Energy data governance for multi-site organizations

A single building has a handful of meters and one or two utility accounts. A portfolio of 80 sites can have several hundred accounts spread across dozens of utilities, multiple commodities, and a mix of monthly bills, interval feeds, and manual spreadsheets. When someone asks a simple question, such as how much electricity the portfolio used last quarter, the answer depends on which accounts were counted, whether estimated bills were included, and how units were converted. Different teams often produce different numbers from the same underlying data.

That is a governance problem, not a spreadsheet problem. Energy data governance is the set of rules, roles, and definitions that make utility data trustworthy across an organization: who owns each data set, what each field means, how quality is checked, and who is allowed to see or change it. Without it, every report starts with an argument about the numbers.

What energy data governance actually covers

General data governance rests on a few well understood pillars, and energy data is no exception. Industry practitioners describe the core elements as accountability, common definitions, quality management, lineage, and controlled access. Applied to utility data, they look like this.

• Ownership and accountability. Every account, meter, and site has a named owner responsible for the completeness and accuracy of its data.
• Common definitions. A shared dictionary states exactly what a 'site', a 'meter', and a 'billing period' mean, so two teams counting the same thing get the same total.
• Data quality rules. Automated checks flag estimated reads, gaps, unit mismatches, and outliers before the data reaches a report.
• Lineage. Every number can be traced back to the source bill or interval file it came from.
• Access and security. People and systems get only the data they need, and changes are logged.

The cost of skipping these steps is not abstract. Gartner has estimated that poor data quality costs organizations an average of 12.9 million dollars per year across all the decisions it touches. In an energy program, bad data shows up as budget forecasts that miss, emissions reports that cannot be assured, and efficiency projects justified on the wrong baseline.

Why the numbers do not reconcile in a portfolio

Most reconciliation failures in multi-site portfolios trace back to a short list of recurring issues. Naming them makes them easier to design against.

• Duplicate or orphaned accounts. A site changes tenants or the utility restructures an account, and the old account keeps flowing data alongside the new one.
• Estimated bills counted as actual. A meter is not read, the utility estimates, and the estimate is later trued up. If both versions land in the data set, usage is double counted.
• Unit and commodity mismatches. Gas billed in cubic meters, therms, or gigajoules; electricity in kWh at some sites and MWh at others. Without normalization, sums are meaningless.
• Missing meters. A submeter or a newly acquired building is never added, so a site looks more efficient than it is.
• Organizational drift. Buildings move between regions or business units, and last year's rollup no longer matches this year's org chart.

None of these are exotic. They are the normal entropy of a large portfolio, and governance is what keeps them from accumulating into numbers no one trusts.

A data model that survives reorganizations

The durable pattern is to separate the physical facts of energy delivery from the reporting structures that sit on top of them. Meters and accounts rarely change. Regions, business units, and reporting groups change often. If your data model ties consumption directly to this year's org chart, every reorganization breaks the history. If it ties consumption to stable meter and account identifiers, you can re-roll the same underlying data into any structure you need.

Ownership and stewardship

Governance works when responsibility is assigned, not assumed. A data steward is the person accountable for the quality and fitness for purpose of a data set, including its metadata. In an energy program, stewardship is usually split: a central team owns definitions and quality rules, while site or regional staff own the accuracy of their own account lists. The central rules keep everyone consistent, and local ownership keeps the data current.

Quality rules that catch problems early

The cheapest place to fix bad data is at ingestion, before it spreads into dashboards, forecasts, and disclosures. A practical quality layer applies a standard set of checks to every bill and interval file as it arrives.

• Flag estimated reads and reconcile them when the true-up bill arrives.
• Detect gaps and overlaps in billing periods per meter.
• Convert every commodity to standard units and validate against expected ranges.
• Compare each period to the site's own history, adjusting for weather where relevant, and flag outliers for review.
• Check that every active meter reported data for the period, and surface any that did not.

These checks turn data quality from a periodic cleanup project into a continuous, mostly automated process. The goal is not zero exceptions; it is that every exception is visible and assigned to someone.

Lineage and access

When an auditor, a finance controller, or an executive questions a number, the fastest way to resolve it is to trace the number back to its source. Data lineage is the record of data's journey from ingestion through transformation to the report, and it is what lets you show that a quarterly total is the sum of specific bills, with any estimates and adjustments clearly marked. Lineage also makes it safe to change a rule, because you can see everything downstream that the change will affect.

Access control is the other half. Utility data can include account numbers, addresses, and, for organizations that bill tenants, information tied to individuals. Governance defines who can view, export, and edit each data set, and keeps a log of changes so the data has an accountable history.

Governance is a process, not a project

The common failure is to treat governance as a one-time cleanup: a consultant reconciles the account list, publishes a data dictionary, and a year later the numbers have drifted again. Governance holds only if it is continuous. New sites are onboarded to the same standard the moment they join. Account changes from utilities are reflected as they happen. Quality exceptions are worked every period rather than at year end. The point is not a perfect data set on one day; it is a data set that stays trustworthy as the portfolio changes.

This is also where automation earns its place. Applying definitions, normalizing units, flagging estimates, and maintaining lineage by hand across hundreds of accounts is not sustainable, and manual handling is itself a source of the errors governance is meant to prevent. The more of the routine work a system handles consistently, the more your people can focus on the judgment calls that genuinely need a human: an unusual site, a disputed charge, a change in reporting structure.

How MartinAI fits

MartinAI collects bills and interval data across any commodity and any utility, then applies the normalization and quality checks described above automatically. Every meter, account, and site sits in one model, so a reorganization re-rolls the same governed data rather than breaking the history. Estimated reads are flagged and reconciled, units are standardized, and every figure keeps a link back to the source document it came from. Site owners maintain their own account lists inside a central framework, which is the split that keeps large portfolios both consistent and current.

The result is that the simple questions have simple, repeatable answers. When someone asks how much the portfolio used last quarter, everyone gets the same number, and the number can be traced to the bills behind it.