Skip to content

Service-Level Commitment

Reliability commitment — instantiates Network Effect Governance

A published, accountable promise about uptime, notice, support, and interface stability, so participants who build livelihoods on the network can depend on it not degrading without warning.

A Service-Level Commitment makes reliability an obligation with consequences. Once participants build businesses, workflows, or public services on top of a network, its uptime and stability stop being the operator's private concern and become the ground those participants stand on. This mechanism converts "we'll try to keep it running" into a measured commitment — defined targets for availability, notice, and support, plus a remedy when they're missed. Its defining move is accountability for continuity: it guarantees the reliability that dependent participants need, and attaches a cost to the operator when that reliability fails, rather than leaving degradation as a silent risk they absorb.

Example

A payment-processing rail carries transactions for thousands of merchants, each of whom loses sales the moment it goes down. Its service-level commitment states measured obligations: availability in the 99.9%-class range, a set number of days' notice before any breaking interface change, a defined support-response time, and service credits when it misses. A small merchant can now run a storefront on the rail knowing an outage is the provider's accountable failure — with a remedy — not an unbounded risk they silently carry. The promise is what makes the dependency safe enough to build on.

How it works

  • Measured objectives. Specific, monitored targets — uptime, latency, support-response — with a stated method of measurement, not vague assurances.
  • Change-notice windows. A required lead time before breaking changes, so dependents aren't surprised by a shift in the ground beneath them.
  • A remedy on breach. Service credits, escalation, or other consequences that make a missed commitment cost the operator something.
  • Interface-stability guarantees. Commitments that the interfaces participants built against won't change without the promised notice.

Tuning parameters

  • Objective targets — how high the bar is set and what exactly is measured; a high number on a narrow metric can be easier to hit than a modest one measured honestly.
  • Measurement method — how availability is defined and attributed; the definition decides whether felt outages count as breaches.
  • Notice-window length — long notice protects dependents but slows the operator's ability to change.
  • Remedy strength — whether the consequence actually stings or is a token credit that changes nothing.
  • Coverage scope — which components and interfaces the commitment covers, and which are excluded.

When it helps, and when it misleads

Its strength is that it lets participants safely depend on the network and disciplines silent degradation — an accountable target is far harder to let slip than a private aspiration. Its failure mode is SLA theater: generous-sounding targets measured so they are never technically breached, availability defined in ways that hide real user pain, or credits so trivial they never change behavior — a metric optimized for the report rather than the experience.[1] The classic misuse is a commitment whose definitions make the operator "compliant" during an outage users plainly felt. The discipline is to measure what participants actually experience, make the remedy sting, and remember that a commitment is only credible to those who could leave if it proved hollow.

How it implements the components

  • resilience_and_dependency_review — the commitment directly answers "what happens if the provider degrades service," turning a dependency risk into an accountable, bounded obligation.
  • network_value_invariant — it guarantees the reliability that must remain valuable to those who build on the network, preserving the invariant an infrastructure rail depends on.

It promises reliability but doesn't set who governs the network — that's Governance Board or Council — nor provide the exit if the promise fails (that's Data Portability Rule), nor apply emergency load control (that's Rate Limit or Throttle).

Notes

An SLC is only as real as its measurement and its remedy: a target measured to be unbreachable, or credits too small to matter, is reliability theater. It also presumes participants have somewhere to go if the promise proves hollow — a commitment gains its force from the credible alternative that portability and interoperability provide.

References

[1] The error budget — the SRE practice of defining an explicit allowance of unreliability against a service objective and treating its consumption as a real cost — is the discipline that keeps a service-level commitment honest, because it forces the target to be measured against experienced reliability rather than a favorable definition. ↩