|Climate data review
Sarah Mitchell · COOHeadline01 / 12
Climate data review | Week 2 of 6 | September 25, 2026

Your protocol exists on paper. It hasn't run for 2026.

One fix reaches Partially ready (6) No single fix (4) Rung length · readiness score

All ten sit in the same band. Point at a rung for that company's score and exposure; the Northwood marker opens the portfolio totals.

For LP disclosure, FY 2025 is achievable once known errors are fixed — the underlying energy data exists for all ten portcos, and management estimates are acceptable under the protocol. For buyer diligence, no portco would pass today: all ten score in the lowest readiness band, 26 to 38 out of 100. Meridian is tied with Sterling for closest, has the most traceable data in the portfolio, and is the most urgent — it enters confirmatory diligence in Q4 2026 on a live $363.9M to $400.9M process (as at 31 January 2026). The risk there is diligence friction, not a quantified discount.

10
of 10 portcos, not diligence-ready
26–38
readiness score range, out of 100
138
facility rows checked
2
of 10 mention climate in board materials
Audience
Sarah Mitchell, COO, Northwood Capital
Author
Kith Consulting
Date
September 25, 2026
Coverage
All 10 active portcos, extended from 4 at no added cost
Edition
Portfolio edition: Kith Climate cohort case, fictional client
Dates
Timetable as planned in September 2026; figures as at the data room’s own dates, shown in brackets
02Act now

Ranked by value at stake, not by data urgency.

Four issues carry high value, a clock within about a year, and a fix within reach. Point at a step for its owner, clock and effort.

03LP-readiness scores

All ten, same band, different distance to move.

Every portco is Not diligence-ready (0–39). Six are one named fix from Partially ready (40+); four are not. Click a bar for the financial exposure the data supports.

One fix reaches Partially readyNo single fix identifiedDashed line · 40, the Partially ready threshold
04How the score works

Eight weighted dimensions. Three readiness bands.

Each portco is scored 0–4 on eight dimensions, then combined into a composite out of 100. This is the methodology behind every number in this deck — reusable next quarter without Kith.

0–39

Not diligence-ready

Foundational gaps; the baseline cannot yet be relied on in diligence. All ten portcos sit here today.

40–74

Partially ready

A usable baseline exists, with specific, namable gaps.

75–100

Diligence-ready

Supports LP disclosure or buyer diligence with minor polish.

The ceiling today is about 63: dimensions 3–6 carry 40% of the weight, and no portco scores above 1 on any of them. Sterling and Meridian sit 2.5 points from Partially ready on unrounded composites — treat the band as a close call and the dimension profile as the more reliable read. The weights were fixed on September 22, 2026, after the first automated checks had run, so with the data in view. They are held unchanged until request responses arrive, so they can't be tuned to results.

05Findings by portfolio company

The four highest-exposure companies, in their own data (six more summarised below).

Ordered by financial exposure at stake. Each finding carries its confidence: Confirmed, Reported or Inferred.

The remaining six portcos, summarized

06Items for Northwood to confirm

Three items can follow after Monday.

The memo lists eight items for Northwood to confirm. The three that unlock the Week 3 baseline are on the decision page (09). These three change grades or later-week work, not Week 3 figures; the rest are in the memo.

07Data requests and timeline

Three tiers, set by what blocks Week 3, not by importance.

Eleven requests go out through Northwood — one per portco, plus the fund-level ask — the moment this memo is reviewed and issued, planned for Monday, September 28. The 79-ask register sits behind every tile.

Tier 1 · blocks the Week 3 baseline · due Monday, October 5

Business days from Monday, September 28. Point at a tier to see its test.

Harvest Ridge

Its corrected submission needs plant-level re-entry from a finance team in the middle of integration.

Patriot

The Evansville and Rockford figures depend on landlords replying — outside management's control. Its template also lists 12 sites that aren't in its organisation chart, so the branch list is a Tier 1 ask.

Ridgeline

Three Tier 1 asks, including an office (Las Vegas) that isn't in the template at all.

If responses are lateThe Week 3 baseline still covers all ten portcos on a stated provisional rule: where the template is known wrong and the questionnaire gives a figure, that figure is used and flagged (NovaCare's two migrated clinics, Patriot's two zeroed branches, Meridian's four zero-gas sites). Where two submissions conflict and neither is known right, a range is shown, as for Harvest Ridge. Everything else uses the submitted template, marked provisional — with a methodology note stating what was assumed and what would close the gap.
08How this was built

Scripted end to end, and Northwood can run it again.

Figures move from raw files to the memo through scripts, not manual re-entry. The scripts use only the Python standard library, and the whole pipeline is checked before issue. Point at a stage to see its output.

91 / 0

verify-pipeline-v1.py, at memo v4: every script re-run, memo figures checked against fresh outputs, 91 checks passed, 0 failed, before issue. The later v2 gate’s last recorded run (27 September, memo v5) passed 133 of 136; its 3 fails were this deck’s citations to an older memo.

151 / 0

Second extraction of all 10 questionnaires: 151 shared figures compared, 0 mismatches.

10

Steps to re-run next quarter (Appendix B), from new submissions and a config-file year change through to the same gate.

Std lib

Every script runs on the Python standard library only — no dependency Northwood's own team would need to install.

Leave-behindEmission factors with source and retrieval date sit in _reference/emission-factors/; the method is set out in _reference/audit-methodology-v2.md. Judgment inputs live in data files, not code, so a change doesn't require touching a script. The one manual step left is questionnaire transcription, checked by the second extraction above; moving the playbook to a Northwood-owned location is part of the Week 6 handover.
09What Northwood needs to decide by Monday, September 28

Three answers unlock the Week 3 baseline.

Appendix · How this was built (process appendix) · D- and A- ids are entries in the disposition log (every finding) and the assumption log (every judgment call)
10Eight stages, four memo versions to the cohort review

Each review made the memo answer your question more directly.

Step height is the stage; bars are the memo's length at each version. Two independent reviews (shaded steps) drove most of the change. Click a step for what happened and the file that shows it. Sourced from Kith's working files and archived drafts, not from the memo. Memo v5 (27 September) and v6 (1 October) followed a closeout check of the source files; this deck was brought in line with v6 for the portfolio.

11Instructions, agents and files

One author, read-only challengers, and every decision logged.

Each lane is a participant; each column is a stage. Solid lines hand work forward; dashed lines are challenges sent back. Point at a marker to see what that participant did at that stage.

Instructions read Authoring pass (Claude Code) Scripts and logs Read-only review agents Kith lead's decisions Still to run
12How the deliverables link

What you receive, what you keep, and what it feeds next.

Evidence on the left, later weeks on the right. The leave-behind column is what lets Northwood re-run this without Kith. Point at any item to trace what it is built from and what it feeds.