skip to content

A provider publishes an availability commitment for a managed service — is that a guarantee, and what does it pay you when missed?

level: juniorimportance: must knowfreq 72%

answer

  1. a contract, not a promise of uptime
  2. remedy, not compensation
  3. credit, not cash
  4. sized by your bill, not your loss
  5. unclaimed usually means unpaid

basics

~20 s

An availability commitment is a contract term, not a promise the service stays up. If the provider's own measurement falls short, the remedy is a service credit against your bill for that service, which you normally have to claim.

solid answer

~50 s

It is a contractual promise about how the provider will be measured and what it owes if the measurement comes out short — not a promise that the service will not go down. The provider computes availability for that one service, in that one region, over a fixed measurement window, from its own instrumentation. If the result falls below the committed level, the remedy is a **service credit**: a share of what you paid for that service in that window, usually tiered so the credit grows as the measured figure falls further. The credit is bounded by your own spend, so it has no relationship to what the outage cost your business, and most providers require you to file a claim within a deadline. Read it as the downtime the provider has budgeted for, not as insurance.

go deeper

for a junior

Recall the three parts: a defined measurement, a committed level, and a service credit as the remedy. Saying "it is a contract about being measured, not a promise of uptime" already answers the screening version of this question.

for a middle

Explain how the credit is sized and why: a share of the charge for that one affected service, tiered by how far the measurement fell, applied against future billing rather than paid in cash, and normally triggered by a claim you file within a deadline.

for a senior

Show you operate it. Name who watches for a bad month, what evidence a claim needs, and how you design on the assumption that the tolerated downtime will actually be used rather than treating the headline figure as expected behaviour.

for a principal

Argue what the commitment is worth in a negotiation. The credit is bounded by spend and never covers exposure, so the leverage is in the surrounding obligations — maintenance notice, incident communication, support response — and in what your own customer contract can honestly promise on top.

## What the instrument actually is An availability commitment — the clause the industry calls a **service level agreement**, or SLA — is part of the contract between you and the platform. It has three moving parts, and confusing them is what makes engineers read it as a guarantee: - a **measured quantity**: a precise definition of what counts as the service being available, usually expressed as eligible time or eligible requests; - a **committed level**: the figure the measurement must reach or exceed over a stated **measurement window**; - a **remedy**: what the provider owes you when the measurement lands below that level. Nothing in that structure obliges the provider to keep the service running. It obliges the provider to *be measured a particular way* and to *pay a defined penalty* when the measurement is bad. A provider that goes down for a long stretch and pays every credit owed has honoured the contract completely. This matters because the word "guarantee" imports an expectation from consumer contracts — that a failure is made whole. Platform commitments deliberately do not work that way, and a candidate who says "the provider guarantees it will be up" is describing a document that does not exist. Note also which instrument this is. A target a team sets for *itself* and tracks internally is a different thing with a different owner; only the provider's contractual commitment carries a credit, and only it is what an interviewer means by "what does the provider actually promise". ## What a service credit pays, and what it does not The credit is the sole remedy in the overwhelming majority of standard commitments. Its shape is consistent across providers even though the numbers differ: | What people expect | What the clause normally says | |---|---| | Compensation for lost revenue | A credit sized as a share of your charge for that one service | | A cash refund | A credit applied against future charges | | Paid out automatically | Claimed by you, within a stated deadline, with supporting detail | | Covers the whole account | Scoped to the affected service, region, and often the affected resource | | Grows with impact | Grows only as the measured figure falls further below the commitment | Two consequences follow. First, the payout is **capped by your own spend**: if a modest monthly charge buys a service that underpins revenue many times larger, the credit cannot approach the damage. Second, an unclaimed shortfall usually pays nothing at all — the obligation is triggered by your claim, not by the provider's incident report, so somebody on your side has to be watching and has to file. Providers write it this way on purpose. A platform sells the same service to enormous numbers of tenants at a published price; accepting liability for each tenant's consequential loss would mean underwriting business risk it cannot see or price. The credit is a symmetric, bounded, predictable remedy — which is exactly why it is a weak one. ## How to use the commitment as an engineer If it is not insurance, what is it for? Three honest uses: 1. **As a statement of budgeted downtime.** The committed level tells you how much unavailability the provider has priced into the service and is prepared to be held to. Design as though that amount will be used. 2. **As a signal about the shape the provider trusts.** Commitments are usually conditioned on running the service in a particular redundant configuration, which tells you which layout the provider believes in. 3. **As an input to what you can promise your own customers.** Your commitment sits on top of everything you depend on, and the arithmetic of that stack is unforgiving. What it is *not* good for is reassurance. "The provider commits to a high figure" is not an availability plan; it is a statement about a contract that pays a small credit if the figure is missed. The engineering work — redundancy, degradation, and knowing what actually breaks — is unchanged by the existence of the clause. ## The reading test When you are handed a commitment, answer four questions before forming an opinion: what exactly is measured, over what window, what is excluded from that measurement, and what must you do to receive the remedy. If you cannot answer all four from the document, you do not yet know what was promised — you know only the headline figure, which is the least informative part of it.

  • Why is the remedy capped at a share of what you paid for that one service?
    Because the provider sells the same service at a published price to a very large number of tenants and cannot see, price, or underwrite each tenant's business exposure. A bounded credit is the only remedy that stays symmetric across every customer, which is precisely why it is small relative to the harm a serious outage does to any one of them.
  • If the credit is that weak, why do buyers still negotiate hard over the committed figure?
    Because the figure is a public statement the provider has to stand behind operationally, and it constrains their own engineering and maintenance practice. Large buyers also use it as leverage for the things that actually matter — advance notice of maintenance, incident communication obligations, and support response times — which are often negotiated alongside it.
  • Who on a team should own watching for a missed commitment?
    Whoever owns the bill, working from the same monitoring the engineers use. The claim needs evidence of impact within a deadline, so it depends on retained telemetry and on somebody noticing that a month was bad. Teams that treat the credit as automatic simply never collect it.

A courier that refunds the postage when a parcel is late is not promising the parcel arrives on time, and the refund is the postage — never the value of what was inside.

saying these in an interview costs you the question

  • Says the provider guarantees the service will not go down
  • Expects the shortfall to refund revenue lost during the outage
  • Assumes the credit appears on the bill without a claim
  • Treats the commitment as insurance against business impact
  • Thinks any outage at all breaches the commitment