when do you put rate limiting on your api? like, whens the right time to do it?

Hey everyone, I've been thinking about API design and wanted to start a discussion around rate limiting. We've seen a lot of growth recently, and while we haven't hit major issues yet, I'm wondering at what point is it wise to implement rate limiting? Is it something to put in place from day one, or wait until you see specific abuse patterns or performance degradation? What are your go-to strategies for implementing it, and what have been your experiences?