When to use GARCH?

[user]

Good day to you guys!

Can you please help me understand when do I use a GARCH model? What are the preliminaries that must be done before I can do a GARCH(1,1) model.
Is it only to be used when after performing residual diagnostics, you have found out that you have a non-constant variance?
I have a data on stock price and I am quite confused.
Here's what I did:
1) I estimated a regression with just a constant for the stock return using LS. [ Estimate Equation--> LS log(return) c ]
2) After which, I did residual diagnostics (normality test - checked for kurtosis, graphed the return series and graphically checked for possible conditional heteroskedastic variance (narrow bands in plot), checked for autocorrelation of squared residuals)
3) Then I decided to do a GARCH(1,1) model.
4) Residual Diagnostics on the GARCH model.

Thank you!