How to Manage Tableau Dashboard Versions in Git When You Need to Know What Changed or How a Dashboard Broke. Plus, How it's Easier with Mighty...
Five Ways Data Changes Can Wreak Havoc on Your Dashboards
Are You Wondering What Caused Your Dashboard to Break? It Could be Unexpected Changes in the Data. What to Look for When Things Break Down.
You’ve built a beautiful dashboard. It meets all the client’s requirements, it functions at peak performance, and offers insights to further the company’s mission. Then one day you get a call asking why your report no longer appears accurate. Maybe it’s missing data, or has numbers that are wildly different than prior months. Either way, you know you have to fix it before your client no longer trusts your dashboard. So you go into detective mode: What could have caused your dashboard to break? Many times, it comes down to unexpected data changes.
When we build dashboards, we are building them from current and historical data. We try to plan for the future, but let’s be honest – we aren’t soothsayers. We can’t know what future changes might impact our data and therefore our dashboards. But we can be ready and know what to look for when things break down. Read on to learn about some common ways changing data can wreak havoc on your dashboards:
1. Incompatible data types
IE: String values in a numeric field
When this happens, it’s important to go back to the source and see what is in the data, then ask the data owner what needs to happen to have the entry corrected. This kind of error typically happens when the data source is a CSV or Excel file, because it's easy to enter multiple types of data into a column.
IE: Data with decimals
After confirming that decimals are supposed to be in the data, you just need to change your data type from Number (Whole) to Number (Decimal) to fix the problem.
2. Null values
Nulls are not treated as zeros in calculations unless specified to do so. In fact, if a null is included in the data, that row will be completely ignored. Tableau has a function, ZN(), which will treat null values as zeros to help solve this problem.
See the example below. In this data there are two columns, A and B. You need to add A and B together, but in one row column B has a null value. If your calculation is simply “A+B” then the result in the row where A=5 but B is null will be null. But if your formula is ZN(A)+ZN(B) then the result will be 5.
To prevent confusion, it's important to discuss with your stakeholders early on whether nulls should be treated as zeros.
3. New Data Values
Maybe you’re working for or with a company that has multiple locations. You need the client to be able to filter by each location, so you add the location pill to the filter shelf and click the “All” button, highlighted below.
Everything is working well, until the company opens a new office and your client is asking why that location’s numbers are not being reflected in the totals and why they aren’t able to filter for that location. That happens because when you just hit the “All” button it doesn’t automatically include any new values that get added to the column. If you always want new values to be included in your dashboards, click the “Use all” radio button instead.
4. A New Year Starts
This one is for everyone who has a dashboard with live data that shows trends over time. I’ve met several analysts who, at the start of the new year, go through and update all their dashboards when the new year hits because colors, calculations, filters, etc. aren’t working. For example, say you have a chart that has three years of data and the most recent is colored blue and the others are gray. If you just have “year” on the color then when the new year rolls around, your prior year will remain blue and your new year will be whatever color Tableau decides to use.
Setting up your calculations and pills is key to not having to make these kinds of fixes. For example, instead of putting “year” on color you can create a flag that says the if year=MAX(Year) then “Y”, else “N” end. If you add this to color then the most recent year will always be colored blue. Setting up filters can be tricky, too. Setting your date filters to “Relevant values” and showing the last three years will mean that the most recent three years will always be showing, instead of manually having to change a Year filter. Little changes like this can save a lot of work come January 1.
5. Tableau Outliers
Outliers, particularly tableau outliers, happen to me a lot when I have a fixed axis. Sometimes I need the axis to be fixed because the start and finish values that Tableau assigns just won’t work for the visual. So I set my axis to fixed and all the data displays perfectly. But then one month something happens (ex. global pandemic) and suddenly the numbers are drastically different than any historical figure. If these numbers are outside of my fixed axis range, then suddenly the numbers are flying off the chart and no one can see them. If your chart suddenly seems off (or completely disappears) and you’ve checked the filters and hidden data and nothing is working, be sure to check the axis range and adjust accordingly.
I wish we could see into the future and know all the ways our data could change. This would save so many headaches when my dashboards suddenly disappear. But alas, I cannot. So instead I use the tips above to try and future-proof my dashboards as much as possible, reducing the amount of time I have to spend troubleshooting data issues.
Find Data Changes Fast with Mighty Canary
Mighty Canary is working to make it easier to find data problems before they wreak havoc on your dashboards.
IN PRIVATE BETA
A Simple Tool to Communicate Data Changes to Your Team
How many times have you had to manually check for data changes to see if they crashed your dashboards?
Mighty Canary is building tools that will allow you to be more proactive when the data changes. Spend less time answering questions and more time building meaningful dashboards that drive your organization forward.
Sign up to try the beta and start building more trust with your stakeholders.