How many decimal places does your graph actually need?
2, 1 or 0 decimal places?
Most of my projects are in the social sciences. Social science measurement is not exact. We’re often creating instruments and tools from scratch to gather information. There are no rulers, tape measures, thermometers, or fancy gadgets guaranteed to give us exact numbers with exact certainty.
Furthermore, I’m often involved in the data collection process. We’re always going to be missing a few surveys. We’re always going to need to estimate a few numbers. That’s how data collection works in the real world.
I would never recommend that someone make a decision based on two decimal places.
Should Country A be awarded more funding because they reached 89.34% versus 89%?
Should Country B’s intervention program get revised because they reached 49.71% versus 50%?
Does it really matter that Country C reached 25.07%? Can’t we simply round down to 25%?
The last thing I want to do is give decisionmakers a false sense of precision.
In real world datasets, I rarely show any decimal places.
This revised graph would score well on the graph has appropriate level of precision section of the Data Visualization Checklist.
Ready to revise your own graphs?
Do not manually round your numbers up or down. You’re bound to make a rounding error or typo. Or worse, you’ll probably want to bang your head against the wall! Rounding by hand is tedious and completely unnecessary.
Instead, use the Add Decimal and Decrease Decimal buttons on your Home tab to automatically adjust the number of decimal places that are displayed. Since your data table is linked to your graph, the graph will get instantly adjusted.
The raw numbers – like that 9.84% – are still there. They’re not lost forever. They’re simply hidden from view.
Are there instances in which you’d intentionally keep 1 or 2 decimal places in your graphs? Share your opinion below.
Dave Paradi
Sep 22, 2015 -
Good discussion to start Ann. If the graph is a column graph with dollar figures and the exact cents are important, then two decimal places are necessary. For example, one of my clients deals with the cost of parts that go into a piece of equipment and cost to the penny is important because every cent in reduced cost is a penny of profit.
When you have financial or operating data that is accurate due to the measurement methods, then I am OK with one decimal place in percentages. As you say, it depends on what you are using the data for and what decision is being made. If your call answer rate drops from 94.4% to 93.7% over one month in a call center, that is something that requires immediate action. Rounding both to 94% would mask an issue that requires action.
When rounding numbers that will add to a total, 100% for example, one needs to be careful in using the add/decrease decimal method. It can introduce an error when the rounded numbers get added. I covered this and a method of eliminating the errors in this article: http://www.thinkoutsidetheslide.com/issue-329-january-20-2015/.
Thanks for continuing to raise important topics to be discussed.
Ann K. Emery
Sep 22, 2015 -
Dave – thanks for sharing.
“An issue that requires action” – yes! Great litmus test for determining whether decimal places will add value or not.
I love your logic and careful thought process behind these decisions. I often see several decimals included in graphs simply because the analyst’s software program included those decimals places by default.
Three cheers for purposeful design.
Dave Paradi
Sep 23, 2015 -
Ann,
I had an example from a client last year where they included six decimal places in every cell in the spreadsheet simply because Excel can do it. They didn’t realize how crazy it was until I asked them about it. Just because the software can do it doesn’t mean you should let it do it.
I agree that presenters should make conscious decisions. Don’t abdicate decision making by accepting whatever the program default is. Have a reason for each decision.
Dave
Sheila B Robinson, Ed. D
Sep 22, 2015 -
I rarely use decimal places unless a greater degree of precision is needed for a specific decision, or the particular data type warrants it, but it’s certainly not often! A place where I would use decimal places? Graphing the performances at the Giant Slalom during the Winter Olympics! 😉 I’m always amazed that one-tenth of a second can separate the 1st place winner from the 2nd.
Ann K. Emery
Sep 22, 2015 -
Yes! I’m putting your giant slalom example in my “fancy gadget” category. Imagine if our research and evaluation measurements were that precise…!
gf
Sep 23, 2015 -
Similar to the ‘issue that requires action’ comment. Places where the decimal can be a significant amount of the over all number. The difference between 2.6 and 2.9 is a large percentage swing of the overall value, once again masking a difference in the data that could cause users to make a different decision.