Purpose of this article
This article is the starting point for understanding estimation in Scaler. It explains the two estimation models Scaler uses — meter-level linear extrapolation and the GRESB estimation model — when each one applies, and why the same asset can show different estimated values in each. Each section links to a full methodology article.
In Scaler, "estimated" can refer to data you enter yourself (via the Manual estimate or Standard consumption (postal code) Monitoring method at meter level) or data Scaler generates automatically. This article covers Scaler-generated estimates. For how your chosen monitoring method affects data classification, see Data reliability methodology.
Why Scaler estimates data
Real-world meter data has gaps: readings that cover part of a month, missing months, delayed invoices. Estimation (also called gap filling or extrapolation) fills these gaps with modelled values so that analytics, benchmarking, and framework reporting can work with a continuous dataset. Estimated values are always clearly identified — they never silently replace measured data.
Estimation is different from normalization: estimation fills missing data, normalization adjusts existing data for comparability. See Normalization methodologies in Scaler.
The two estimation models
Meter-level linear extrapolation | GRESB estimation model | |
What it does | Fills gaps in energy meter data using a weighted average of that meter's recent history | Validates and fills gaps in reported data according to GRESB's official estimation rules |
Resource types | Energy meters only | Energy, water, and waste meters (GHG emissions follow from the estimated energy) |
Where you see it | Analytics Portal (Estimated view) and Data Collection Portal monthly exports | GRESB Asset Spreadsheet and the accompanying Estimations Audit Report |
Governing logic | Scaler's own methodology: 6-month lookback, recency-weighted daily averages, confidence scoring | GRESB's published estimation rules, including the 20% cap and 3-month cap |
Read more |
Meter-level linear extrapolation
Scaler's default estimation model for analytics and data exports. It detects missing periods at the meter level, looks back up to six months of actual consumption, and fills gaps using a recency-weighted daily average. Every estimate carries a confidence score.
You'll find this data in Analytics Portal → Portfolio → Performance → Energy (select View → Estimated (Scaler algorithm)) and in the estimated_linear_extrapolation_consumption column of the monthly export from Data Collection Portal → Portfolio → Meters & Consumption.
Full methodology: Meter-level linear extrapolation methodology
GRESB estimation model
A separate model that applies only when generating a GRESB report with the Include estimated data toggle enabled. Instead of optimizing for analytical continuity, it enforces GRESB's data estimation rules — including the 20% cap on estimated data and the three-month cap across both reporting years combined — and can optionally fill remaining gaps with linear extrapolation up to those limits.
Every GRESB export generated with estimates includes an Estimations Audit Report showing exactly how each reported figure was built.
Full methodology: GRESB estimation model methodology
Workflow context: Completing a GRESB submission using Scaler
Why the two models produce different values
The same asset and period can show different estimated values in analytics than in a GRESB report. This is expected, not an error:
- The meter-level model optimizes for a complete, continuous time series and fills every gap it can.
- The GRESB model is constrained by GRESB's rules: estimates are capped at 20% of the longest continuous period of actual data, no more than three months may be estimated across both reporting years combined, and non-compliant, client-entered estimates are excluded rather than passed through.
Do not reconcile analytics estimates against GRESB report values one-to-one. If you need to trace how a GRESB figure was built, use the Understanding the GRESB Estimations Audit Report article.
Estimates you enter vs estimates Scaler generates
Every meter has a Monitoring method that describes how its consumption data is collected. Values recorded under these methods are entered by you — Scaler does not generate or modify them. The monitoring method also determines how a meter's data translates to actual or estimated in framework reports such as INREV SDDS:
- Measured methods (e.g.
Smart meter,Invoice,Conventional meter,Standard consumption (cluster average)) translate to actual data.
- Modelled methods (e.g.
Manual estimate,Standard consumption (postal code)) translate to estimated data.
The estimates generated by Scaler's two estimation models — meter-level linear extrapolation and the GRESB estimation model — are separate from and additional to this client-entered data. For classification and scoring detail, see Data reliability methodology.
Additional resources
- Meter-level linear extrapolation methodology — full model logic, confidence scoring, and limitations
- GRESB estimation model methodology — GRESB rules and how Scaler enforces them
- Completing a GRESB submission using Scaler — end-to-end GRESB workflow
- Understanding the GRESB Estimations Audit Report — trace how each reported figure was built
- Normalization methodologies in Scaler — how normalization differs from estimation
- Data reliability methodology — monitoring method classification and scoring