables, say country, profit, time and product type, you may start to feel that the ‘Lego box’ of tools Power BI provides you isn’t large enough. You may have to find workarounds that wouldn’t have been necessary in languages like Python. However, for my business requirements, I’m happy to accept this trade-off.
Personally, I may not accept it for the rest of my career, and a part of me actually enjoys the cerebral challenge and complete freedom Python provides, but it’s good to know that Power BI is there as a tool which can quickly give business analysts visually appealing results.
