seniorData platformPro answerGrounded in authoritative sources

When you propose a new observability signal, how do you justify it, why that metric, and what does it cost to collect?

This question is asked because monitoring proposals are cheap to write and expensive to live with, and the interviewer wants to know which side of that you have been on. There is a recognisable weak answer: a list of interesting metrics with no stated threshold, no cost, and no owner. The strong version runs the other way round, starting from an action somebody takes at three in the morning and working back to the smallest signal that triggers it. PostgreSQL happens to make the cost side unusually concrete, because the available instrumentation has visible and very different price tags. This page covers a ninety-second structure, the cost dimension people forget, and what to commit to after the signal ships.

What the interviewer is scoring

They're listening for the decision the metric drives, the threshold, the collection overhead and cardinality, and the cheaper alternative you rejected, not a dashboard wish-list.

In short: Every signal has to earn its keep: it must map to a decision, and its collection cost and false-positive rate have to be smaller than the pain it prevents.

ProFrom a real interview loop

The spoken answer and the reasoning behind it

The question, the intro, what the interviewer is scoring, the reasoning outline, what you would verify and the follow-up probes are all open above. Pro unlocks the spoken answer to “When you propose a new observability signal, how do you justify it, why that metric, and what does it cost to collect?”, the reasoning behind it, and the verification that backs it.
  • The full 90-second answer, written first person, the way you would actually say it
  • How I reason through it: the mechanism, the decision points, and where the claim stops
  • Verification steps grounded in the documented behaviour and the source, with no lab claim attached.
  • Worked responses to the 3 follow-up probes listed above, plus the traps that lose the point

Card required. Cancel before day 7 and you are not charged.

Compare plans

How to reason through it

  • Start from the decision or action the signal enables; if nothing changes when it moves, do not collect it.
  • State the threshold before it ships, otherwise it is a graph rather than an alert.
  • Price collection: overhead, cardinality, retention, and who pays it.
  • Address false positives; a noisy signal that gets muted is worse than none.
  • Name the cheaper alternative you rejected and why this one wins.

What I would verify

  • Ask who acts on it and what they do; an unowned signal is not a signal.
  • Estimate cardinality at the highest dimension anyone will actually ask for, not the lowest.
  • Review after it fires: if nobody acted, either the threshold moves or the signal comes out.
  • Check whether existing instrumentation already answers the question and is simply not being read.

Follow-ups they push on

  • Give me a signal you decided not to collect.
  • How do you set the threshold if you have no history?
  • What is the real cost of high cardinality?

The probes are open. Pro carries the spoken answer, the reasoning behind it, and the verification steps, including a worked response to each of these.

Concepts tested

Observability designCost vs impactSignal vs noise

Learn it, run it, then say it

Three steps, in order. Nothing here is a detour.

Questions that go with this one