Availability Report Template: Copy, Fill In, Ship
Key takeaways
- A complete availability report has eight fields: reporting period, monitored services, per-service uptime, incident log, root causes, SLA or error budget status, response times, and a plain-English summary.
- Compute uptime as (total minutes minus down minutes) divided by total minutes, per service, never as a blended average across probes.
- A 99.9% SLA gives a monthly error budget of 43.2 minutes; report budget consumed and remaining on a single line.
- Executives need five fields (uptime, SLA status, incidents, downtime, root causes); engineers need the full report plus per-probe detail, false alarm counts, and alert delivery times.
- Report scheduled maintenance separately rather than excluding it, because customers experienced the downtime even if your monitoring was paused.
An availability report is a one-page summary of your uptime percentage, incident log, and error budget for the month, formatted so stakeholders can read it in two minutes. Below is a copy-and-paste availability report template, every field explained, and a fully filled-in example using realistic sample data, so you can produce a stakeholder-ready report in under 15 minutes.
What fields does an availability report template need?
A complete availability report has eight fields: reporting period, monitored services, per-service uptime percentage, incident log with timestamps and duration, root cause for each incident, SLA or error budget status, alert response times, and a short plain-English summary. That is the whole structure. Anything else is decoration, and anything less leaves stakeholders asking follow-up questions you could have answered up front.
You do not need a tool to produce this. Hosted status pages start around $29/month and will not generate this report for you anyway. A spreadsheet and your monitoring history will.
The template (copy and paste this):
Availability Report — [Month Year]
Reporting period: [first day] – [last day]
Services monitored: [list each probe/endpoint]
Uptime by service
| Service | Uptime % | Downtime (min) | Incidents |
|---|---|---|---|
| [API] | [99.97] | [13] | [2] |
| [Dashboard] | [100.00] | [0] | [0] |
Incident log
1. [Date, time UTC] — [duration] — [what customers saw]
Root cause: [one sentence]
SLA status
Target: [99.9%] | Actual: [99.97%] | Error budget used: [30%]
Response times
Mean time to acknowledge: [4 min] | Mean time to resolve: [38 min]
Summary
[2–3 sentences: what broke, why, what changed.]
Fill the uptime and incident fields from your monitoring export, then write the summary last. For the SLA math behind those percentages, see how to measure availability with clear evidence.
Reporting period and probe coverage. State the exact date range (e.g., "2025-05-01 00:00 UTC to 2025-05-31 23:59 UTC") and list every probe: its target, protocol, and check interval. A report covering only a homepage URL check hides failures in your API or database. Protocol-level probes (TCP, ICMP, DNS) catch the outages a URL ping misses, which is why the probe list belongs in the report itself. If you changed probe coverage mid-month, note the change date.
Uptime percentage and how to compute it. Uptime = (total minutes − down minutes) / total minutes × 100. Use one consistent granularity, usually per-minute probe results, and state it. A service down 43 minutes in a 43,200-minute month is 99.900%. Report per-service uptime, not a blended average, because your API and your marketing site rarely fail together. For how probe design affects these numbers, see cloud availability monitoring: an engineering guide.
Incident log fields. Each incident needs: ID, start time, detection time, end time, affected service, severity, root cause, and customer-facing summary. Severity labels keep the log honest; use a simple two-tier scheme rather than a taxonomy you will not maintain.
| Severity | Definition | Example |
|---|---|---|
| SEV-1 | Customers cannot complete a core workflow | API returning 5xx for 22 minutes |
| SEV-2 | Degraded: slow, partial, or single-region | Checkout p95 latency above 2s for 40 minutes |
Error budget and SLA line. If you promise 99.9% uptime on an SLA, your monthly error budget is 43.2 minutes. Report budget consumed, budget remaining, and whether any SLA credits were triggered. A single line: "Error budget used: 43 of 43.2 min (100%); SLA credits issued: yes." If you blew the budget, say so plainly; stakeholders trust the report more, not less. Google's SRE materials describe the same budget mechanics from the operator's side at sre.google.
What does a completed availability report look like?
Here is the filled-in report for a fictional B2B SaaS called Invoiceflow, covering October 2025. This is what your finished report should look like once you paste in your own numbers.
Availability Report — Invoiceflow, October 2025
- Reporting period: 1–31 October 2025 (744 total minutes in the month, 44,640 minutes)
- Services covered: API (api.invoiceflow.com), dashboard, billing webhook, PostgreSQL primary
- Uptime per service: API 99.97%, dashboard 100%, webhook 99.89%, database 100%
- Overall availability (weighted): 99.95%
- Incidents: 2 total. Incident 1: webhook failures, 41 minutes, 14 October, 02:13–02:54 UTC, caused by a provider IP allowlist change. Incident 2: elevated API latency, 23 minutes, 27 October, degraded not down.
- Error budget: 44.6 minutes allowed at 99.9% SLA; 41 minutes consumed; 8% remaining
- False alarm count: 1 (a single failed probe during a routine deploy, cleared on retry)
- Alert delivery: median time from failure to PagerDuty notification: 48 seconds
- Actions for next month: add a probe on the invoice-export endpoint; raise webhook probe frequency from 1 to 5 minutes
The weighted availability line is the one worth quoting. It is weighted because the webhook handles fewer requests than the API, so a raw average would flatter the month. The error budget line does the same job in reverse: it tells a stakeholder whether the 41-minute outage burned through the month's tolerance or fit inside it. Both numbers take two minutes to compute from incident logs and turn a vague "we were up basically all month" into something a customer success manager can paste into a renewal email. If your probes only hit a homepage URL, the per-service lines are guesses; protocol-level checks against each endpoint are what make the uptime figures defensible, which is the difference between URL pingers and real endpoint probes (see endpoint availability monitoring).
The figures above are illustrative, not real customer data. Pull uptime percentages from your monitoring tool's monthly summary, incident details from your alert history, and compute the error budget as (1 − SLA) × minutes in month. Producing this report manually in a spreadsheet costs you nothing beyond an hour.
Which fields do you cut for executives vs. engineers?
The two audiences need different reports. Executives want the number and the trend; engineers want the probe-level detail that explains the number.
Executive version: 5 fields. Keep it to overall uptime percentage, SLA status, total incidents, total downtime, and a one-line root cause per incident. That fits on one page and answers the only question a board asks: did we keep our promises, and did anything threaten churn? Skip false alarms, probe configs, and response-time percentiles. If a stakeholder wants more, they can ask.
Engineering version: everything plus probe-level detail. Engineers get the full report plus the fields executives never see: per-probe uptime (so API probes are judged separately from the marketing site), p95 response times, false positive counts with the cause, alert delivery times, and error budget burn against your SLA target. Monitoring the right things means probing critical paths at the protocol level, not just checking that a homepage returns 200. If your current tool only does URL checks, you cannot produce this section honestly; protocol-level probes are the reason StatusPulse exists, and the free tier (5 probes, 1 page) or Starter plan at $5/mo covers a small team's report.
For a deeper treatment of probe types and what each detects, see website availability monitoring: a technical guide to uptime and reliability.
How do you generate the report automatically each month?
If your monitoring tool already records uptime, response times, and incident timestamps, the report is just a query over that data. Assembling it by hand each month means copy-pasting from dashboards, which is where transcription errors creep in.
Most monitors expose an API or scheduled email export with per-probe uptime percentages and incident logs. Map those fields to the template once, then script the export on the 1st. If you run protocol-level probes (TCP, DNS, database checks rather than a homepage ping), pull each probe separately so the report shows which layer failed, not just that "the site" was down. DNS failures deserve their own line item; see our guide to DNS monitoring for high availability.
Before sending, check three things, every time. First, confirm the reporting window matches the calendar month your stakeholders expect. Second, cross-check the incident count against your alert history: a probe flapping counts as one incident in the report, not forty alerts. Third, verify scheduled maintenance is excluded from downtime, or your SLA number will be wrong and someone will notice. If your alerting runs through a paging tool, its incident export (for example, PagerDuty's incident API) usually maps directly onto the incident log fields above.
What makes an availability report misleading?
Three reporting habits inflate uptime numbers: excluding scheduled maintenance, averaging across probes, and monitoring only the homepage.
Excluding scheduled maintenance. If you pause monitoring during a maintenance window, that downtime vanishes from your SLA math. Your customers still experienced it. Report it separately, but report it.
Averaging across probes. A US-East probe and an EU probe can produce 99.5% and 98.5% uptime. Averaging them hides a regional outage that broke things for half your customers. Report per-region figures, or report the worst probe.
Homepage-only coverage. Your homepage can be up while your API or auth endpoint is down. Probe the paths customers actually pay for, and report on those.
FAQ
How is uptime percentage calculated for a monthly availability report?
Uptime = (total minutes − down minutes) / total minutes × 100, using one consistent granularity such as per-minute probe results. A service down 43 minutes in a 43,200-minute month is 99.900%. Compute it per service rather than as a blended average, because your API and marketing site rarely fail together and averaging hides that.
What is a good uptime number to report to customers?
Report the number you actually measured, not the number you wish you had. A 99.9% SLA is a common target, giving you 43.2 minutes of monthly downtime. If you beat it, say so; if you missed it, state the miss plainly. Stakeholders trust an honest miss more than a rounded number that unravels under questioning.
Should scheduled maintenance count as downtime in an availability report?
Report it, but separately from unplanned downtime. If you pause monitoring during a maintenance window and exclude those minutes from your SLA math, the outage still happened from your customers' perspective. A separate maintenance line keeps your SLA figure honest while acknowledging the planned work, and it prevents someone from noticing the gap later.
How often should I send an uptime report to stakeholders?
Monthly is the default for most teams: it matches SLA error-budget windows and is frequent enough that incidents are still remembered. Weekly reports suit teams mid-incident-recovery or with aggressive SLAs. Quarterly is too coarse, since a single bad month gets averaged away. Keep the reporting window aligned to the calendar period stakeholders expect.
Can I build an availability report for free?
Yes. A spreadsheet plus your monitoring history covers it: pull uptime percentages from your monitor's monthly summary, incident details from alert history, and compute the error budget as (1 − SLA) × minutes in the month. Hosted status pages start around $29/month and will not generate this report anyway, so manual assembly costs nothing but an hour.