Service-level objective

Fixed retrieval budgets over-retrieve for easy queries and underserved hard ones.

Their adaptive approach hits 95% SLO compliance versus 30% for the best static baseline.

The fix

for over-retrieval is adaptive retrieval budgets.

And so this is where I get into, you know, exposure management, right? Because There I'm able to say, look, you know, we are going to move faster.

We're going to turn around a, you know, a 12 hour SLO for public facing criticals.

Right.

Or 24 hour SLO, whatever service level objective.

Right.

It's not an SLA unless you're a service organization.

Okay.

And there's all the SLAs and not SLOs.

We stand behind the SLAs.

There's actually, you know, financial teeth on them kind of thing where if we don't meet those.

So the system detected that they were burning through their monthly allowance of failed starts faster than expected, and it paged the on-call before the user-facing impact became obvious.

So SLOs are not just about measuring, they're about triggering action.

Exactly.

And that is the angle I want to drill into today.

Most people think SLOs are just a dashboard you show management.

But the real power is in using them to catch problems early, before they become incidents.

And the key to that is something called burn rate alerting.

They rolled back the config, and by the time customers noticed, it was already fixed.