Turning Y.1731 Insight into Automated Fault Diagnosis - with the Engineer at the Centre

Continuity, delay, loss, and availability measurements provide a clear, objective view of service health across access paths and network nodes. Many engineering teams already seek to distill this data into simple operational states - red, amber, green - to guide fault diagnosis and prioritisation.

At NetMinded, we see this not as the end of the story, but as the starting point for intelligent automation. By structuring performance data through thoughtful Prompt Feature Engineering (PromptFE), we unlock opportunities for automation that span detection, diagnosis, and remediation at scale.

For example, Y.1731 performance monitoring plays a critical role in modern Fiber to the Home (FTTH) operations. PromptFE can embellish performance data with real-time signals which enables automated systems to move beyond static health indicators and take timely, corrective action.

From Performance Monitoring to Action

Once a path or node moves into an amber or red state, experienced engineers instinctively know what to do next.

They ask:

• Is this likely a path issue or a node issue?

• Is the behaviour transient or persistent?

• Which additional signals will actually help explain this?

That intuition is not random. It is the result of deep familiarity with how the network behaves under real-world conditions.

Our goal with MNOC is to make that understanding explicit and reusable, without forcing engineers to change how they think.

Encoding Engineering Judgement into Data

MNOC scoring is NetMinded’s approach to PromptFE - shaping data so that meaning is embedded before it reaches automation or AI systems.

In an FTTH environment, for exaple, this allows the engineer - the Subject Matter Expert (SME) - to define:

• What red, amber, and green truly represent in their network

• How confidence and persistence should be expressed

• When an issue should be treated as localised or systemic

• Under what conditions deeper diagnostics should be triggered

This logic lives in the data analysis layer, not in dashboards or one-off scripts. The result is data that carries judgement, not just measurements.

How This Works in Practice

A typical MNOC-style workflow might look like this:

• Y.1731 telemetry is continuously collected and curated.

• That data is transformed into stable, SME-defined operational states.

• Those states are enriched with context: scope, duration, confidence.

When a state indicates potential degradation, the data product itself signals that additional information is now relevant - for example:

• OLT performance counters

• OLT transceiver metrics

• Interface- or vendor-specific telemetry

Crucially, the decision to look deeper is not hard-coded into an automation engine. It is expressed by the engineer and embedded directly into the data product.

This produces a 1st-class MNOC data product that can drive:

• human fault diagnosis

• automated workflows

• or Agentic AI systems coordinating further investigation

All using the same shared understanding of what the data means.

Designed for Engineers - and for AI

Large Language Models (LLMs) and Agentic AI systems are powerful, but they do not reason well with raw telemetry alone.

MNOC data products act as a translation layer:

• raw measurements become scored states

• noisy signals become contextual indicators

• transient events become meaningful patterns

This dramatically simplifies downstream automation and reduces the risk of brittle or opaque behaviour. At the same time, the engineer remains in control of semantics, thresholds, and evolution over time.

Scaling Expertise, Not Replacing It

MNOC is built on a simple premise:

The best automation reflects how experienced engineers already reason about their systems.

By embedding SME judgement into Prompt Feature Engineered data products, NetMinded enables automation to operate at machine speed - while staying grounded in human understanding.

Y.1731 performance monitoring becomes more than a reporting tool. It becomes the foundation for explainable, trustworthy, and scalable network automation.