Most businesses find out their website is down when a customer says so. That is not bad luck; it is a monitoring design that never defined what “working” means. A useful uptime strategy starts with the journeys customers depend on, assigns a person to every alert, and leaves evidence that checks actually ran.
This checklist is written to be copied into an operations manual. Each item names an owner, the evidence it produces, how often it is reviewed and who signs it off. Adjust the names to your team, but do not leave the fields blank.
Start with the user journey, not the tool
A ping that returns a status code proves very little. The site can be “up” while checkout fails, the certificate has expired for mobile users, or a payment callback is silently timing out. Define availability around what people actually do.
- List the journeys that matter most. Typically: homepage and category pages, search, account login, add to cart and checkout, payment callback, and the admin panel. Owner: operations lead. Evidence: one-page journey map with URLs and expected outcomes. Frequency: reviewed when the site changes. Sign-off: business owner.
- Write a pass rule for each journey. State the expected status code, a visible content marker, a response-time ceiling and the certificate condition. “Page loads” is not a pass rule. Owner: technical lead. Evidence: test definition per journey. Frequency: on release. Sign-off: operations lead.
- Separate the check types. Availability checks answer “can it be reached?” Functional checks answer “does the transaction work?” Performance checks answer “is it fast enough?” Infrastructure checks answer “is the server healthy?” You need all four, but they fail differently. Owner: technical lead. Evidence: check-type summary. Frequency: quarterly. Sign-off: operations lead.
- Publish planned maintenance windows. A window without a start, end and owner becomes an excuse for silent downtime. Owner: operations lead. Evidence: calendar entry and change ticket. Frequency: per change. Sign-off: business owner for customer-facing work.
Build the check inventory
The inventory is the contract between the business and the monitoring service. A short list that is tested beats a long list nobody trusts.
- Synthetic check: homepage and two deep pages, every minute, from at least two networks or regions. Owner: operations. Evidence: check configuration and 30-day history. Frequency: continuous. Sign-off: monthly review.
- Transaction check: a scripted checkout using a test account and a test payment method, at a frequency your provider supports. Owner: ecommerce lead. Evidence: successful test order reference and timestamp. Frequency: every few minutes. Sign-off: monthly review.
- Certificate and DNS: expiry, chain validity, and resolution from public resolvers. Owner: technical lead. Evidence: expiry report and resolver output. Frequency: daily check, weekly review. Sign-off: monthly.
- Third-party dependencies: payment gateway, SMS or email provider, CDN and any API the checkout calls. Owner: technical lead. Evidence: dependency register with status links. Frequency: continuous where possible, weekly otherwise. Sign-off: quarterly.
- Server and queue health: disk, memory, database connections, queue depth and failed jobs. Owner: infrastructure lead. Evidence: dashboard with thresholds. Frequency: continuous. Sign-off: monthly.
- Backup completion: the job ran and the artefact exists. This is not the same as a restore test, which belongs in a separate schedule. Owner: infrastructure lead. Evidence: job log and storage listing. Frequency: daily. Sign-off: monthly.
Give every alert an owner and a next action
An alert sent to a shared inbox is a notification, not a control. Define severity, a first-response target and a named escalation path for each level.
- Severity definitions: a full outage, a broken critical journey, degraded performance and a single failed non-critical check should not wake the same person. Owner: operations lead. Evidence: severity matrix with examples. Frequency: annual and after a major incident. Sign-off: business owner.
- Routing and escalation: name the primary responder, the backup and the person who decides on customer communication. Owner: operations lead. Evidence: on-call rota and escalation list. Frequency: monthly. Sign-off: operations lead.
- Runbook link on every alert: the first three diagnostic steps and the rollback option should be one click away. Owner: technical lead. Evidence: alert-to-runbook mapping. Frequency: per new check. Sign-off: operations lead.
- Alert tests: trigger a test alert and time how long acknowledgement takes. An untested route is an assumption. Owner: operations. Evidence: test log with timestamps. Frequency: monthly. Sign-off: operations lead.
Make customer communication part of the runbook
Monitoring tells you something is wrong; it does not tell customers what to do. Decide in advance who updates the status page, which channels carry the message, how often updates appear and what the post-incident note will say.
Keep three templates ready: investigating, identified and resolved. Include what is affected, what still works, the next update time and where to get help. If personal data may be involved, route the decision to the incident lead rather than the marketing inbox.
Connect monitoring to changes
Most avoidable outages follow a deployment, a DNS edit, a certificate renewal or a plugin update. Link every change ticket to a monitoring watch period and a rollback trigger. After a release, run the transaction check manually once and watch the error rate for a defined period. If the check fails, roll back first and investigate second.
Review the evidence monthly
A monthly review should take less than an hour and produce decisions, not a dashboard tour. Review alert volume by type, false positives, incidents found by customers rather than monitoring, expiring certificates, failed backups and open remediation items.
Every quarter, test one restore, compare check coverage against the journey map, review third-party dependencies and look at on-call load. The operations lead and business owner sign the review; the evidence pack is retained so an auditor, insurer or new manager can see what was actually checked.
What this checklist prevents
- Monitoring the homepage while checkout is broken.
- Alerts that arrive in a channel with no named owner.
- Certificate expiry discovered by a browser warning.
- Alert blindness caused by maintenance that was never suppressed.
- Backups that are assumed to exist because a job was scheduled.
- No record that monitoring ran at all.
First week actions
- Write the journey map and choose the five most important paths.
- Assign one owner for monitoring and one backup.
- Configure synthetic and transaction checks for those paths.
- Trigger a test alert and fix the escalation route before go-live.
- Put the monthly review on the calendar with a named sign-off.
If you need help turning this into a monitored service with a documented maintenance history, start with Managed Technology & Cybersecurity. The pricing page explains how monitoring and retainer work are scoped. For the response side of an outage, read how SME incident response changes as the business grows.
Sources
- Google, SRE Book: Monitoring Distributed Systems, accessed 17 November 2026.
- Prometheus, Alerting practices, accessed 17 November 2026.
- IETF, RFC 9110: HTTP Semantics, accessed 17 November 2026.
- NIST, SP 800-61r3: Incident Response Recommendations and Considerations for Cybersecurity Risk Management, accessed 17 November 2026.
Responses (0)
No comments yet. Add the first one below.
Leave a response
Your email address will not be published. Required fields are marked *.