On the fifth day of dashboard week, my viz coach gave to me, a Power BI break from sanity.
While for the most part The Information Lab sticks to Tableau and Alteryx as its ETL and Visualization softwares of choice, there comes a time when every Data Schooler must stick their toe into the uncharted waters of Power BI.
Our data today came from a real fruit vendor who posted three years of transaction information on Kaggle. The challenge was to build a KPI dashboard for the store owner using Power BI.
Toying with Power BI
I have never used Power BI. My only experience with Microsoft data visualization is from Excel. I basically haven’t looked back since starting to use Tableau. However, many large companies the world over rely on Powe BI to deliver their business intelligence insights.
It’s amazing how two programs that perform similar functions (Tableau and Power BI) can have such different interfaces and strategies for use. While Tableau is based on understand how your fields create visualizations through the x and y (or row and column) axes, Power BI moves in the opposite direction. Start with the kind of chart you want to make, then fill in the appropriate fields. It was a little hard to wrap my brain around. Basically, Power BI wants to do a lot of legwork for you, but if you don’t understand what it’s trying to do for you, it’s easy to get very lost.
After getting a little overwhelmed at the Power BI interface, I decided to do a little data investigation back in Tableau. Mostly, I was trying to understand how to calculate the sales per line in the data set. This was easy to investigate via a parameter in Tableau that switched between different price fields. I have no idea how I might have done that in Power BI.
After approximating how to calculate total sales, I moved over to Power BI. I wanted to use the KPI functions to show how the sales per month per category changed from the previous year’s value. In Tableau, I would have whipped these values out with some nifty LODs or table calculations. In Power BI, I was lost. Luckily, I still had Alteryx to help me with aggregations and multi-row formulas. That way, I created a file that had a record for each month and category with sales, cost and units.
I brought that into Power BI along with the original file and then I got stumped on how to match up these files. Using Power BI’s data modeling functions you can specify the relationships between tables. However, I got confused on whether Power BI though the connections I was drawing were joins, something akin to a blend, or another process entirely.
After a lot of deleting and re-uploading data sources, I finally seemed to get the fields to match the way I wanted. Here’s my sketch of the dashboard I was planning to build.
I really struggled to get the filtering to work across multiple data sources, something I could have done easily with Tableau’s data blending feature. I also took a long time to figure out that I needed to add a ‘Slicer’ visualization if I actually wanted to put the filter in the dashboard itself. Who could have guessed? Finally, everything fell out approximately according to plan, here’s my final viz.
Not my most beautiful viz, but a decent one for three hours in an entirely new program.
With that, it’s sayonara to Dashboard week. I feel like the week has been very valuable for me in two primary ways. 1) It feels much lower stress to quickly write and publish a blog, and 2) I feel far more comfortable with basing my visualizations off of designs I see elsewhere on the web or in books. Best of luck to the next cohort!