Start with up/down monitoring across all critical systems: network, infrastructure, applications. This gives you the baseline to see where problems originate, not just where they surface.