Begin by labeling your diagram to identify the loops. Then identify the nodes from which branches, well, branch from

Then add a current for each unique branch. Each component should end up having a unique current flowing through it. Like this:
For convenience I chose to number the currents to correspond to the resistor number in the branch (the exception is the current I
1 that flows through
both R
1 and R
3 because they are in the same branch -- I picked the lower resistor number for the current number).
You could add a forth loop that traverses the outer perimeter of the circuit, but it is not necessary for solving the circuit: once you have every component "touched" at least once by some loop, you have enough loops.
Now you need to apply your KVL rules to write the equations. Do a "KVL walk" around each loop summing the potential changes due to voltage sources and the individual currents in the components. Then write KCL for each of the nodes. You'll find that you need to use one less node than you've identified (try it, you'll see why).