IPMVP Baseline Statistics — Model Quality & Validation | EC.DATA

Baseline Statistics — Model Quality

Statistical methods for validating the quality and reliability of IPMVP energy baseline models.

Key Statistical Metrics

• R-squared (R²) — proportion of variance explained by the model (target: >0.75)
• CV(RMSE) — coefficient of variation of root mean square error (target: <20%)
• t-statistic — significance of each independent variable (target: >2.0)
• Confidence interval — range around savings estimate at 90% or 95% confidence
• Fractional savings uncertainty — ratio of uncertainty to total savings

Data Requirements

IPMVP recommends minimum 12 months of baseline data, covering all seasons and operating modes. Monthly data needs at least 12 points; daily data improves model precision significantly.

Baseline Statistics in practice

IPMVP baselines stand on regression statistics — CV(RMSE), R², t-stat, autocorrelation. EC.GAIA computes them automatically and refuses to publish a baseline that fails ASHRAE Guideline 14 thresholds.

How EC.DATA operationalises Baseline Statistics

EC.GAIA implements IPMVP option discipline as a workflow: the engineer picks the option (A, B, C, D), declares the boundary, sets the independent variables, and the platform enforces the statistical thresholds (CV(RMSE), R², t-stat) before allowing publication. Baseline Statistics fits into that workflow as a specific stage with its own evidence requirements.

Reports export in IPMVP-conformant format and can be signed by a CMVP using the EC.IAM credential, so customers receive an audit-grade savings report without the partner having to assemble it manually.

Common pitfalls when working with Baseline Statistics

Baseline Statistics M&V failures rarely come from arithmetic; they come from boundary, data quality, or independent-variable choices.

• An option-C model with a CV(RMSE) above 20 % is statistically too noisy to publish — EC.GAIA blocks it.
• Ignoring autocorrelation inflates the apparent confidence interval and produces savings claims that do not survive review.
• Forgetting non-routine adjustments (occupancy change, production volume swings) lets unrelated effects masquerade as savings.
• Stipulated values in option A drift over time; revisit them annually.

Where Baseline Statistics connects across EC.DATA

Baseline Statistics touches every layer of the EC.DATA stack: telemetry capture in EC.Node; visualisation and alerting in EC.EMS with EC.Alerts; tariff translation in EC.Bills; savings verification in EC.GAIA; and field-device fleet governance in EC.IoT. Solution work originates in EC.Solution Design Studio; partner and customer training live in EC.Academy.

Frequently asked questions about Baseline Statistics

How does EC.DATA expose Baseline Statistics to partners?

Baseline Statistics fits inside EC.GAIA's IPMVP workflow; the platform enforces the statistical thresholds before publication.

Do I need a separate license to access Baseline Statistics?

No. Baseline Statistics is part of the core EC.DATA platform; partners get it as part of their standard licence and white-label it under their own brand for their customers.

Where do I learn more about Baseline Statistics on EC.DATA?

Start with the EC.Academy track this page belongs to, then explore the related EC.DATA platform modules linked above. The EC.DATA changelog announces new capabilities and the EC.Academy session catalogue tracks every recorded session.

How EC.DATA applies this in production

The concepts in this lesson are not theoretical — they are operationalised every day inside the EC.DATA platform across deployments in 10+ countries on 3 continents. The module most directly tied to this track is EC.EMS, working alongside EC.GAIA and EC.Bills to translate the underlying physics, protocols, and methodology into a working production system.

Every reading in EC.DATA flows through the same lifecycle: telemetry is captured at the meter or sensor, normalised by the EC.Node edge gateway (which speaks Modbus RTU/TCP, BACnet, OPC-UA, MQTT and pulse counting natively), buffered locally for offline resilience, then delivered to the cloud where EC.EMS stores it as 1-minute resolution time-series. From there, EC.Bills reconciles metered kWh against the utility invoice, EC.Billing allocates consumption to tenants or cost centres, EC.Alerts watches for anomalies, EC.PQ scrutinises waveform quality, and EC.GAIA applies machine learning for forecasting and root-cause analysis.

That integration is what differentiates EC.DATA from the patchwork of disconnected tools most facilities run today. Because every module shares the same data warehouse and the same role-based permission layer, a finding in one module is immediately actionable in another — a tariff change in EC.Bills can adjust demand-alert thresholds in EC.Alerts, a setpoint override in EC.BMS is automatically measured for energy impact in EC.EMS, and an IPMVP baseline is established once and reused across reports forever.

The team behind EC.DATA — described in more depth on the Who We Are page — combines former Fortune 500 energy consultants, field commissioning engineers, and software developers, with a deliberate hiring policy that requires every senior product role to have prior experience on the customer side of an energy programme. The platform is what we wish had existed when we ran those programmes ourselves; the academy is the public-domain version of the training material we built internally to bring new hires up to speed.

If you want to see the platform in action, the free assessment, the savings calculator, and the Solution Design Studio are open without an account; the partner programme is the route in for ESCOs, facility-management firms, commissioning agents, and utilities that want to deliver EC.DATA under their own brand.