Useful trend data needs stable point identity, intentional collection, accurate units and timestamps, quality control, retention rules, and an operating owner.

Begin with the questions the data must answer

List the actual use cases: verify a sequence, diagnose comfort or process conditions, tune schedules and setpoints, test staging and resets, find simultaneous heating and cooling, support fault detection, calculate energy or water performance, establish a measurement-and-verification baseline, document commissioning, manage alarms, investigate a failure, or plan capital work. For each use case, name the system, point relationships, operating modes, expected duration, required accuracy and resolution, recipients, decision, and consequence of bad or missing data.

This prevents two common failures: collecting everything without a review process and collecting too little to reconstruct the event. Trend one system as a relationship—not a list of isolated values. A chilled-water question may need equipment status, enable command, supply and return temperatures, flow, valve position, pump speed, differential pressure, outdoor conditions, load proxy, setpoints, alarms, and electrical demand on the same clock. The qualified controls and engineering teams remain responsible for design, programming, cybersecurity, and system safety.

Give every point a stable identity and meaning

Build a hierarchy that can survive graphics changes and vendor turnover: portfolio, facility, area, system, equipment, device, point. For each item, record a stable unique identifier, readable name, description, physical location, equipment association, source controller or meter, network object where appropriate, point type, measurement or command role, engineering units, scale and multiplier, expected range, enumeration values, polarity, writable status, sample method, alarm role, tags, and source of truth. Distinguish measured, calculated, commanded, setpoint, status, override, alarm, and quality values.

DOE’s EMIS metadata guidance identifies inconsistent names, spelling and label errors, enumeration differences, hidden parameters, wrong or missing units and scaling, and proprietary naming as barriers to analysis and transfer. Adopt portfolio and project conventions for equipment and point types while preserving the original vendor or device reference. A readable label such as “AHU-2 Supply Air Temperature” helps an operator; a stable ID and structured tags help software retain the relationship after a graphic, room name, controller, or integration changes.

Specify collection, timestamps, and operating context

Choose fixed intervals, change-of-value collection, event capture, or a combination according to system dynamics and the question. Document interval, deadband or change threshold, recording trigger, start and stop conditions, minimum and maximum frequency, precision, aggregation, time zone, daylight-saving treatment, clock source and synchronization, and how missing, late, duplicate, estimated, overridden, out-of-range, or bad-quality values are represented. One universal interval will either hide fast events or create unmanageable noise.

Coordinate points that must be compared so their clocks and collection methods produce an interpretable sequence. UFGS controls guidance provides a useful federal example: points schedules document descriptions, hardware and object identifiers, settings, ranges, network exchange, overrides, trends, and alarms; performance-test data identify the included points and collection basis and can use different intervals for different needs. Those provisions do not establish a universal private-project interval or retention period. The project should set both from its systems, risks, contracts, and use cases.

Govern the historian, integrations, retention, and exports

Document the path from device or controller through the BAS, gateway, meter platform, historian, EMIS, fault-detection tool, analytics layer, dashboard, work-order system, data lake, and report. Record protocol, source ownership, mapping, transformation, aggregation, synchronization method, network-load constraint, failure and retry behavior, latency, quality flags, access, and the boundary between raw and derived data. DOE recommends considering batch synchronization of trend logs where constant polling would burden the controls network.

Set retention by purpose and obligation. Short high-resolution data may support tuning; longer normalized intervals may support seasonal analysis, baseline comparison, maintenance history, compliance, or capital decisions. Identify the owner for each tier, minimum usable period, raw versus transformed record, file format, schema, export cadence, archive, backup, recovery test, privacy or tenant boundary, contractual or regulatory constraint, deletion or legal hold, and migration plan before a vendor or platform changes. Test an actual export and restoration rather than accepting a promise that the data is “in the cloud.”

Operate one quality and change-control loop

Review missing periods, frozen values, impossible ranges, incorrect units or multipliers, clock drift, duplicate or swapped points, failed communications, unstable sensors, overrides, mapping errors, equipment modeled under the wrong system, and calculations that no longer match the installed sequence. Compare trends with field observations, spot measurements, calibration, meter totals, utility records, commissioning tests, operator knowledge, and known operating events. Keep source, date, reviewer, issue, corrective action, verification, and effect on prior analysis.

Control changes to equipment names, graphics, sequences, setpoints, alarms, controllers, sensors, meters, network addresses, gateways, data mappings, integrations, derived calculations, sampling, retention, and software. Record the request, reason, approval, implementation, test, rollback, effective date, affected history, and documents updated. When equipment is replaced, preserve the old identity and time boundary while linking the successor; silently reusing a point name can make years of data look continuous when the sensor, units, location, or calculation changed.

Make data governance part of building operations

Assign accountable ownership across facility operations, controls, information technology, cybersecurity, energy or sustainability, commissioning, vendors, and portfolio management. Define who can view, command, configure, map, export, share, and delete; how credentials and service accounts are controlled; how remote access is approved; how vendor offboarding works; and who responds when the collection pipeline fails. Keep operating documents, sequences, standard procedures, point and asset records, network diagrams, and cybersecurity protocols current as the building changes.

Use a recurring review that connects data quality to action: open control and sensor issues, unresolved alarms, analysis requests, completed corrections, commissioning and measurement results, work orders, savings or performance claims, and capital decisions. DOE’s EMIS technical resources recommend quality checks, issue tracking, and repeated training as needed. The goal is not a perfect taxonomy sitting in a consultant’s file. It is a record that lets the next operator, engineer, lender, equity partner, or owner understand what happened, trust the usable evidence, and know where uncertainty begins.

Primary sources

Sources and technical references.

Source review current as of September 2, 2026.