Today was our fourth day of #Dashboardweek and we had to look into drought levels in the United States. For the first time in the week I ran out of time, which means I’m breaking the rules and posting this nearly one hour after the deadline. Before going through why I ran out of time I will introduce you to the task and the data as usual. The task that Andy gave us today was fairly simple, go to https://droughtmonitor.unl.edu, get the comprehensive data calling the API and produce a dashboard. However, compared to the previous days we actually had a lot of interesting data to play with, a few million rows (so many rows that the first time I tried to publish, Tableau Public told me I can’t because it was over 15M).
The Prep
The data prep seemed very easy, but it took me pretty much all morning. The API had al ot of different parameters and I wanted to get the data at its deepest level of granularity which was counties. However, to get the county data I also had to find the FIPS codes of the different counties and this is when I ran into a few issues:
- The FIPS are meant to be 5 digits and some of them were 4. The solution for this was very simple, add a 0 at the beginning of those that started with 4. However, it took me ages to realise this was the problem.
- After solving that problem, I noticed there was a small whole in my map, one of the counties hadn’t been downloaded. I had to do a little research to notice the county has had change name and FIPS code recently. This took me another bit of time.
- Every time I had to fix something like this I had to run the workflow again, which takes like 10 minutes.
- Because I had to pivot my data for the calculation this generated many rows, therefore when I tried to publish my viz to Tableau public I wasn’t able. I had to go back to the workflow and limit the number of years.
The building.
I was very excited for the building part today as I really like having a lot of data to play with. I started creating a map, even trying to create my own background in MapBox. After trying different things and making a decent map, I realised it wasn’t really what I wanted. The problem I had with the map is that is not very useful when you are using dates, as they can only display one date at a time. Also, pages shelves don’t work in Tableau Public. Then I started to explore different avenues, radial charts, joy plot tutorials and when it was over 15:00pm I decided to go for a horizon chart. However, it got in my mind that it had to be a phone version, which meant lots of extra time and something I had never done before. Obviously, I wasn’t going to meet the deadline.
Learning:
- Plan for success! Today I planned but didn’t do it very thoroughly, therefore I wasn’t able to prevent the possible issues I could face. With a bit of extra care, I would have noticed that my plan had a few holes.
- The scoping of the task wasn’t great. I should have taken an easier route that considered the allocated time. instead of trying to take all in.
- The scoping of the data wasn’t great, by now I should remember that Tableau Public doesn’t allow 15M rows and that working with that number of rows is not a good practice.
- When you are time constrained maybe you should limit to what you already know and not try to learn things on the way.
- If you are time constrained don’t go for over complicated visualisations.
This was the result:
We will see tomorrow how much I learnt today as we have another day of #Dashboardweek and I will be on penalty! (we get less time if we take extra time the day before).
If you have any doubts or comments, feel free to use the box below or contact me in Twitter @DiegoTParker
