The acceptable amount of unreliability a service can have before reliability work must take priority over feature work.
An error budget is the inverse of an SLO. If your SLO is 99.9% availability over 30 days, your error budget is the 0.1% (43 minutes) you're allowed to be down before you've broken the contract. Teams use the budget as a forcing function: while it's healthy, ship features; once it's burned through, freeze features and invest in reliability until the budget recovers. The framework keeps reliability and product velocity in measurable balance.
Without an error budget, reliability is a vague aspiration that always loses to feature pressure. With one, the team has a concrete number that triggers a concrete shift in priorities, and engineering leadership has a defensible reason to push back when product asks for more features in a quarter when reliability is already in deficit.
See the part of the platform that handles error budget in production.