- 01Understand
- 02Prioritize
- 03Acting
A company can have five tools and still not find the cause of an incident. Maturity of observability and PMA is not measured by license: it is measured by the ability to move from a symptom to an action with evidence.
This checklist allows you to get a first photograph in 15 minutes. It does not replace a technical assessment, but discovers priorities.
How to Rate
I answered every statement with:
- 0: there is or is not known.
- 1: exists partially or depends on people.
- 2: is documented, tested and measured.
There are 20 questions and a maximum of 40 points.
1. Coverage and objectives
- Critical services are identified and have owner.
- Each critical service has a measurable SLO or expectation.
- The user experience is observed in addition to the infrastructure.
- External dependencies and critical routes are documented.
Sign of risk: the computer monitors servers, but does not know which transaction affects the business.
2. Instrumentation and quality
- Metrics, logs, and traces share service names, environment, and version.
- Logs are structured and contain correlation context.
- Traces cross critical services without relevant cuts.
- There are controls for sensitive data and cardinality.
Sign of risk: each label production equipment in a different way or secrets appear in logs.
3. Detection and alerts
- Urgent alerts represent an imminent impact or risk.
- Each alert has owner, severity and playbook.
- Percentiles and SLOs are used instead of isolated averages.
- Noise, false positives and alerts are measured without action.
Sign of risk: the guard channel receives hundreds of warnings and major incidents are discovered by users.
4. Research and response
- It is possible to navigate from an alert to related logs and traces.
- Deployments and changes appear in the timeline.
- Development, operations and security share incident identifiers.
- It measures time to detect, diagnose and recover by severity.
Sign of risk: each incident starts by asking for catches of different dashboards.
5. Government and costs
- The volume and cost of telemetry can be attributed to equipment or services.
- Retention and sampling depend on criticality and use.
- Dashboards, alerts and pipelines have owner and periodic review.
- Incident lessons generate verifiable changes in instrumentation.
Sign of risk: the bill grows, but no one knows what data can be removed.
Outcome
| Score | Level | Reading |
|---|---|---|
| 0–10 | Reagent | Visibility depends on isolated people and tools |
| 11–20 | Instrumented | There are data, but little correlation and government |
| 21–30 | Operable | Critical services are researched with repeatable processes |
| 31–36 | Managed | SLOs, costs and quality are measured regularly |
| 37–40 | Adaptive | Telemetry improves with incidents, changes and business objectives |
The total score is indicative. A zero in data protection or coverage of a critical transaction may be more urgent than several points in dashboards.
Evidence that should accompany the score
Don’t assign a “2” just because a person claims that the practice exists. I asked for verifiable evidence:
| Capacity | Minimum evidence |
|---|---|
| SLO | definition, window and active board |
| Correlation | a trace connected to your logs |
| Alerts | owner, playbook and execution log |
| Costs | consumption attributed by service or equipment |
| Improvement | change closed after an incident |
The evidence prevents assessment from measuring perception. It also makes it possible to repeat it: if a practice depends on a single person or a dashboard that no one maintains, it should be rated as partial. Keep links, responsible and verification date next to each response.
What to prioritize according to the level
Reactive: create a baseline
I chose a critical experience, defined owner and four signals: traffic, errors, latency and saturation. Document the scaling flow.
Instrumented: correlate
Normalize service.name, environment and version. Add trace_id to logs and instrument the main route with OpenTelemetry.
Operable: manage SLOs and costs
Replaces infrastructure alerts without impact by symptoms. Apply error budgets, criticality sampling and consumption attribution.
Managed: automate with controls
Automatize enrichment, tickets and reversible actions. I measured whether automation reduces times without increasing risk.
Adaptive: Connecting with business and security
It relates telemetry to conversion, operation and risk. I shared context with the SOC and reviewed coverage for changes in architecture.
90-day road map
First 30 days
- Baseline services, tools and costs.
- Selection of a critical experience.
- Definition of SLO and owner.
- Proof of a known fault.
Days 31 to 60
- Instrumentation and correlation.
- Redesign alerts.
- Resource conventions and logs.
- Decision-oriented service board.
Days 61-90
- Sampling, retention and budget.
- Playbooks and guards.
- Measurement of times by stage.
- Template to incorporate the following service.
How to prevent the assessment from being in a PDF
I turned each priority gap into a measurable hypothesis. Example: “If we correlate log and checkout traces, the time to locate the cause will drop from 45 to 15 minutes.” Try it with a controlled exercise and record the result.
Maturity grows when the cycle repeats itself: observe, learn, change and measure again.
Next step
CTL can perform an assessment of observability and APM with architecture, operation, security and costs. Complete the SOC diagnosis o agenda a meeting.
Frequently Asked Questions
Should the score be the same across the company?
No. The depth must follow criticality and risk. Similar services should be compared and priority given to those that sustain critical experiences.
Does maturity depend on using a single platform?
No. Several tools can work if they share identity, processes and governance. A unique platform also does not solve poor quality data.
How often should the checklist be repeated?
Quarterly or after relevant changes. It is also advisable to review it after incidents to verify whether the improvements were implemented.