How the recorded data source of each month changes PropScore's estimated share and GRESB export verdict

Question

PropScore records a 「データ取得方法」 (data source) for each month of energy data. As more months are recorded as something other than measured, how do the 「推計割合」 (estimated share) on the emissions screen and the verdict on the GRESB export screen change? Consumption is identical for every property and every month, so only the data source differs.

Method

• Product and screens: PropScore's property list, a property's energy data (CSV import), the emissions screen and the GRESB export screen. The interface language was switched to Japanese for the recording, and the screen text is quoted as displayed. The language switch is a browser-side setting and changes no product data.
• Hypothetical inputs: under the existing hypothetical portfolio 「架空ポートフォリオ 研究ノート」, we created 5 hypothetical properties named 「架空物件 k=…」 (hypothetical property k=…), with k = 0, 1, 3, 6, 9. All share the address 「東京都千代田区」 (ward level only), use オフィス (office), structure RC, gross floor area 1000 m², year built 2000, 5 floors above ground and acquisition date 2020-04-01; only the name differs. None of them is a real building.
• Monthly data: every one of the 12 months of 2025年度 (fiscal year, month 4 to month 3 of the following year) got 10000 kWh of electricity. For each property, the data source of the last k months was 「前期実績流用」 (prior-period figure reused) and of the other months 「実測」 (measured); every row also carried a 「推計理由」 (reason). Consumption is identical by construction.
• Input route: the monthly grid on screen cannot edit the data source, so we imported a CSV (60 rows) with the columns 「物件名」「対象月(YYYY-MM)」「電力(kWh)」「データ取得方法」「推計理由」 through 「CSVインポート」 on the energy data screen. The screen showed 「60件インポート完了」 (records imported). After the import we opened each property's monthly grid for 2025年度 and checked that each month's data source was displayed as entered.
• Operations: on the emissions screen we selected 2025年度 before pressing 「排出量計算」 (calculate emissions), and recorded each property's estimated share. We then selected the same year on the GRESB export screen, recorded what it displayed, and opened 「プレビュー」 (preview). We did not download the workbook or press 「新規提出」 (new submission).
• Date and production revision: 2026-10-15. Cloud Run revision propscore-00030-vhj (commit 3fecf87). The revision was the same before and after the recording. The results are the output of that production revision.

We confirmed the product's rules by reading the code at that commit. What follows describes PropScore's rules, not GRESB's own rules.

• Estimated share: 1 −（実測の月数 ÷ 12）, i.e. one minus the number of measured months divided by twelve. It counts months per property and year, with no weighting by consumption, and a month with no data counts like an estimated one.
• Every source other than measured counts the same: 「請求期間按分」, 「前期実績流用」, 「床面積按分」 and 「不明」 are all counted as estimated in the same way.
• Display: the stored share is shown as a percentage with 1 decimal place. Beneath it appears 「推計なし（全月実測）」 (no estimate, all months measured) when every month is measured, and 「入力値が推計」 (inputs estimated) when every month has data and at least one is not measured.
• GRESB export verdict (per property and year): blocked when the estimated share exceeds 0.20 (rule code R2_CEILING_EXCEEDED); blocked when the estimated months (share × 12, rounded half up) of this year and the prior year together exceed 3 (R3_TWO_YEAR_CAP). The prior year counts only when a prior-year emission estimate exists.
• When there is no data: the estimated share counts a month with no data like a month not measured (the denominator is always 12 months). The estimated-months rule, by contrast, reads the previous year's emission estimate rather than its energy data, and counts the previous year as 0 months when there is no estimate. The last item under 「この確認で保証されないこと」 on the screen said the same: 「2年間の上限は、PropScoreに存在する前年度のデータのみで評価します。前年度が無い場合は当年度のみで計算するため、実際の件数より少なく出ることがあります。」
• The workbook is per portfolio: if any property-year is blocked, no workbook is produced for the portfolio. The 「Excelダウンロード」 (Excel download) button can be used only when the screen's 「全体データ完成度」 (overall data completeness) is at least 50% and no property is blocked.
• The preview is read-only: opening it records nothing in the submission history.

Results

The emissions screen.

Property	Months not 実測 in the grid	推計割合 (estimated share)	Label under the share
架空物件 k=0	0	0.0%	推計なし（全月実測）
架空物件 k=1	1	8.3%	入力値が推計
架空物件 k=3	3	25.0%	入力値が推計
架空物件 k=6	6	50.0%	入力値が推計
架空物件 k=9	9	75.0%	入力値が推計

The GRESB export screen (same year).

Property	List of property-years that cannot be submitted	Rule codes shown in the list
架空物件 k=0	not in the list	—
架空物件 k=1	not in the list	—
架空物件 k=3	推計割合 最低 25.0%	R2_CEILING_EXCEEDED
架空物件 k=6	推計割合 最低 50.0%	R2_CEILING_EXCEEDED, R3_TWO_YEAR_CAP
架空物件 k=9	推計割合 最低 75.0%	R2_CEILING_EXCEEDED, R3_TWO_YEAR_CAP

• The list was headed 「3件の物件年度はGRESBに提出できません」 (property-years that cannot be submitted to GRESB), followed by this text: 「これらの数値は、GRESB 2026年版ルールが禁止する手法で作成されたか、上限を超えています。修正されるまでExcelエクスポートは無効です。提出用ワークブックは全件一括であるため、1件でも不可の物件があれば何も出力しません。」
• Each list item showed the rule detail in English. The item for 「架空物件 k=6」, for example, read: "R2_CEILING_EXCEEDED: 50.0% of the reporting period is estimated, above GRESB's 20% ceiling; R3_TWO_YEAR_CAP: 6 month(s) in FY2025 alone exceeds GRESB's hard cap of 3 months across both reporting years"
• At the top of the screen, 「全体データ完成度」 (overall data completeness) read 「100%」, and below it 「データ完成度が50%未満のためエクスポートできません」 was displayed (export not possible because data completeness is below the threshold).
• In 「プレビュー」 (preview), 「全体データ完成度」 read 「39%」 and the status label read 「準備中」 (in preparation). After the preview was closed, the 「提出履歴」 (submission history) read 「提出データがありません」 (no submissions).

How to read this

• Consumption is identical by construction, so the differences in the tables come from the data-source input alone.
• The property with 1 month not measured (estimated share 8.3%) did not appear in the list of property-years that cannot be submitted; the properties with 3 or more such months did. With these inputs, the boundary we observed lies between k=1 and k=3; values of k in between were not measured.
• For k=3 only the estimated-share rule code was shown; for k=6 and 9 the estimated-months rule code was shown as well.
• 3 properties were listed, but because the workbook is per portfolio, no workbook for this portfolio's 2025年度 could be produced, including the properties that were not listed.
• In three places the displayed text and the rule in the code differ.
  • 「データ完成度が50%未満のためエクスポートできません」: the code shows this sentence whenever download is not allowed, whatever the reason, and a blocked property alone stops the download regardless of completeness. Here the reason was a blocked property and the completeness shown was 「100%」, so the sentence and the code's reason differ.
  • The first item under 「この確認で保証されないこと」 (what this check does not guarantee): 「estimated_ratio は「何か月分のデータが存在するか」から算出されるため、システムが外挿した割合を測っており、入力そのものの質は測っていません。顧客が自ら推計して入力した月も、ここでは実測として数えられます。したがって以下の割合はすべて下限値です。」 It says that months a customer estimated are counted here as measured. The code counts any month whose data source is not measured as estimated. The months entered as 「前期実績流用」 were counted as estimated, so the item and the code's treatment differ.
  • 「全体データ完成度」: the value at the top of the screen is the average share of months with data per property, while the value in the preview is the share of indicators that have data. The same heading shows two different quantities.
• Emission values are outside the scope of this note.

Limits

• All inputs are hypothetical; this is not a result for any real property or real report.
• We covered 1 portfolio, 2025年度 and electricity only. No energy data was entered for the previous fiscal year and no emission calculation was run for it. The estimated-months rule (R3_TWO_YEAR_CAP) counts the previous year as 0 months when the property has no previous-year emission estimate. The rule appeared for k=6 and 9 because the current year's estimated months alone exceeded the cap of 3 months, and the rule detail says so ("alone"). The part of the rule that adds a previous year was not exercised by these inputs.
• The values of k measured were 0, 1, 3, 6, 9 only.
• The only non-measured data source used was 「前期実績流用」. That the other sources count the same way was confirmed in the code, not measured on screen.
• We did not press 「Excelダウンロード」, and we did not confirm from the screen whether the button could be pressed.
• The screen text refers to GRESB's rules; we did not check whether it matches GRESB's own rules. This note consulted no GRESB document and is not advice on whether to submit.
• The verdicts are those of the production revision at the time of measurement and may change with product updates.

How to reproduce

• Sign in to PropScore, choose a portfolio and create properties (required fields as above).
• Prepare a CSV with the columns 「物件名」「対象月(YYYY-MM)」「電力(kWh)」「データ取得方法」「推計理由」 and import it through 「CSVインポート」 on a property's energy data screen. Do not leave the data source or the reason blank.
• On the emissions screen, select the fiscal year and press 「排出量計算」.
• On the GRESB export screen, select the same year, read the screen and open 「プレビュー」.

Data attribution

• All inputs in this note are hypothetical data created by the author; no value from an external dataset is published.
• The figures published are values displayed on PropScore's screens.
• See the PropScore documentation; where each DXA Labs product gets its data is summarised in Data sources and measurements (Japanese).