A monitoring service the client never sees is a cost. The monthly report is how it becomes a deliverable — including in the months when nothing broke.
A monthly SLA report states, for each client asset, the measured availability against the tier the client is on, every incident with when it was detected and resolved, the response-time trend, certificate status and deploys — in one page the client can read in two minutes. Availability is computed as (total minutes − downtime minutes) ÷ total minutes; on 99.9% a 30-day month allows 43 minutes of downtime, on 99.5% about 3 hours 36 minutes, on 99% about 7 hours 12 minutes. Everything in the report comes from monitoring data already being collected; the work is choosing the tier honestly and presenting the month plainly.
An SLA tier is a promise, so make one you can keep with the hosting you actually use. A site on shared hosting with no redundancy should not be on 99.9%. The tier sets the error budget the month is measured against:
| 99% | ≈ 7 h 12 min downtime allowed per 30-day month. Marketing sites, brochure sites. |
|---|---|
| 99.5% | ≈ 3 h 36 min. Small web apps, portals with business-hours use. |
| 99.9% | ≈ 43 min. Checkout, booking, anything transactional. |
| 99.95% | ≈ 22 min. Only with redundant hosting and an actual response process. |
State whether the measurement is 24×7 or business hours, and whether announced maintenance is excluded. Both are legitimate; both must be written down before the first report.
Availability = (minutes in period − confirmed downtime minutes) ÷ minutes in period. Downtime starts when a failure is confirmed — usually two or three consecutive failed checks, to exclude single blips — and ends at the first successful check after. With five-minute checks the resolution is coarse; say so, and do not report to three decimal places on data measured every five minutes. If maintenance is excluded, subtract announced maintenance windows from both numerator and denominator.
Two figures the report should also carry per incident: MTTD, from the first failed check to the incident being opened and someone alerted, and MTTR, from open to resolved. Detection time is the number most vendors never show — and the one that most clearly demonstrates value.
One page. Charts optional; the table with the minutes is what the client's finance person reads.
The quiet month is the one agencies skip, and the one that matters most: it is the evidence the retainer is working. The report is short — availability 100%, zero incidents, 8,640 checks passed, certificate renewed on the 14th, two deploys with no regression, p95 stable. Ten lines. Send it with the same subject line as every other month. Clients who receive twelve quiet reports renew; clients who receive one report after an outage wonder what they paid for the other eleven months.
Lead with it. "Availability 99.62% against a 99.9% tier — not met. One incident, 2 h 44 min, caused by the hosting provider's storage failure on the 9th; detected in 4 minutes, client informed in 11, resolved when the provider restored service. Actions: moved the site to the provider's redundant tier on the 12th." Then the standard sections. If there is a service credit, state it without being asked. Nothing rebuilds trust like a bad month reported before the client noticed it was bad.
| Formula | (period minutes − confirmed downtime minutes) ÷ period minutes |
|---|---|
| 99.9% | ≈ 43 minutes downtime per 30-day month |
| 99.5% | ≈ 3 h 36 min |
| 99% | ≈ 7 h 12 min |
| Confirmed downtime | From 2–3 consecutive failed checks to the first success |
| Seven sections | Summary · availability vs tier · incidents · MTTD/MTTR · response time · certificates · changes |
| Cadence | Every month, including quiet ones |
For each asset: the SLA tier and measured availability with downtime minutes against minutes allowed; every incident with start, detection, resolution and one-line cause; mean time to detect and resolve; response-time percentiles against baseline; certificate status; and the month's deploys. One page, sent every month regardless of whether anything happened.
Divide the minutes the service was confirmed available by the total minutes in the period. Confirmed downtime runs from the check that confirmed the failure — usually the second or third consecutive failed check — to the first successful check after. On a 30-day month, 99.9% allows 43 minutes of downtime, 99.5% about 3 hours 36 minutes, and 99% about 7 hours 12 minutes.
About 43 minutes in a 30-day month, 8 hours 46 minutes in a year. At 99.5% it is about 3 hours 36 minutes a month; at 99% about 7 hours 12 minutes; at 99.95% about 22 minutes.
The one your hosting and response process can actually deliver. Marketing and brochure sites are usually 99%; small web apps 99.5%; transactional sites such as checkout or booking 99.9%, and only with redundant hosting. State whether measurement is 24×7 or business hours and whether announced maintenance is excluded.
Yes — it is the most important one. A quiet month is the service working, and the report is the only evidence the client receives. Ten lines showing 100% availability, checks passed, certificate status and deploys, sent on the same day as every other month, is what makes a maintenance retainer renew.
Mean time to detect is the average time from a failure starting to it being noticed and someone alerted; mean time to resolve is from detection to the service being restored. Reporting both shows the client how fast you learn about problems, which is the part of the service they cannot see any other way.
Merik produces the monthly SLA report per client from the monitoring it already does. Each asset carries a declared SLA tier, and its health score is an error budget against that tier — so "met or not met" and "minutes remaining" are computed, not assembled. Incidents carry their detection and resolution times; deploys from GitHub and Vercel webhooks sit on the same timeline; vendor status feeds mark which incidents were dependency-caused; certificate expiry is checked daily.
The report is generated monthly per client, and the private status page gives the client the same picture between reports. Because assets belong to clients in the same workspace as projects, quotes and invoices, the retainer that includes monitoring is billed from the same record. See the Digital Operations module.
SLO vs SLA, the minutes arithmetic, a health score as budget remaining, ship or stabilise.
When it earns its place, what to show, private per-client pages, and the honesty rules.
Per-client structure, what to check, who gets the alert, and pricing the retainer around it.