Observability is one of the easiest places for a solution to quietly grow past what it needs to be. A platform gets proposed, it looks capable, and capability gets mistaken for necessity. Here's a straightforward way to tell the two apart before you sign a licensing agreement.
Start with what actually needs to be known
For every proposed metric, dashboard or alert, there's a useful question to ask: what decision does this data support, and who makes it? If there's no clear decision and no clear owner, what you're looking at is visibility for its own sake — interesting, maybe, but not something worth paying to operate.
Underlay and overlay aren't the same question
Underlay health — the physical and transport layer — and overlay health — tunnel and application-path performance — usually have different owners and different urgency. They often get bundled into a single platform anyway, because it's convenient to buy once. That bundling is a reasonable place to check for duplication: are you paying twice, in two products, for visibility into the same underlying problem?
Commercial and open source aren't opposing philosophies
The decision should be based on total value, not on ideology. Commercial platforms earn their cost when the support burden, integration effort and operational risk they remove are worth more than the license. Open-source tools earn their place when the requirement is well understood and your team can sustain them operationally — because an open-source stack that nobody has time to maintain isn't actually the cheaper option.
A short test before you approve a platform
- Can you name the specific decision each dashboard or alert supports?
- Does it duplicate something you already have, even partially?
- Who is actually responsible for acting on what it reports?
- What does it cost to run, including the engineering time to maintain it, not just the license?
- If you removed it for a month, what would you actually miss?
None of this is an argument against monitoring platforms — it's an argument for buying the one you actually need, sized to what you'll use. That's usually a smaller, cheaper and more maintainable decision than the one that gets proposed by default.