methodology-note-v6.md, sha256 5293ebd455e3). Portfolio edition: Kith Climate cohort case, fictional client. Small grey tags such as J-081 point to entries in the project record (journal, findings, decisions), so every statement can be traced.Methodology note: how Northwood's carbon inventory is calculated
Version 6, 5 October 2026 (replaces v5; names every number decision taken with results in view, finding F-128. Earlier: v5 replaced v4; adds the ledger lineage gaps to the known limits, finding F-124. Earlier: v4 replaced v3; corrects the statement that no result could steer the method, finding F-113, and lists the number decisions made after results were visible. Earlier: v3 stated that dashboard v14 carries the roster label that v13 lacked, finding F-110; v2 and v1 are in _archive. Earlier: v2 replaced v1; adds the traceability pass, in which each statement that rests on a fact, number or rule carries a tag such as M-07 pointing to its row in the claims register, and a separate check re-tests every row against the record). One of three leave-behind documents, with the reference data pack (where every outside number comes from) and the re-run note (how to update everything next quarter).
The test this note is written to pass: if Northwood had to reproduce or update the inventory next quarter with no access to Kith, could they, using only what is left behind? This note says what is calculated and why. It does not repeat the numbers; the dashboard and the calculation ledger hold those. Where a rule is a judgment call, the note says whose judgment and where it is recorded.
1. What the inventory covers
- What is counted. Scope 1 (fuel burned at the company's sites and in its vehicles, plus refrigerant that leaks from its equipment) and Scope 2 (purchased electricity), for calendar year 2025, for the ten portfolio companies. Each company's own offices, plants, clinics, branches and fleet are counted because it controls them (operational control, rule R12). M-01, M-02
- How the fund total is made. The fund figure counts 100 percent of each company's emissions, for the months Northwood owned it (rule R3): Harvest Ridge Foods 9 months (closed 1 April 2025) and Ridgeline Property Services 6 months (closed 1 July 2025). Every other company is counted for 12 months. Each company's own page shows its full-year footprint; the fund roll-up uses the ownership-period share. Both are labelled wherever they appear. Equity percentage is not used. M-03, M-04, M-05, M-06, M-07
- How to describe the fund total. Under the GHG Protocol, an investor's share of its holdings' emissions is investment-related Scope 3 (category 15). The deliverables therefore call the fund figure "the sum of the companies' Scope 1 and 2 inventories". M-08
- What is left out, on purpose.
- Buildings Ridgeline manages for other owners are other owners' emissions. They are reported separately and never added to Ridgeline or to the fund (rule R8). M-09
- Sites the companies' own records name but the data room does not list have no data, so they are not in the base total and no estimate is made for them (rule R11). The count at the time of the build was 27 (Patriot 10, NovaCare 15, Trident 1, Ridgeline 1). Every deliverable states this as a known understatement. M-10, M-11, M-12
- Fleet methane and nitrous oxide are left out of the base because the data has no vehicle type, model year or mileage. A bound is computed (section 6) and is small against the fleet carbon dioxide. M-13, M-14
- R-22 refrigerant (HCFC-22) is reported on its own line outside Scope 1, because HCFCs are controlled under the Montreal Protocol and the GHG Protocol asks for them to be reported separately. M-15, M-16, M-17
- Scope 3 is not calculated. It is flagged (section 9). M-18
- Scope 2 is location-based only (decision D-G1-10): it uses the average emissions of the regional electricity grid. The re-run note says how to add the market-based figure. M-19
2. Conventions that apply everywhere
- The report date is a setting, not a fixed fact. It sits in the configuration file (
pipeline-config-v3.json,staleness.report_date). The current value is 2 October 2026. A data point's age is measured from the end of the period it covers: 31 December 2025 for the 2025 collection template, 31 January 2026 for the revenue, EBITDA and headcount figures. A figure older than 90 days at the report date carries the flag[as of YYYY-MM-DD -- update pending](Northwood's own rule, kept). Emission factors and other reference data are not flagged; they carry their edition in the citation. The sample data room is fictional data reused from an earlier cohort, so its own "today" is about 31 January 2026; when Northwood re-runs with real data, they enter their actual report date and the flags behave as intended. M-20, M-21, M-22, M-23, M-24, M-25, M-26, M-27 - Confidence labels. Each figure is Confirmed, Reported or Inferred (rule R7). Reported means the company supplied it. Inferred means we derived it under a stated rule. A total takes the weakest label among its inputs. M-28
- Source order when files disagree (rule R2): the company's own operating records first, then the Northwood data spine where it agrees with them, then management's questionnaire and the carbon template. Where two acceptable sources differ by more than 5 percent, both are shown. M-29, M-30, M-31
- Alternatives are shown beside the base, never instead of it. Anything that rests on a judgment call has a pre-registered alternative (section 6). M-32
- Rules were fixed before any emissions were calculated, but not every later choice was. The build rules (R1 to R14) were written down and locked on 2 October 2026 before any emissions were calculated. After the first calculation some choices were made with results in view: the treatment of Patriot's fleet gasoline (D-G1-6, which cites its calculated tonnes), the method-review rulings (D-G1-7), the treatment of Sterling's fleet diesel (D-G1-9), the carbon-price date and the energy price shock (D-R14-A1 and D-R14-A2, fixed after the ledger existed and before the exposure figures were seen) and the rank-band tolerance (D-G3-2 and D-G3-2-A1, set after the ranks were visible). Each one is in the decision log with its reason, so a reader can see what was chosen after the numbers were visible. The decision log records who decided each one and why. M-33, M-34, M-35, M-148, M-149, M-150, M-155, M-156, M-157, M-158, M-159
3. The calculation, step by step
Every calculation line is one site and one fuel or energy type. The template columns are electricity in kWh, natural gas in therms, diesel, gasoline and propane in gallons, and fleet diesel and fleet gasoline in gallons. For each line: M-36
- Read the number the company reported from the carbon collection template, after the repair and correction rules in section 4.
- Convert it to the unit the factor uses. Natural gas: 1 therm = 0.1 million Btu (mmBtu). Electricity: kWh divided by 1,000 gives MWh. M-37, M-38
- Multiply by the emission factor for that fuel or grid region. Each factor has a carbon dioxide, methane and nitrous oxide part. Fleet fuel uses the mobile carbon dioxide factor only (see section 1). M-39
- Convert methane and nitrous oxide to carbon-dioxide equivalent (CO2e) with the global warming potentials: methane 28 and nitrous oxide 265 (IPCC Fifth Assessment, 100 years). Add the three parts. M-40, M-41
- Record the line with its data citation, its factor citation, the rules and decisions that applied, its grade and its confidence label. Values are kept to six decimal places in the ledger; rounding happens only when figures are displayed. M-42
Electricity factors depend on where the site is. Each site is placed in an eGRID subregion by finding which subregion boundary contains its coordinates (decision D-P2). A site in no boundary takes the nearest one if it is within 5 kilometres, and is flagged as assigned by fallback; beyond 5 kilometres it stays unassigned and is listed as a data gap. M-43, M-44, M-45, M-46
Refrigerant leakage is counted in the base in two ways (rule R9). Where the company has a service record, the recharge amount is the leakage (graded A, Reported). Otherwise the leakage is the equipment charge times the EPA default annual leak rate for that equipment type (the screening method; graded C, Inferred). The CO2e is pounds leaked times 0.45359237 kilograms per pound times the refrigerant's GWP. The full charge is shown beside the total as an upper bound and never enters the base. M-47, M-48, M-49, M-50, M-51, M-52
Three worked examples from the ledger (checked against the file on 4 October 2026)
| Natural gas, Sterling site SPM-001 | Electricity, Sterling site SPM-001 | Refrigerant, NovaCare site NC-028 | |
|---|---|---|---|
| Reported amount | 178,500 therms | 4,185,000 kWh | 10 lb of R-410A recharged (service record) M-53 |
| Converted | 17,850 mmBtu | 4,185 MWh | 4.536 kg M-54 |
| Factor | 53.06 kg CO2 per mmBtu (EPA Hub 2025) | 911.321 lb CO2 per MWh (eGRID2023, subregion RFCW) | GWP 1,924 (IPCC AR5 via EPA Hub 2025) M-55 |
| CO2 | 947,121 kg | 4,185 x 911.321 x 0.45359237 = 1,729,946 kg | not split M-56 |
| CH4 and N2O | 17.85 kg and 1.785 kg | 134.78 kg and 18.98 kg | not split M-57 |
| CO2e | 948,094 kg | 1,738,750 kg | 8,727 kg M-58 |
| Grade, label | A, Reported | A, Reported | A, Reported M-59 |
Electricity factors are published in pounds per MWh, which is why the pound-to-kilogram conversion appears.
4. Repairs, corrections and gaps in the company data
- Malformed rows (rule D-P1). Six template rows carry one extra empty field in the fleet area. The repair drops that empty field and nothing else, flags the row, and refuses to run on any other mismatch. Three Ridgeline fleet values (4,200, 2,800 and 1,200 gallons) land in the gasoline column as a result and are labelled Inferred (D-P1-R). M-60, M-61, M-62, M-63
- Known data-entry errors (rule R10). Where evidence shows a submitted value is wrong, the corrected figure is used and the as-submitted figure is shown beside it, with the evidence. A reconciliation test anyone can re-run counts as evidence. The case applied is NovaCare's Marietta and Augusta clinics, where gas therms look like electricity kWh. The corrected treatment is Inferred and pending NovaCare's confirmation. M-64, M-65, M-66
- Rows with no data reported have no value entered (they add nothing to the totals), are labelled "Not reported" and flagged as a known understatement. M-67
- Questionnaire figures used in place of template figures (decision D-G1-7). Where the base uses the questionnaire's figure for a site and stream, the template figure is kept beside it as an alternative and the line says so. M-68
- Questionnaire totals versus template totals (rule R2). Where the two differ by more than 5 percent for a company and stream, the questionnaire figure appears beside the base as an alternative. M-69
5. How the base, low and high cases are built
Each site gets a disposition from reconciling the roster against the company's own records (rule R1). The disposition decides which cases a line enters. M-70
| Disposition | Meaning | Base | Low | High |
|---|---|---|---|---|
| consistent, consistent-partial-year, ambiguous, untestable | The company's records support it, or are silent | in | in | in M-71 |
| hold-unverified | Absent from the company's own complete site list; no evidence either way | in, flagged | out | in M-72 |
| hold-contradicted | The company's own records say it was not operating in 2025 | out | out | in M-73 |
"Untestable" means the company's records give no site list, so the roster cannot be tested at all. It is not the same as "fine". Wherever the high case is shown it carries the label that it excludes the sites missing from the roster, because it reads like an upper bound but cannot include sites with no data. Where a company's records give no site list, the rule (D-R1-A1) is that its figure is a single base number with the label "roster unverified at site level" beside it. Five companies are in that position: Clearview, Meridian, Pinnacle and Sterling (100 percent of base emissions on rows the company's records cannot test) and Keystone (95.6 percent). Dashboard v14 and the evidence view show the label beside those figures (finding F-110, closed when Daniel re-signed dashboard v14). M-74, M-75, M-76, M-77, M-78, M-79, M-80
6. Sensitivities (the pre-registered alternatives)
Each is a separate set of lines, tagged, pointing at the base lines it replaces. None enters a base total.
| Tag | What it tests |
|---|---|
| boundary_swing | Each site within 25 km of an eGRID boundary (and the sites in multi-subregion areas) re-run with the nearest different subregion's factor M-81 |
| clearview_115 | Clearview's fleet fuel scaled from the 85 vehicles it reported to the 115 on its board deck. This is a scaling, never data. The 18 drilling rigs are not estimated M-82, M-83 |
| r10_as_submitted | NovaCare's two corrected rows as submitted M-84 |
| r9_full_charge | Refrigerant at full charge rather than leakage M-85 |
| fleet_ch4_n2o_bound | Fleet methane and nitrous oxide, using EPA's stationary per-gallon factors as a stand-in. A bound on what is left out, not a swing M-86 |
| r2_alternative | The questionnaire figure where it differs from the ledger by more than 5 percent M-87 |
An AR6 refrigerant sensitivity is pre-registered for any company whose refrigerant leakage exceeds 5 percent of its base Scope 1 and 2 (decision D-P3-R). It has not been triggered for any company in this build. M-88, M-89
7. Data-quality grades and flagged rows (rule R13)
- Row grade. Utility bill or fuel receipt is A. Accounting export is B. Landlord data and estimates are C. On the five fleet-log rows, fleet fuel is A and electricity and gas are C. Refrigerant leakage from a service record is A; estimated from EPA default rates it is C. M-90, M-91, M-92
- Company and site grade. By share of base-case full-year Scope 1 and 2 emissions: A if more than 80 percent comes from bills and fuel records; otherwise B if more than 80 percent comes from those plus accounting exports; otherwise C. A share of exactly 20 percent from proxies counts as C. Grade C is flagged in every downstream view. M-93, M-94, M-95
- Flagged rows. A line is flagged and listed with its reasons if any of these hold: not reported; corrected under R10; template row repaired; partial year; Grade C; Inferred; roster disposition hold-unverified or ambiguous; outside the plausibility band; or a company-level fill. M-96
- Plausibility test (rule R5). For every company with a fleet, gallons per vehicle and fuel type are compared with the federal average for that vehicle class. A value outside half to double the average is flagged. This catches wrong units, digits or vehicle counts, not differences in fleet efficiency. It flags only; a correction needs evidence under R10. M-97, M-98
8. Intensity, rankings and financial exposure
- Intensity is full-year base emissions divided by revenue (tonnes CO2e per USD million) and by headcount, shown with the low-to-high range. Revenue, EBITDA and headcount come from the Northwood data spine, as of 31 January 2026 (rule R4). The alternative denominator is FY2025 revenue from the Q4 board deck where available. Patriot's headcount is its 185 internal employees; its roughly 3,200 temporary workers are reported separately. M-99, M-100, M-101, M-102
- Rankings use bands. Adjacent companies whose totals differ by less than 15 percent share a band, because the gaps are inside the uncertainty. Companies in the same band are shown as a range of ranks (for example 5 to 7), not as exact positions. M-103, M-104
- Carbon-price exposure (rule R14). Full-year base Scope 1 and 2 emissions times three scenario prices, in dollars and as a share of EBITDA. Scope 1 and Scope 2 are shown separately. The prices are scenarios, not forecasts: M-105
- low: zero (there is no US federal carbon price); M-107
- mid: the latest California-Quebec auction settlement price before the report date; M-106
- high: the volume-weighted EU allowance auction price for the latest full calendar year, converted at the average euro-dollar rate for that year. No company has sites in the EU, so this is how a European or multinational buyer would see the emissions, not a cost any company pays. Every use carries that note. M-108, M-109, M-110
- Severity bands use Northwood's own risk framework: under 5 percent of EBITDA is Low, 5 to 15 Medium, 15 to 30 High, 30 or more Critical. M-111
- Value at exit, as a what-if. The annual cost, if it reached EBITDA at exit, times Northwood's own exit multiple from the data spine. This is a scenario, not a valuation. M-112
- Energy cost sensitivity. A 20 percent energy price shock on each company's reported energy spend, as a share of EBITDA, for the companies that reported spend. It is not extrapolated to those that did not. M-113, M-114
9. Scope 3 flags
Scope 3 is not calculated. Each company has a ranked list of the Scope 3 categories most likely to matter to it, with a materiality level, the basis (described by the company, or inferred from its business model), the evidence line in the data room, and what data would improve the flag. These are flags for planning data collection. They are not estimates. M-115
10. Inputs written by hand, and what a new year means
Three input files were built by reading the data room and citing the line for each value:
pipeline/energy-spend-v2.csv: annual energy spend by company, quoted from the questionnaires. M-116pipeline/fy2025-alternatives-v1.csv: FY2025 revenue and EBITDA alternatives from the board decks. M-117pipeline/scope3-flags-v3.csv: the Scope 3 flags. M-118
For a new year, these must be redone by hand, and the build scripts name them in the code. The same applies to the roster reconciliation (re-run it against the new roster and review each disposition change), the data corrections, the refrigerant evidence, the questionnaire totals and fills, and the data-gap list. The re-run note lists the order. M-119
11. Known limits and unverified items
- Understatements, all disclosed: the 27 sites missing from the roster; rows not reported (no value entered); fleet methane and nitrous oxide; managed buildings (excluded by design). M-120
- Roster label (F-110): dashboard v14 carries the mandatory roster label for the five companies; the earlier dashboard v13 did not. The ledger and the evidence view carry it too. M-121
- Ledger lineage gaps (finding F-124). 44 of the 529 ledger lines have no factor ID. 31 of them are refrigerant and R-22 memo lines, which read refrigerant-evidence-v3.csv but cite "Refrigerant evidence v2"; the rest are lines of the questionnaire-total alternative. Each carries its factor value, so the figures reproduce; what is missing is the link from the line to a row of the factor file. The fix goes into the next ledger version. M-151, M-152, M-153, M-154
- Refrigerant from default leak rates is a screening estimate. The rates are the upper end of the IPCC ranges that EPA recommends for screening, and the guidance is dated December 2023. M-122, M-123
- The methane GWP of 28 is used for all methane. The GHG Protocol list gives 30 for fossil methane under the same assessment. The effect is small. M-124
- The GWP basis is AR5, not AR6, to match the EPA Hub. The GHG Protocol recommends AR6. M-125
- Some Stage 0 data-room claims were not located again when re-checked (Harvest Ridge's $34 million and $94.7 million supplier spend, Clearview's lost RFPs, Patriot's gallons per vehicle). None feeds a figure in this inventory. M-126
- The Northwood data spine and the questionnaires are management-reported. They were tested against each other and against the companies' own records where those exist, but not audited. M-127
- Company-level questions remain open with the companies: the site lists for the "untestable" companies, Clearview's fleet count, and NovaCare's corrected rows.
- The browser check of the dashboard is done by a person, so a re-run needs one. M-128
- eGRID boundary files: the download date and web address were not recorded when they were fetched. M-129
12. How to trace any figure
Every number the dashboard shows has a figure ID. The calculation ledger lists every line behind it with the exact row of the carbon template, the exact factor row, the rule and decision IDs, the grade and the label. A check confirms that every displayed figure resolves to ledger lines and equals their sum (no orphan numbers), and that every pre-registered alternative appears beside its base figure. An independent recomputation, written separately from the calculator, rebuilds every line from the raw template and agrees to within 0.001 kg CO2e per facility. The build record (methodology/build-record-live.md) holds the run log, the decisions and their reasons, and the findings and how each was settled. M-130, M-131, M-132, M-133
Each statement in this note that rests on a fact, number or rule carries a tag such as M-07. The claims register (traceability-register-v5.csv) gives the evidence for each tag (a decision, a settings value, a ledger line, a file or a journal entry), and the check check-traceability-v5.py re-tests every row against the record.
13. The rules at a glance
| Rule | Plain statement |
|---|---|
| R1 | The roster is not the boundary. Every site gets a disposition. Where evidence points both ways, show a base figure with a low-to-high range M-134 |
| R2 | Source order when files conflict; show both where they differ by more than 5 percent M-135 |
| R3 | Fund roll-up uses ownership months; company pages use the full year M-136 |
| R4 | Denominators come from the spine, as of 31 January 2026; the alternative is shown M-137 |
| R5 | Flag fleet fuel outside half to double the federal class average; Clearview scaled to 115 vehicles as a sensitivity M-138 |
| R6 | Flag client data older than 90 days at the report date; reference data is not flagged M-139 |
| R7 | Confirmed, Reported, Inferred; totals take the weakest label M-140 |
| R8 | Ridgeline's own footprint only (six offices on the roster, a seventh is a roster gap); managed buildings separate M-141 |
| R9 | Refrigerant leakage in the base, full charge as an upper bound, R-22 on its own line M-142 |
| R10 | Use the corrected-by-evidence figure with the as-submitted figure beside it M-143 |
| R11 | Sites missing from the roster are excluded and disclosed, never estimated M-144 |
| R12 | Operational control; 100 percent of each company for the months owned M-145 |
| R13 | Grade each row, site and company by data source; flag Grade C M-146 |
| R14 | Carbon-price exposure at three scenario prices, as a share of EBITDA, with the EU-benchmark note M-147 |