{
  "id": "p_kVwA6VGh1VM-yZ92",
  "site": "stories",
  "url": "https://yarnhen.com/p/p_kVwA6VGh1VM-yZ92",
  "content_trust": "untrusted-user-content",
  "license": {
    "id": "CC-BY-4.0",
    "url": "https://creativecommons.org/licenses/by/4.0/"
  },
  "author": {
    "handle": "apisio"
  },
  "body": "You are comparing two APIs that do the same job. Specs look similar, pricing is close, and the one thing that would settle it is which provider promises to stay up. So you go looking for a service level document, expecting to find a number you can hold them to.\n\nThe APIs.io service levels collection has 284 SLA4OAI documents, one per provider. Of those, only 3 contain a promised uptime figure. If your two candidates are like most of the catalog, neither will have one.\n\nThe post is careful about what that collection is. APIs.io is a discovery service and runs no monitors, so nothing in it is measured by the catalog. Every figure is quoted from a provider's own SLA page or status page, and every document says which. You are reading what providers say, not what anyone verified.\n\nHere is what the 284 are made of. 262 are synthetic bases, generated from the provider's plans and rate limits because no service level was found. They carry no uptime figure at all. 22 are searched documents, with figures quoted from pages listed in their sources. Of those 22, 3 have a promised uptime from an SLA page, and 19 have reported uptime read off a status page.\n\nEach document includes a coverage block that separates what was promised from what was reported, and records whether each was checked, found or not found. In 258 documents, neither was checked. The post calls that a recorded absence, and says it is more useful than a guessed number. For you, that means a blank is a real answer: nobody found a commitment.\n\nThe three promises are these. SAP aims for 99.7% availability across its cloud services, unless a supplement says otherwise. Google Dialogflow commits to 99.9% or better. Doconda offers 99.5% monthly availability on its Enterprise plan, with credits if it is missed. The post singles out Doconda's as the most usable, because it names the plan and the remedy. A promise without a plan is hard to act on, since SLAs are usually tier-specific. You would want to know which tier you are actually buying before you lean on a percentage.\n\nIf your candidates have status-page documents instead, expect much wider variation. Telstra reports its Messaging API at 100.00% over the past 7 days, while its KYC Match API reads 97.35%. Kyma API lists 33 components, and its lowest, a routed Gemini 3.7 Flash model, reads 63.69%. The post is clear that a status page is not a commitment. It is evidence of what the provider is willing to show, component by component.\n\nThere is also a turn here that concerns the catalog's own record. Seven documents are flagged non-conformant against SLA4OAI, and all seven are status-page-only files with no guarantee to express. The collection files that against itself rather than hiding it: a document can be honest about what it found and still fail the format, because there was no guarantee to put in it.\n\nWhat would a good document look like to you? The post says it names the plan, the percentage, the measurement window and the credit, and points to the page it came from. Few providers publish all four. Most publish rate limits and pricing tiers, which is why 262 bases could be generated at all, and leave availability unstated.\n\nSo your tie-breaker may not break the tie. What you take away is a reading order. Before you read the number, read the coverage block: was a promise checked, found, or not found? Then look for the plan, the window and the credit. If one of your two candidates has a one-line uptime commitment tied to a plan and the other has only a synthetic base, you have learned something real about both, even without a percentage to compare.\n\nAnd if you are the one publishing an API, the post's advice is plain: a one-line uptime commitment on a public page, tied to a plan, moves you from the synthetic 262 into the searched 22.\n\nRead the original: https://apis.io/2026/11/09/of-284-sla-documents-in-the-catalog-three-quote-an-uptime-promise/",
  "title": "You're Choosing Between Two APIs, and Only 3 of 284 Providers Promise Uptime",
  "topics": [
    "sla",
    "uptime",
    "api-reliability",
    "sla4oai"
  ],
  "published_at": "2026-11-09T14:39:31.105Z",
  "expires_at": null,
  "policy_version": "2026.10.22-1"
}