← Week 6 portalMaster file in the engagement folder:
my-work/week-6/fund-level/library/methodology/materiality-scoring-v1.md
my-work/week-6/fund-level/library/methodology/data-quality-rubric-v1.md
my-work/week-6/fund-level/library/methodology/narrative-voice-v1.md
my-work/week-6/fund-level/library/methodology/anonymisation-convention-v1.md
my-work/week-6/fund-level/library/methodology/scope3-pcaf-method-v1.md

Materiality scoring (house method, v1)

Record ids. D-6-n is a Week 6 decision, J-n a journal entry, F-0n a review finding, DR6-n a data request, A-n an assumption; all are in the engagement record.

Purpose. Decide which register concepts the report must cover in depth. The client's existing materiality assessment is the input; this method converts it, it does not redo it.

Financial materiality (IFRS S2 basis). Score each climate topic, per company, on two axes from 1 to 5:

Score = magnitude x likelihood. Material if 12 or more, or if magnitude is 5. Bands are house defaults, unverified against any client; the owner may reset them per client and the setting is recorded.

Lens test. A topic stays only if it moves EBITDA, reduces deal risk (regulatory, physical, reputational, diligence) or matters to a buyer at exit. Record which.

Extending to double materiality (ESRS). Add an impact score per topic: severity (scale, scope, irremediability, 1-5) x likelihood (1-5). A topic is material for ESRS if either the financial or the impact score crosses its threshold. Keep the two scores in separate columns; never average them. Report to an EU LP which topics are material on impact only.

Output. settings.material_topics lists the concept ids. QA Q03 checks each appears in the coverage matrix and the narrative.

Re-run: re-score only topics whose company, scenario or regulation changed; carry other scores with their original date.

Data-quality rubric (grades A, B, C, v1)

GradeDefinitionExample
AMeasured or primary data from the company's own systems, for the full period, with a named source documentMetered energy or fuel invoices for twelve months; audited revenue
BPrimary data with gaps filled by a documented method, or measured data for part of the period extrapolatedTen months of invoices extrapolated; supplier-specific factor with average activity
CEstimated from proxies: spend-based factors, floor-area benchmarks, sector averages, or modelled outputsSpend-based Scope 2; physical-risk loss from a scenario model

Rules

Narrative voice (house style, v1)

Anonymisation convention (v1)

Scope 3 estimate and PCAF financed emissions: method (v1)

House method for the two figures added at close-out (decisions D-6-61, D-6-64, D-6-65, D-6-66). Both are computed by the intake step (my-work/week-6/pipeline/intake-v1.py), never typed. Every input is a file in my-work/week-6/pipeline/ or the data room.

Scope 3: spend-based estimate (grade C)

What it covers. Operating purchases (mainly GHG Protocol category 1, with 4 and 6), taken together because the cost lines in the data room do not separate them. Capital goods are not captured, since capex is outside operating cost. For the four clinic, office and staffing companies the purchases are mainly rent (closer to category 8, and partly already in Scope 1 and 2). The other categories are not estimated.

Steps, per company.

  1. Annual costs. Take one month of cost lines from portco_financials.csv (revenue less EBITDA), divide by that month's revenue, and multiply by the company's LTM revenue (portfolio_companies.csv, current_revenue).
  2. Less labour. Labour = headcount x the BLS QCEW 2024 average annual pay for the company's sector (bls-average-pay-v1.csv, with the URL of each figure). For companies whose cost of revenue is mostly worker pay (the two staffing businesses, the two clinical businesses and the testing business, D-6-67), labour is the larger of that figure and the cost-of-revenue line. The rule and reason for each company are in scope3-spend-map-v2.csv.
  3. Emission factor. Multiply what remains (purchased spend) by the EPA Supply Chain GHG Emission Factors v1.3 (built on USEEIO; NAICS-6, kg CO2e per 2022 USD, with margins) for the supplier sector named for that company in the spend map. The factor file is useeio-supply-chain-factors-naics-usd2022-v1.csv, downloaded from EPA's data commons (DOI 10.23719/1531143) with its SHA-256 in the record.
  4. Range. Low and high cases set labour at 1.3 times and 0.8 times the BLS pay (benefits on top, or lower pay), which moves the purchased spend. Where the cost-of-revenue line is used as labour the range has no width. The range is a sensitivity to pay only, not a confidence interval.
  5. Fund total. The sum of the company estimates, each weighted by the months Northwood owned the company in 2025, matching the fund Scope 1 and 2 rows.

Known limits. Spend is in 2025 dollars and the factors are in 2022 dollars, with no inflation adjustment, so the estimate is likely high. Purchased spend includes energy bills already counted in Scope 1 and 2, and rent. One supplier sector stands for all of a company's purchases. One month of costs is scaled to a year. Weights are calendar-2025 ownership shares applied to a twelve-months-to-January-2026 estimate; the windows differ by one month. The range reflects the pay assumption only, not factor or spend uncertainty. Clearview's estimate rests on the USEEIO waste management and remediation sector factor (562000), which includes waste-disposal emissions. It is likely high for subcontracted remediation, and Clearview is about 43% of the portfolio total. Keystone's January 2026 cost lines do not reconcile with revenue less EBITDA (the lines give less cost than the difference), so its estimate may be high; the data request covers it.

How to improve it. Data request DR6-16 asks each company for supplier spend by category and for headcount including contractors. With those, replace step 3's single sector with one factor per spend category.

Financed emissions: PCAF Part A, unlisted equity

Attribution factor, per company = fund equity invested / (total equity invested + current debt), from portfolio_companies.csv (USD millions). The data room gives equity at entry and debt today, not a current equity value, so the factor mixes the two dates (D-6-64).

Financed emissions, per company = attribution factor x the company's full-year Scope 1 + 2 (Week 3 ledger). PCAF uses the investee's annual emissions, so part-year companies count for the full year here, unlike the ownership-period fund rows.

PCAF data-quality score: 2 where the company's emissions grade is A or B (primary activity data with published factors), 3 where it is C (proxies). That gives score 2 for seven companies and 3 for three. This is a house mapping not yet checked against PCAF's score tables. Scope 3 of the companies is not attributed (investee Scope 3 excluded).

Totals. The fund row is the sum of the ten companies; Fund I and Fund II rows split the same sum by the fund named in the spine.

Re-run

Update the spine, the financials and, if the sector or rule changes, the spend map; re-read the BLS figures each year and record the new URL; then run the steps in the re-run note. Every number moves with its inputs.

Portfolio edition: Kith Climate cohort case, fictional client.