# Clustering and why it might not be a good idea to normalize your data?

I came across this post on Quora and the first answer cleared up why it might be a good idea to always normalize your data, but I want to understand what happens conversely. More specifically, what types of negative impacts would normalization of data have when using clustering?

• We normalize so that a feature doesn't become artificially important just because it has a larger scale than the others. You might instead wind up amplifying a truly unimportant feature that initially has a very small range. But 'unimportant' is determined by the goals of the study. Mar 29, 2021 at 22:03
• @AryaMcCarthy right, I think I understood that. But I want to know in what situations it might be a bad idea to normalize data in the context of $k$-means? EDIT: Sorry, by your second sentence, you mean if we normalize, we might give weight to a feature that may not be as important if we left our data not normalized? Mar 29, 2021 at 22:05
• That's exactly the context. If you have an unimportant feature that's now affecting how your data are clustered, your clusters are less meaningful than if you hadn't scaled the data. EDIT: Yep, that's what I mean. Mar 29, 2021 at 22:07
• There's an example at stats.stackexchange.com/a/140723/919.
– whuber
Mar 29, 2021 at 22:10
• Mar 29, 2021 at 22:31