The MDM sits between the Head-End System and the utility’s business systems. It is where raw meter data becomes reliable, validated, billable information. Getting the MDM selection and architecture right determines whether your AMI investment delivers ROI.
The MDM’s Three Core Jobs
1. VEE — Validation, Estimation, and Editing
Raw meter data is dirty. Gaps from communication outages, register rollovers, meter replacements mid-period, and clock errors all require handling:
- Validation: Rule-based checks — consumption > 0, consumption < physical maximum, no negative deltas, status bits checked
- Estimation: For missing intervals, substitute using historical profile, calendar-weighted average, or metered proxy (neighboring meter in same DMA)
- Editing: Analyst override with audit trail. Every manually edited interval must be flagged permanently so downstream systems know the data provenance
2. Data Repository
An MDM stores 2–5 years of interval data for every meter. For 500,000 meters at 15-minute intervals with 4 channels, that is ~1.7 billion rows/year. Time-series database design is non-negotiable — row-oriented databases (SQL Server, MySQL) fail at this scale.
3. Business System Integration
The MDM distributes validated data to:
- CIS/Billing — interval reads for TOU billing
- GIS — spatial asset data enrichment
- SCADA/OMS — outage detection from last-gasp events
- Analytics platform — load forecasting, loss analysis, demand response
Integration uses IEC 61968 CIM (Common Information Model) or REST APIs. Mature MDM vendors publish a CIM conformance profile.
MDM Architecture Patterns
Centralised MDM
Single MDM instance handles all meters, all utilities (for multi-utility). Simple to operate, single vendor relationship. Risk: single point of failure, vendor lock-in.
Federated MDM
Separate MDM instances per utility or per region, with a thin orchestration layer. More resilient, but more complex data governance.
Cloud-Native MDM
The 2024+ trend: MDM as a SaaS platform (Oracle Utilities OUDA, Itron EE, VertexOne). Eliminates infrastructure management. Requires robust network connectivity and data residency compliance review.
Leading MDM Vendors
| Vendor | Strength | Typical Market |
|---|---|---|
| Oracle Utilities MDM | Deep CIS integration, large utility deployments | Tier-1 electric utilities, NA/Europe |
| Itron Enterprise Edition | Native Itron HES integration, water/gas/electric | Multi-utility, global |
| Honeywell Temetra | Multi-utility, strong water/heat focus | Water/heat utilities, Europe |
| Siemens eMeter | Grid-edge analytics, TOU billing | Large electric utilities |
| VertexOne (DataPlatform) | Cloud-native, fast deployment | Mid-market utilities, NA |
| Trilliant | Communication-agnostic, open API | Greenfield AMI deployments |
Key Questions When Selecting an MDM
- What is the tested scale (meters × intervals × years) in production references?
- What VEE rules are configurable vs hardcoded?
- Is there a published CIM conformance profile?
- What is the HES integration mechanism — file drop, real-time API, or direct DB?
- How are audit trails stored and queried?
- What is the cloud/on-premises deployment model?
Frequently Asked Questions
What database architecture should we use for storing 2-5 years of interval meter data at 15-minute resolution?
Time-series database design is non-negotiable; row-oriented databases like SQL Server and MySQL fail at the scale of ~1.7 billion rows/year for 500,000 meters. Purpose-built time-series databases are required to handle the volume and query patterns typical of MDM deployments.
How does an MDM handle missing meter intervals caused by communication outages or meter rollovers?
The MDM’s VEE (Validation, Estimation, Editing) function substitutes missing intervals using historical profile data, calendar-weighted averages, or metered proxy data from neighboring meters in the same DMA. All manually edited intervals must be flagged permanently with audit trails so downstream billing systems can track data provenance.
What validation rules should be applied to raw meter data in the VEE process?
Standard rule-based checks include consumption > 0, consumption < maximum register capacity, and rate-of-change thresholds to catch spike anomalies. The MDM should flag data that fails these rules for estimation or analyst review rather than allowing dirty data to reach billing systems.
Which integration standard should we require our MDM vendor to support for business system connectivity?
IEC 61968 CIM (Common Information Model) is the preferred standard for distributing validated data to CIS/Billing, GIS, SCADA/OMS, and analytics platforms; mature vendors should publish a CIM conformance profile. REST APIs are an acceptable alternative for greenfield deployments.
What are the trade-offs between centralized, federated, and cloud-native MDM architecture patterns?
Centralized MDM is simple to operate with a single vendor relationship but creates a single point of failure; federated MDM is more resilient but adds data governance complexity; cloud-native (SaaS) eliminates infrastructure management but requires robust network connectivity and data residency compliance review.