The actual problem
Every product, hardware and software alike, can send alerts. That is precisely the difficulty: in a single outage the host, the storage, the database and the application all speak up at once, and in that volume the one message naming the cause gets lost.
What a monitoring estate has to do
It is made of senders and receivers, and its job is correlation: relating messages to each other so that fifty events become one incident. Only then is analysis possible at all.
Two kinds, and both are needed
Availability monitoring answers whether something is running. It reports the outage.
Performance monitoring answers how well it is running. It reports the trend leading to the outage, and is therefore the half that prevents one.
Run only the first and you reliably learn that something is broken. Run both and you learn it beforehand.
What to watch
An alert nobody acts on is worse than no alert: it trains the team to ignore messages. Before rollout, settle which message reaches whom and what happens next.