What the new SLA contract actually says
A standard 2026 freelance SLA clause reads something like: "Service Provider guarantees 99.9% monthly uptime measured at the API endpoint, with credits of 10 percent of the monthly fee per 0.1 percent below target." That single sentence quietly demands 24 hour pager coverage for a one person business.
99.9 percent uptime allows 43.2 minutes of downtime per month. 99.95 percent allows 21.6 minutes. 99.99 percent allows 4.32 minutes. A solo freelancer on a single Vercel or Cloudflare account, sleeping eight hours, traveling on weekends, and depending on third party APIs is mathematically going to breach 99.95 the first time their upstream LLM provider has a rough afternoon. AWS posted a regional Lambda outage of 4 hours 11 minutes on October 20, 2025, and OpenAI logged a 71 minute incident on December 11, 2024. Either event alone burns through a 99.9 percent budget for the entire month.
Clients writing these clauses are not malicious. They are reusing language from their SaaS vendor contracts because the freelancer''s deliverable started behaving like a SaaS vendor. The fix is to negotiate the SLA early, not to accept it and hope for the best.
Why 2026 made this a freelance problem
Two shifts collided. Freelancers started shipping automations through n8n, Make, Zapier, and custom MCP servers, which keep running on the freelancer''s infrastructure long after the project closes. At the same time, finance teams at mid market companies tightened vendor risk policies, partly in response to the Bench shutdown in late 2024 and the wave of single founder SaaS shutdowns that followed.
The result is a procurement template that asks every vendor for an SLA, including the freelancer who built one automation last quarter. PWC''s 2025 third party risk survey reported 68 percent of mid market firms had added SLA requirements to vendor onboarding in the previous 12 months. Most of those firms had no way to tier the requirement, so the same 99.9 percent clause goes to AWS and to the freelancer with two clients.
The three SLA structures that work for a solo operator
There is no version of "I promise 99.9 percent" that works for a freelancer with no on call rotation. The three structures below redistribute the risk so the SLA is honest and survivable.
Structure 1: business hours SLA
Define uptime only during business hours in the client''s primary timezone, typically 9 to 5 weekdays. Off hours are best effort. This cuts the monitored window from 720 hours per month to roughly 160, which makes the same percentage figure 4 to 5 times easier to hit. Clients running internal tools or back office automation usually accept this immediately because their own users are not online overnight either.
Structure 2: response time SLA, not uptime SLA
Replace the uptime promise with a response time promise. Something like: "Critical incidents acknowledged within 4 business hours, resolved within 1 business day. Non critical incidents acknowledged within 1 business day." This shifts the contract from a thing the freelancer cannot fully control (third party uptime) to a thing they can (their own responsiveness). Clients who actually care about responsiveness over raw availability accept this.
Structure 3: shared infrastructure SLA
Pass the underlying uptime through to the cloud provider. "Service availability matches AWS / Cloudflare / Vercel published SLA for the underlying services used in this deployment. Freelancer is responsible for application level availability above that baseline." This is honest, easy to audit, and explicitly carves out the upstream outages that would otherwise blow your SLA for reasons that have nothing to do with your work.
The carve outs that protect you
Whatever SLA structure you sign, four carve outs need to live in the contract.
Third party outage exclusion. Downtime caused by AWS, Cloudflare, Vercel, OpenAI, Anthropic, Stripe, or any other named vendor in the dependency tree does not count against the SLA. Without this, an OpenAI degradation that lasts 90 minutes destroys your month.
Scheduled maintenance window. Two to four hours per month, announced 48 hours in advance, does not count toward the downtime budget. This is what gives you the room to deploy updates and run database migrations without quietly burning your monthly cushion.
Force majeure. Standard contract language, but explicitly call out internet routing failures, regional power outages, and government network disruptions. A 2024 Tier1Research report tracked 174 country level internet outages globally in calendar 2024 alone. Some of those will hit your client traffic.
SLA credit cap. Whatever credits you offer for missed SLA, cap them at 20 to 30 percent of one month''s fee. Without a cap, a 30 minute outage in a 99.99 percent contract can technically refund the entire month plus penalties. Caps stop a small slip from becoming an unbounded financial event.
What this looks like in pricing
The SLA changes the price, and the price needs to reflect it.
| SLA tier | Practical cost to freelancer | Suggested monthly premium | |---|---|---| | Best effort, no SLA | 0 | 0 | | Response time only, business hours | Low: monitoring + working hours pager | 10 to 20 percent over base | | 99.5 percent business hours uptime | Moderate: monitoring + alerting + 4 hour on call window | 25 to 40 percent over base | | 99.9 percent 24 by 7 uptime | High: on call rotation, backup operator, instrumentation | 75 to 150 percent over base | | 99.95 percent or higher | Not appropriate for solo: requires a team | Decline or refer out |
The premiums are not arbitrary. The cost of monitoring at 24 by 7 with Pingdom or Better Stack alone runs $20 to $80 per month. A second freelancer on a paid backup rotation costs another $300 to $800 per month. PagerDuty for solo operators starts at $19 per user per month for the lowest pager tier. Together that is $400 to $900 in fixed cost before you have done any actual work.
When to walk away from an SLA negotiation
There are three signals that the deal is not going to work.
The client refuses to accept third party outage exclusions. If they expect you to be on the hook when OpenAI breaks, the deal is structurally bad and will end with finger pointing.
The client wants a 99.95 percent or higher uptime guarantee on a sub $5,000 monthly retainer. The math does not work. The premium needed to cover that risk would push the retainer well past their budget, so either the SLA tier drops or you decline.
The procurement team will not negotiate the template. Some larger clients have legal teams that cannot deviate from boilerplate. Either accept the boilerplate, walk away, or route through a smaller subcontract with a partner who already has the right insurance and pager coverage.
One operational note
Get errors and omissions insurance regardless of SLA tier. Hiscox, Next Insurance, and Thimble all sell solo freelancer E&O policies in the $35 to $90 per month range as of late 2025. A single SLA dispute can cost more than five years of premiums, and clients increasingly ask for proof of coverage during vendor onboarding anyway. Have the certificate ready.
FAQ
Can I just refuse to sign any SLA at all? Sometimes yes, especially on smaller engagements. On anything over $3,000 per month or with a mid market client, expect the SLA conversation to happen. Going in with the three structures above gives you a counter offer instead of a flat no, which is much more likely to close the deal.
What if the client offers to indemnify me for third party outages? Take it in writing, in the contract, with specific named vendors. Verbal assurances during the sales process disappear when an incident actually happens and a procurement person who was not on the original call is now reading the contract literally.
Does this affect freelance work that is not automations or live software? Less directly. If your deliverable is a finished one shot artifact like a design file, a video, or a written report, SLAs rarely apply. The trap is for work that keeps running, which is most automation and integration work in 2026.
How do I measure my own uptime cheaply? Pingdom, Better Stack, UptimeRobot, and Cronitor all have starter tiers between $0 and $20 per month that handle 10 to 50 checks. Better Stack''s status page is included on lower tiers and makes the monthly SLA report straightforward to produce.
What goes in the monthly SLA report I send the client? Total monitored minutes, total downtime minutes, incident summary with cause and resolution, third party outage exclusions applied, and uptime percentage for the month. Keep it to one page. Send it on the same day every month, ideally the day before the invoice.
Written by Delivvo Editorial · June 5, 2026
More from the blog →