5 Clues That Something Is Wrong with Your Data
Published: January 21, 2015
Author: Caitlin Halpert
Just because you’re looking at data in a spreadsheet, pivot table, or even an engine UI, does not mean it’s correct or should be taken at face value. Here are some clear hints that something is wrong with your numbers and requires a deeper dive:
1. Any metric declined more than 100%
Pulling data without looking it over might lead you to believe, say, that clicks are down a lot, but any negative number beyond -100 is a flashing neon light that says “CHECK YOUR FORMULAS.”
A change of -136% for clicks, for instance, is impossible unless you are operating in a land with negative advertising where mythical Google makes you pay to get non-clicks. The lowest possible change is -100% (because then you’ll be at zero). There’s no maximum possible increase. Any change approaching -100% variance should be investigated with extreme skepticism.
2. Two alternative views of the same data just report different numbers
It’s often helpful to transform your regular reporting into an alternative view to get a better picture of a specific trend. When completing a task like this, cross-check a couple cells or data points with an existing report. Do December 2014 clicks in Report A match December 2014 clicks in Report B? No? Then you’ve got problems that need to be reconciled.
3. Just one metric is showing astronomical change
Clicks are up 2504%, but all other metrics fluctuated +/- 20%? The data is wrong because MATH. If spend and clicks are up 50%, you’re probably okay, but you should know why those changes happened (see #4).
4. Metrics had a big (>20%) change & you didn’t make any major account adjustments
If performance took a big swing week over week but you haven’t touched the account (why haven’t you touched the account?!?!), then the data is probably wrong, or something else is going on. Bottom line: investigate.
5. You see data points with a value of “0”
This is a biggie – you should only have 0 spend if there were 0 clicks and you should only have 0 clicks if you turned something off. Was it off? No? You’ve got problems. This one’s usually the easiest to fix (incorrect date range, maybe?).
If you find any of the above, check your formulas, check another data source, check tracking, and check your notes. Chances are you’ll quickly find the source and you’ll end up with much cleaner and more accurate reports, allowing you to make better decisions with better data.