ISO 55013 Changed the Rules on Asset Data Quality. Most EAM Architectures Aren't Ready.

When ISO published ISO 55013 in 2024 alongside the revised ISO 55001, it was the first time the international asset management standard family addressed data quality as a formal governance obligation — not just a best practice. ISO 55013 specifies how organisations should manage asset data to support decision-making, including the quality controls, lineage tracking, and governance structures required when AI drives maintenance recommendations.

For organisations managing a mixed portfolio of infrastructure, the standard surfaces a specific and uncomfortable question: who governs the data for assets that sit at the boundary between your two EAM systems?

The boundary where data quality deteriorates

In most utilities and industrial operators with mixed portfolios, linear infrastructure — pipelines, distribution networks, transmission lines, rail corridors — lives in one data environment. It is GIS-native, spatially organised, and modelled as a network. Discrete assets — pump stations, substations, treatment plants, maintenance depots — live in another system: hierarchical, organised around equipment classifications, maintenance intervals, and component genealogy.

This split reflects a genuine representational difference. Linear assets require network topology to express condition and risk; discrete assets require parent-child relationships and equipment taxonomies to do the same. Two different structures, two different systems, two different data models.

The boundary between them is where data quality deteriorates fastest. Compressor stations on a gas transmission network. Pump stations on a water distribution system. Substations on a transmission grid. These boundary assets appear in both systems — often incompletely, sometimes with conflicting records, almost always without a shared asset identifier.

A predictive model drawing on incomplete or inconsistent data from boundary assets is producing recommendations with an unacknowledged confidence gap. The model doesn't flag what it doesn't know. Neither, typically, does the operator.

What ISO 55013 actually requires

ISO 55013 is organised around a data management framework that covers data strategy, data quality, data governance, and data architecture. For AI-assisted maintenance specifically, three requirements stand out:

Data completeness. All data inputs to a maintenance decision model should be identified, and gaps documented and actively managed — not left to accumulate at system boundaries.

Data lineage. The provenance of data feeding AI recommendations should be traceable: where it originated, how it was processed, and when it was last validated against physical reality.

Data governance. Clear ownership of data quality should exist across the full asset portfolio, with defined accountabilities at each stage of the asset lifecycle.

For a mixed-portfolio operator running separate systems, none of these requirements can be fully satisfied at the boundary. Take Compressor Station 7 on a hypothetical gas transmission network. Does the pipeline operations team own its condition data, or the facilities maintenance team? In a two-system architecture, the honest answer is frequently neither — ownership defaults to whoever last updated it, in whichever system they were logged into that week.

ISO 55013 does not offer a workaround for this. It requires accountable ownership. The data governance gap at the boundary is not a process problem that training will fix; it is an architectural consequence of managing one portfolio across two incompatible data models.

Why AI amplifies the problem

The shift toward agentic EAM — AI agents reasoning autonomously across maintenance workflows rather than surfacing advisory dashboards — amplifies the data governance problem rather than dissolving it.

An AI agent optimising inspection schedules across a gas distribution network needs to evaluate compressor station condition as part of the network state. Not merely "is the station operational?" but "what is the degradation trajectory of its primary compression units, and how does that change the network's redundancy capacity over the next maintenance window?" That question requires current, quality-controlled condition data from the discrete asset system, linked to a live network topology model.

If that data is incomplete or structurally incompatible with the network data, the agent is not optimising — it is generating recommendations at false confidence levels, with no mechanism to surface what it is missing.

ISO 55013's data quality requirements do not distinguish between human-driven and AI-driven decisions. They apply equally to both. If anything, the case for quality-controlled, lineage-tracked asset data is stronger when AI is making consequential recommendations — because an agent cannot exercise the informal scepticism a senior engineer brings when a recommendation feels off.

The governance fix requires an architecture fix

Solving the data quality problem at the linear-discrete boundary is not a matter of better ETL pipelines or tighter API contracts between two legacy systems. Synchronisation can align records; it cannot resolve the underlying representational conflict between a network graph and a component hierarchy. Boundary assets will continue to be governed by whichever system last touched them.

A platform that natively holds both representations — where a pump station at the network terminus is simultaneously a node in the network graph and a root element in an equipment hierarchy — can apply consistent data governance across both contexts. Completeness checks, lineage tracking, and quality controls operate on a single asset record, not on two partial approximations of it in two systems.

For a deeper look at why linear network data and discrete asset hierarchies require different but complementary data structures, see our knowledge guide on linear asset management.

The practical implication for ISO 55001:2024 alignment

The 2024 revision of ISO 55001 and the companion ISO 55013 together raise the bar for what "good" asset data governance looks like. Organisations undertaking recertification or initial alignment in 2025 and 2026 are encountering data quality requirements that most EAM implementations — particularly those built up over years across two or more systems — were not designed to satisfy.

The question is not whether to invest in better data governance. Most asset-intensive organisations know they need it. The question is whether the platform architecture they are governing can actually deliver unified, quality-controlled data across all the asset types they operate.

For operators managing both network infrastructure and facility assets, the answer to that question starts with the data model, not the governance policy written on top of it.

Curious how UniEAM handles both? Request a demo.