To test ISP SLA and uptime you need timestamped, repeatable measurements from your location — not a single afternoon speed test. Business circuits often promise 99.9% monthly availability (roughly 43 minutes downtime); consumer plans may offer no credit at all unless you document prolonged outages. Whether you are negotiating a credit on fiber or proving chronic congestion on cable, the workflow is the same: define the metric, measure on a clean path (preferably wired), log continuously, and compare against the contract. speedtest.doctor provides free stability runs, uptime monitoring, and test history on self-hosted api.speedtest.doctor infrastructure.
What SLAs actually promise
Read the fine print before buying monitoring hardware. Common clauses include: availability (link up), latency caps to the ISP core, packet loss thresholds, and mean time to repair (MTTR). Many consumer terms exclude "planned maintenance," "customer equipment," and "events outside our control." An SLA you cannot measure is marketing — know which numbers are enforceable.
Translate percentages to minutes. 99.9% monthly ≈ 43.8 minutes allowed downtime; 99.95% ≈ 21.9 minutes; 99.99% ≈ 4.4 minutes. Exceed the budget and you may qualify for service credits — if your logs meet their format requirements.
What to measure (and what to ignore)
- Link up/down — modem sync lost, WAN DHCP failure, total loss of ping to ISP gateway.
- Throughput collapse — sustained download or upload below contracted tier (harder to enforce on "up to" consumer plans).
- Latency and loss spikes — matter for VoIP SLAs; use ping and packet loss tools.
- Ignore — single Wi-Fi dropout, one failed speed test, DNS typos, your router reboot during firmware update.
Build a measurement rig
Ideal home lab: a small always-on device (Raspberry Pi, NUC, old laptop) connected by ethernet to the router, on a UPS so house power cuts do not fake outages. Schedule probes every 1–5 minutes: ICMP to a well-known dual-stack target, HTTP GET to a neutral host, and periodic stability throughput samples. Label the rig with site ID and timezone UTC in logs — support teams hate ambiguous timestamps.
No spare hardware? Run manual Connection Doctor sessions during incidents and save exports. Weaker evidence, but better than memory alone.
Using stability tests for SLA evidence
A one-shot speed test captures a snapshot. Stability testing runs longer windows, surfacing variance, dropouts and sustained throughput — closer to how SLAs reason about service quality. Run stability during business hours for a week before escalating chronic issues; attach graphs showing repeated collapse at peak times.
Uptime monitoring for reachability
Throughput can be fine while DNS or HTTP fails. Configure uptime monitor checks against critical SaaS endpoints and your own status page. External vantage complements on-prem probes: if the world can reach your site but your office cannot, the fault is local or ISP-ward; if everyone fails, the target is down.
Correlate with ISP and modem logs
Export modem event logs (T3 timeouts, SNR drops, DHCP renew failures) and align timestamps with your monitor. A line sync loss at 02:14 UTC matching your ping gap is compelling. Without modem data, ISPs often blame "your router" — bypass with a laptop directly on the modem for one wired test during the next incident.
Documenting an outage for support
- 01Start/end times in UTC and local timezone.
- 02Symptoms — no sync, high loss, partial throughput.
- 03Topology — wired vs Wi-Fi, modem model, account tier.
- 04Attachments — stability CSV, monitor screenshots, test history.
- 05Ticket number and prior visits — pattern matters for chronic cases.
Business vs consumer enforcement
Business circuits include explicit credits — often 5–15% of monthly fee per SLA miss, capped monthly. Consumer reps may offer goodwill credits without citing SLA math. Know your contract class before arguing percentages. MSPs should store per-client monitors and white-label embedded widgets on status portals for transparency.
Automating with the Measurement API
Scale beyond manual clicks: api.speedtest.doctor exposes programmatic throughput, ping and diagnostic endpoints documented at /docs. Cron a probe every five minutes, push JSON to Grafana or your PSA. Consistent methodology beats exporting screenshots from random consumer speed-test apps that use different CDN paths.
False alarms to filter
Power blips, local firewall updates, laptop sleep, and VPN reconnects look like outages. Tag maintenance windows. Use dual probes — one on LAN, one external — to reduce noise. Is It Down? helps confirm whether a destination is globally unreachable versus unique to your path.
When SLA testing says upgrade or leave
Months of stability logs proving evening congestion or weekly multi-hour outages justify tier changes, technology swaps (cable to fiber), or competitive quotes. Arm sales calls with data, not frustration. Symmetric upload needs and low jitter for video — measure with diagnosis, not anecdotes.
SLA math: credits and measurement windows
Business contracts usually calculate availability over a calendar month, excluding approved maintenance windows announced 48–72 hours ahead. Downtime is often defined as complete loss of connectivity to the ISP gateway, not "Netflix felt slow." Understand whether your SLA uses active probing from the ISP side or only credits customer-reported outages — asymmetric definitions cause disputes. Keep your own probes so you are not solely dependent on the provider's portal graphs.
For throughput SLAs (less common on residential), compare sustained median from stability runs against the contracted minimum, not peak bursts from a 15-second test. Document three consecutive failures before claiming breach — one bad run is noise; a pattern is evidence.
Redundancy and failover testing
Sites with dual WAN or 4G failover should test each path independently. An SLA on the primary fiber link is meaningless if you never fail over to document backup behavior. Use traceroute after failover to confirm routing changes, and uptime monitor targets that ping through each interface. MSP runbooks should list which VLAN or SSID maps to which circuit so on-call engineers reproduce customer tests.
Latency SLAs and VoIP
Some business contracts cap round-trip latency to the provider edge at 25–40 ms. Consumer plans rarely do, but VoIP-heavy offices should still log ping and jitter during incidents. An "up" link with 200 ms spikes fails real-time apps even when HTTP monitors show green. Pair uptime checks with latency thresholds in your internal runbooks — speedtest.doctor's Connection Doctor surfaces both in one pass.
Store monthly summaries: total probe count, failed probes, longest outage window, and median latency. That one-page summary is what account managers accept faster than raw gigabytes of logs. Include your account number, service address, and CPE serial on every attachment so tickets are not bounced for missing metadata. Consistent UTC timestamps prevent timezone disputes with offshore support teams worldwide.
Tool quick links
Start measuring uptime with evidence
Run stability tests, configure monitors, and export results you can attach to ISP tickets.
Frequently asked questions
- What is an ISP SLA?
- A Service Level Agreement is a contract clause stating minimum availability (uptime), response times for outages, and sometimes throughput guarantees. Business fiber often advertises 99.9% or 99.95% monthly availability; consumer cable rarely includes enforceable SLAs unless you pay for business class.
- How do I prove my internet was down?
- Collect timestamped evidence: external ping or HTTP checks from multiple vantage points, modem sync logs, and wired speed or stability tests showing zero throughput or 100% packet loss during the window. Screenshots without timestamps are weak; automated logs are strong.
- How often should I test ISP uptime?
- For SLA disputes, probe every 1–5 minutes from a wired device on a UPS-backed monitor. For home troubleshooting, hourly stability samples plus manual tests during incidents are enough. Match frequency to your SLA measurement period — monthly credits need month-long data.
- Does Wi-Fi outage count against ISP SLA?
- Generally no. SLAs cover the service delivered to the demarcation point (ONT/modem). Document wired ethernet tests at the router or modem to separate in-home equipment from last-mile failures.
- Can I automate ISP monitoring?
- Yes. Use speedtest.doctor's Measurement API (/docs) or embed widgets (/widgets) to schedule throughput and ping checks. Pair with /web/monitor for HTTP reachability alerts. Store JSON results in your ticketing system for audit trails.