> ## Documentation Index > Fetch the complete documentation index at: https://documentation.kodelabs.com/llms.txt > Use this file to discover all available pages before exploring further. # Baselines and weather normalization > What baselines are in EnerG, how long they should run, and how weather normalization separates real savings from weather swings A **baseline** is a fixed historical reference window you measure future usage against. Weather normalization adjusts that comparison for hot and cold years. Together they let you claim savings that reflect operations, not weather. This page explains both concepts. To create baselines, see [Baseline analysis configurations](/products/energ/analytics/baseline-portfolio-configurations). For parameter definitions, see [Calculations and parameters](/products/energ/reference/calculations-and-parameters#weather-normalization). ## What a baseline is A baseline captures a period of stable, well-metered operation. EnerG compares later usage against it for savings, intensity, and compliance reporting. You set baselines at two levels. * **Portfolio configurations** train weather-normalized models across many buildings at once. * **Building windows** set energy, water, and waste reference periods for one site. ## Baseline length Choose a one-year or two-year window based on data stability and reporting policy. | Length | When to use it | | --------- | ------------------------------------------------------- | | 12 months | Standard reporting; reacts faster to operational change | | 24 months | Smooths anomalies; averages duplicate calendar months | Each category needs at least 12 consecutive months of complete billed data before EnerG accepts a baseline. Partial months or missing commodities disqualify the window until you repair the data. A baseline built on incomplete data biases every savings number that follows. Close gaps before you finalize a window. ## Why weather matters Raw year-over-year comparisons punish cold winters and reward mild ones. A building can run efficiently and still post higher usage in a harsh year. **Weather normalization** removes that noise. EnerG trains a regression model that relates usage to degree days, then adjusts actual usage to a normal-weather equivalent. ## Degree days and base temperatures Degree days measure how far outdoor temperature departs from a base temperature each day. * **Heating degree days (HDD)** accumulate when it is colder than the heating base. * **Cooling degree days (CDD)** accumulate when it is warmer than the cooling base. Base temperatures often sit near 65 °F. EnerG pulls them from [Building information](/products/energ/buildings/building-information) or lets you set custom values per configuration. ## How normalized comparisons read A weather-normalized configuration produces four usage figures you compare on the [Baseline analysis overview](/products/energ/analytics/baseline-portfolio-overview). | Figure | Meaning | | ------------------------ | -------------------------------------------------- | | Baseline usage | Usage from the baseline window | | Actual usage | Measured usage in the reporting period | | Weather normalized usage | Actual usage adjusted to normal weather | | Expected usage | Usage the model predicts for the reporting weather | Model fit reports as an R-squared value. Configurations below your tolerance flag as low correlation and need review before you trust the numbers. ## Next steps Train weather models and set building baseline windows. See how baselines feed M\&V savings and forecasts. Look up weather averages, R-squared tolerance, and base temperature settings. Compare baseline, actual, and normalized usage across the portfolio.