The Problem of Setting the Variance of the Observational Error Term

I am newbie in Bayesian estimation and now I would like to ask is there any method tricks that I can follow when setting the size of the prior mean of the variance of the observation error disturbance term when doing Bayesian estimation? For example, refer to the units of the observed data?

A prior is supposed to be set before seeing the data. A very common approach is to follow the precedents in the literature. They should give you an indication.