Week 2 of 6, Climate Work Residency
Data audit
Is Northwood's climate data good enough to disclose to LPs (the fund's investors) and to survive a buyer's diligence, and what must be fixed first?
Highlights
Climate data review deck
Ranks four issues by value at stake and shows every company's readiness score on one page.
Open (opens in a new tab)Climate data quality memo
The full argument, with each finding labelled Confirmed, Reported or Inferred and an owner for each fix.
Open (opens in a new tab)Brief check: actions taken
Walks the five brief items against what was done, so a reader can see nothing was skipped.
Open (opens in a new tab)The case
Northwood's climate protocol exists on paper but has not run for 2026. I audited the data of all ten companies, not just the priority four, and scored each for readiness.
For LP disclosure, FY 2025 is achievable once known errors are fixed. For buyer diligence, no company passes today: all ten score 26 to 38 out of 100. The ranking is by value at stake. Meridian's exit data, Harvest Ridge's resubmission, Sterling's site-level data for defense bids and the LP climate section come first.
All deliverables
Each opens in a new tab, in the Kith Climate style it was delivered in.
- Climate data review deck (12 pages) (opens in a new tab)The presentation to Sarah Mitchell, Northwood's COO (fictional): headline, value-ranked issues, readiness scores, decisions needed and how it was built.
- Climate data quality memo (opens in a new tab)The written audit of all ten companies, with the value ranking and the readiness method.
- Brief check: actions taken (opens in a new tab)A page tracing each item of the Week 2 brief to what was done.
Data and method
The method notes, data and re-run steps that let the client reproduce the work without us.
- Data quality memo (Markdown) (opens in a new tab)Markdown fileThe memo source text, for reuse and diffing.
- Data request: Clearview (opens in a new tab)Markdown fileThe baseline data Kith needs from this company, with tiers set by what blocks the Week 3 baseline.
- Data request: Harvest Ridge (opens in a new tab)Markdown fileThe baseline data Kith needs from this company, with tiers set by what blocks the Week 3 baseline.
- Data request: Keystone (opens in a new tab)Markdown fileThe baseline data Kith needs from this company, with tiers set by what blocks the Week 3 baseline.
- Data request: Meridian (opens in a new tab)Markdown fileThe baseline data Kith needs from this company, with tiers set by what blocks the Week 3 baseline.
- Data request: Novacare (opens in a new tab)Markdown fileThe baseline data Kith needs from this company, with tiers set by what blocks the Week 3 baseline.
Show all 14
- Data request: Patriot (opens in a new tab)Markdown fileThe baseline data Kith needs from this company, with tiers set by what blocks the Week 3 baseline.
- Data request: Pinnacle (opens in a new tab)Markdown fileThe baseline data Kith needs from this company, with tiers set by what blocks the Week 3 baseline.
- Data request: Ridgeline (opens in a new tab)Markdown fileThe baseline data Kith needs from this company, with tiers set by what blocks the Week 3 baseline.
- Data request: Sterling (opens in a new tab)Markdown fileThe baseline data Kith needs from this company, with tiers set by what blocks the Week 3 baseline.
- Data request: Trident (opens in a new tab)Markdown fileThe baseline data Kith needs from this company, with tiers set by what blocks the Week 3 baseline.
- Fund-level data request (opens in a new tab)Markdown fileThe items only Northwood can answer, plus the plan for issuing the company requests.
- Presentation script (opens in a new tab)Markdown fileThe talking notes for walking Sarah through the deck.
- Cover email to Sarah (opens in a new tab)Markdown fileThe email that accompanied the memo and deck.
How it was built
- The work ran through scripts, not manual re-entry: raw files, then checks, then the memo figures. The scripts use only the Python standard library, so Northwood can re-run them.
- A verification script re-runs the pipeline and tests the memo's figures against fresh outputs. At the last recorded run 133 of 136 checks passed; the 3 fails were deck citations to an older memo version, declared as known.
- All ten questionnaires were extracted twice and compared: 151 shared figures, 0 mismatches.
- Read-only review agents challenged the draft memo. Later, opening five files nobody had read changed the exit date and settled a name ambiguity.
- Limits are stated. I set the readiness weights and materiality thresholds after the first findings were known, so they are held fixed until responses arrive.
My part
- I ranked the issues by value at stake, not by data urgency, which set what Northwood fixes first.
- I set the readiness weights and materiality thresholds, and held them fixed until the companies' responses arrive.
The AI's part
- Claude Code wrote the pipeline scripts, the logs of findings and assumptions, and drafted the memo, deck and eleven data requests.
- Review agents (assurance, client-voice, deck-quality, process) read the drafts and raised the fixes I decided on.
Across the engagement
From Week 1. The proposal priced four priority companies and promised a data audit. The audit covered all ten, with the priority four marked.
Into Week 3. The Week 3 inventory ran on the same ten companies and 138 facility rows. Where data was late it used the stated provisional rule from this week.