TONS of people build a system, launch it, and then have basically no idea what’s happening once it’s running.
Are people using it? Is it working? Is something broken?
You can monitor a hundred different things, but these are the three I always start with.
1. Is anyone actually using it?
At the most basic level, you should know whether people are using your product.
For a normal app, that might be daily active users, signups, or some other core product action.
For a pure backend system, it might just be the number of requests coming in.
The exact metric depends on the product, but you need some simple high-level number that tells you:
Is this thing actually being used?
This is also useful for capacity planning. Once you understand where your system is likely to break, you can watch usage grow toward that limit and plan for the coming scale instead of being surprised by it.
2. What’s breaking?
You should be monitoring failures across the core surface area of your system.
For an API, that usually means watching your non-200 responses.
Not every 400-level response is a problem. A user sending a bad request is normal.
But if you suddenly see a huge spike in 404s, something might be wrong. Maybe you shipped a bug and your frontend is sending users to an endpoint that doesn’t exist.
And obviously, a spike in 500s is something you want to know about immediately.
The point is simple: don’t wait for users to tell you your product is broken.
3. Are your third-party dependencies working?
Modern apps depend on a ton of third-party services (mine usually use 10+).
Payment providers, email services, authentication platforms, AI APIs, databases, analytics tools, and everything else.
Any one of these can fail.
Maybe you hit a usage limit. Maybe billing fails. Maybe an API key expires. Maybe the provider itself goes down.
And when one of those dependencies is critical to your stack, a single failure can break a huge part of your product, so monitor them.
There are obviously a hundred other things you can monitor, but this is a pretty good starting point.
The bare minimum is knowing whether your product is working and, when it stops working, finding out before your users do.
Hope you guys are having a good week. Big launch coming from me soon.
Arjay