GRESB estimation model methodology

This article explains the methodology behind Scaler's GRESB estimation model: how Scaler validates estimated data against GRESB's data estimation rules and, where permitted, fills gaps using linear extrapolation. Read this if you want to understand how the numbers in your GRESB Asset Spreadsheet were built.

When generating a GRESB report with Include estimated data enabled, three modes are available:

GRESB estimations exist for the same reason as Scaler's analytics estimations — filling gaps in meter data — but they play by GRESB's reporting rules and cover all GRESB performance data: energy, water, and waste meters are estimated, and GHG emissions in the report follow from the estimated energy. A GRESB submission covers two years: the reporting year and the year before it (the like-for-like year). The model fills missing data at the edges of those years so an asset's coverage improves, but only as much as GRESB allows:

Estimations are recomputed whenever your data changes, so newly arrived actual data automatically replaces estimates. This works for both calendar-year and fiscal-year reporting.

Scaler enforces GRESB's published estimation rules automatically in the two GRESB-compliant modes. The rule numbering below is Scaler's own, kept consistent across our documentation and the audit report's Explanation(s) tab — GRESB's guidance does not number its rules.

Rule 1 — 20% cap on estimated data. Estimated data must not exceed 20% of the longest continuous period for which actual data is available. Non-contiguous stretches of data are never added together.

Rule 2 — three-month cap. At most three months (90 days) may be estimated across both reporting years combined, even when the 20% rule would allow more. How the budget is split between the two years is free.

Rule 3 — two consecutive reporting years. Estimation operates within pairs of consecutive years: the reporting year and its like-for-like year. A pair only qualifies when both years have data, and both years are filled, keeping the like-for-like comparison complete. Assets without full-year coverage (more than 355 days) in both years are excluded from like-for-like calculations.

Rule 4 — same data type. Estimation may only be based on data of the same performance aspect, area type, and utility type. Scaler guarantees this by construction: each meter is extrapolated exclusively from its own history, so types can never mix.

Rule 5 — overlapping coverage window. When an asset has multiple utility types or meters with differing coverage, data availability is the overlapping period where all of them have actual data. The 20% cap is measured against that window.

Rule 6 — gap in the middle. Scaler only extrapolates the start and end of a year — mid-year gaps are never filled automatically. Closing a mid-year gap with real data or with your own estimate, and choosing which segment to report when a gap cannot be bridged, remains your decision.

Rule 7 — no full-year estimation. A reporting year with no actual data at all may not be estimated from the previous year. Scaler never generates estimates for a year without measured data; the asset reports 0% coverage for that data type instead.

Rule 8 — no prorating across area. Estimates are never derived from other area types, other assets, floor-area proration, or aggregated consumption. Every extrapolation is based on the meter's own measurements only.

Rule 9 — no generic data sources. National benchmarks, statistics, averages, or other sampling methods are never used as a basis for extrapolation — only the meter's own historical readings. Entries you record via Standard consumption (postal code) pass through as your own input under the client-provided estimate modes; their compliance is your responsibility.

Rule 10 — waste tonnage exception. Waste tonnage may be derived from bin count, bin volume, fill level, and pickup frequency; GRESB's minimum-missing-data rules do not apply to waste. Bin-based registration counts as collected data, not an estimate — despite the monitoring method's name, Estimation (Number of bins) is not treated as estimated data. It is never capped, it remains in the report when Include estimated data is toggled off, and Scaler's extrapolation never builds on bin-count-derived values: periods covered by bin registration are neither estimated on top nor used as history for estimating other periods. Tonnage-based waste data (invoices, hauler reports) is gap-filled and capped like any other meter.

Rule 11 — disclosure of estimation use. GRESB requires participants to disclose, per performance indicator, the estimation methodology, the time-based estimated share, and the reason for estimating. The Estimations Audit Report gives you all three: it breaks every reported figure into actual, manual (full and capped), and extrapolated components with a percentage estimated per value. Estimates reported in one GRESB year are expected to be replaced by actuals the next — Scaler handles this automatically, since estimations are recomputed whenever new actual data arrives.

The percentage estimated is time-based, as GRESB requires: estimated days as a share of the days with actual data availability — the same period the 20% cap is measured against. It is not volume-based: a high-consumption estimated month and a low-consumption one contribute equally.

Every GRESB download generated with estimates includes an Estimations Audit workbook (file suffix Estimations Audit.xlsx) breaking each reported figure into actual, capped manual estimates, and linear extrapolation components, with the time-based Percentage estimated per value.

Red-flagged cells indicate that Scaler trimmed a manual estimate that would have exceeded GRESB's caps — the reported value itself remains compliant.

See Understanding the GRESB Estimations Audit Report for the full column reference.