You may want to look at the Artin approximation theorem.  Roughly, it says (in the context of varieties, say) that any phenomenon you observe at the level of completions is already achieved etale locally. 

E.g. suppose that $X$ is an algebraic curve over $k$ (an algebraically closed field) and $x$ is a closed point of $X$ such that completed local ring $\hat{\mathcal O}_{X,x}$ is a product of two factors.  Then Artin's theorem says that there is an etale n.h. of $x$ which is the union of two branches passing through $x$.

This has the practical consequence that notions such as $x$ is a node'' can be defined either by a condition on the completion $\hat{\mathcal O}_{X,x}$ or by an etale local condition.  The former is more elementary, and usually easier to check; the latter is often more powerful in proofs, because it is more tightly connected to the geometry of the whole curve $X$.