Data Doctor Download: The Advice Column No One Asked For

Dear Data Doctor,

My organization just bought Tableau, and I have no idea where the EX**L to start. They have 9,000 reports (okay, maybe a slight exaggeration, but really, it seems like a wall of endless to-do’s) to port over all my prior reports. There’s one of me, and it seems like too many of them. All the deadlines are yesterday and I have no idea what to do with this tool. What can I do?

– Outnumbered and out of time

Outnumbered,

It’s a hard role you play as the lone analyst. There’s a few options, but it’s going to require higher level support. First, you need champions, starting with the crew in your organization that helped select Tableau (and have them read this playbook to get started). The reasons will help you decide your next move, but also to buffer you as you build out dashboards.

Second, you can’t reasonably port all the reports, and nor should you. Some are likely relics from before, created by someone that left years ago and 2 other people have potentially taken that role and left in that time. Make sure dashboards are being used, then add to the list. Look for overlaps and start migrating the highest value reports. Ideally, “Tableau-ify” them as much as possible.

Lastly, if you can, get outside help. So often, we analysts are asked to make amazing dashboards in tools we hardly know without fully realizing the capabilities or best ways. I may know how to make music, but if my instrument is the guitar, that doesn’t mean I can just immediately play the piano beautifully. It helps, yes, but there are still many things to learn. A tutor or even a bandmate can make all the difference in the world. And, yes, when you have this many reports, sometimes you need a whole orchestra.

Hopefully this helps you undig a little.

Data Doctor

Dear Data Doctor,

I spend all my time doing ETL from flat files. I export the information out of all our programs, and am an absolute pro at Prep. Seriously, once certification for Prep rolls out, I promise you, I will be first-in-line to get my creds because I am that good! Except, frankly, I’m a little tired doing Olympic-level lifting for reporting. Between endless formula finagling, and “Excel Extravaganza” (this is an actual 15-day not-holiday we observe), it’s hard to squeeze the data into a useful form. Is there something else my org should consider?

– Sifting through spreadsheets

Dear Spreadsheets,

If there was ever a data source more prolific than Excel, it’d be CSV. The pain is real, the peril in organizing structured-but-not-structural sources is one that keeps many data agony aunts (AKA consultants) hopping. But, no, it doesn’t have to be this way.

In the ideal world, you’d have some type of data mart, lake or other comparable body-of-water data organizer. These are databases designed to take in a plethora of data, organize it in one manner for storage, and then feed a limited amount of ETL. Rather than attempting to pace Usain Bolt, you can attempt an exercise that doesn’t require breaking records and reduces your strain.

Data models are key. You want data organized at set “grains” for each table (1 row can easily be described as one thing, not a myriad of fuzzy metrics where headers have similar names but different years or other details across 20 columns). Spreadsheets have no concept of grain, and nor do they require or reward it. But databases do. They also like when you create keys, so things can be joined in a logical fashion and split out in ways that reduce duplication. Then, you can even pull it into Tableau’s data model and avoid further data heroics. Databases typically pay for themselves, both by reducing ETL load and by reducing risk of data loss. After all, Olympians are worth a gold medal and so much more.

Data Doctor

Doc,

You gotta help me. My org is stuck on pie charts and I hate them. Not practical pie charts – no, the ones with too many slices to be useful, but hey, bring a plate and you get served maybe? The legend is bigger than the chart. Is there any way to break this habit when they won’t listen to me?

– Pies are for dessert

Pies,

Sometimes a little bad behavior goes a long way. Start with your legend. Make it a bar chart and put the labels IN the bars. Add whatever delimiter you like: a | goes a long way and include your numbers. Toss in a highlight action, which is key to this whole trick. The highlight helps show how absurd this pie can be. Wait a month or 2 and swap it to a set that ONLY shows that one slice and make your pie a donut. Swap out the highlight for a set action that toggles the single slice. Ensure this magic trick colors both the bar and the donut. Dine on bliss as you phase out the pie. And yes, you can ultimately drop the donut too in this manner.

Doc “Donuts Delight” Data

Tune in next time for more data dramas! Can’t get enough advice? Check out the Tableau forums for more advice archives or comment below and our agony aunt(s) may answer them.