“Fun” with Statistically Valid Sample Sizes
Say you’ve been asked to order enough chocolate for a room of 100 random people. But, you’re getting the good chocolate, so it’s important not to order more than will be needed.
How many people in the room want chocolate? You can ask, but only one at a time. How many people do you need to ask before you know?
If you ask just one person and they want chocolate, do you order chocolate for everyone? What if that one person you ask says no. Do you order no chocolate? This may not go over well.
If you ask 50 people, and 30 of them said they wanted chocolate, and 20 said “no thank you”, how much would you order for the whole 100 person group? 60 pieces? Maybe a few more in case you’re a little off. How confident would you be that you were going to have enough?
These questions actually have industry standard answers used in many different scenarios.
Statistics tells us:
To be 80% certain that you know how everyone will answer your question. How about that buying a little extra just in case? That’s a good idea. Buy 5 extra pieces, to be exact. With this table, the best you can do is +/- 5% (Error Bands).
Can’t get that many people to respond? That’s okay. It just means there is more guesswork involved for you. If you only need to be 75% certain your right, to within +/- 10%, you only need 25 people to respond.
The conversation around what’s “Statistically Valid” can get technical very quickly. But as a quick reference, tables like the ones above let you know if you’re having a conversation about a single person’s issue or an issue that they share with many others.
Want to know more, check out these great resources from Qualtrics, a leader in Customer Experience.
More About Sample Size – https://www.qualtrics.com/experience-management/research/determine-sample-size/
Need a specific number?: – https://www.qualtrics.com/blog/calculating-sample-size/