Slide 1 notes. Sarah, this is the close-out of Week 6. The brief was to turn one engagement into something Northwood keeps: a disclosure system, not a one-off report. All four gates are signed, including a final gate after every open item was closed or handed to you. The final freeze hashes 85 files, including 2 of the 10 board pages; the other eight come from the same builder and passed checks B01 to B09. Nothing has gone to any LP, lawyer or portfolio company; the sends are your decisions and they are on the last content slide. The link on this slide opens a portal holding every final v2 deliverable.
The answer first
You now hold three things that work together
An LP climate report
Anonymised, in TCFD order, with four framework indices. Now with Scope 3 and financed emissions. It has had 22 checks run: 18 pass, 4 not applicable, none fail, and is final.
Your own answer desk
A dashboard, plus a data pack your own Claude answers from, citing the data row behind every answer.
A system that repeats
A reference library, an engine and nine re-run steps that rebuild every output from next year’s data.
21,778 t
Scope 1 and 2, calendar 2025, ownership-period basis
12,494 t
Financed emissions following PCAF; Scope 3 estimated at about 289,000 t (grade C)
$12.6M
Physical-risk expected loss, gross: Fund I $1.9M, Fund II $10.7M
10 of 10
Companies with no board oversight, climate role or target
Week 6 readout · Portfolio edition · 2
Slide 2 notes. Three things, and they are designed to be used together. The report is what LPs will see. The dashboard and data pack are how you answer their follow-up questions without calling us. The library and re-run steps are how next year’s report gets made. The headline numbers: 21,778 tonnes of Scope 1 and 2, counting the two part-year companies only for the months Northwood owned them. Financed emissions, following the PCAF equity method with stated proxies, are 12,494 tonnes, and the spend-based Scope 3 estimate is about 289,434 tonnes, graded C and not to be added to Scope 1 and 2. Physical climate risk is about $12.6M of expected loss over the holding periods, gross of insurance: $1.9M for Fund I and $10.7M for Fund II. The finding that matters most for exit readiness is the last one: no company has board oversight of climate, a climate role or a target. The report says so plainly, because a buyer will ask. Each of these ties to value: a recurring cost a buyer sees at exit, multiplied by the exit multiple, is value a buyer may take off the price, and the board pages show this for each company.
Slide 3 notes. This is the way into everything. Each tile opens the portal at that deliverable. The pages show the final v2 files (the final freeze hashes 85 files, including 2 of the 10 board pages; the other eight come from the same builder and passed checks B01 to B09), and each page names its master file in the engagement folder, so what you read here is what sits on file. The order runs from what you will use most, the LP report and your dashboard, to what you will use least often, the library and the process log. The machine-readable files, such as the evidence map and the query-pack data, stay in the engagement folder rather than the portal. Most portal pages are internal and name companies; only the LP report and its indices are written for LPs.
How they work together
One data table feeds every output
Inputs
Data room and frozen Weeks 3 to 5
Intake
124 rows, each sourced and graded
D1 + engine
Frameworks mapped, rules applied
22 checks
Run on every draft, logged
D3 LP report, for LPs
D4 Dashboard, for you
D5 Board pages, for boards
D2 closes the loop
The register shows each gap for each company; the 16 data requests turn the gaps into questions. Answers become next year’s inputs.
No figure is typed
Every number in the report, dashboard and board pages resolves from the same data table, so the three can never disagree.
D6 records the run
The process log keeps what worked and what it took, so the next cycle starts from evidence.
Week 6 readout · Portfolio edition · 4
Slide 4 notes. This is how the pieces connect. Everything starts from the data room and the outputs we froze in Weeks 3 to 5: emissions from Week 3, EBITDA effects from Week 4, physical risk from Week 5. Week 6 did not recalculate those; it read them. The intake step turns them into one data table of 124 rows, each with a source and an A, B or C quality grade. The engine then maps that table against the library’s framework files and applicability rules, and the 22 checks run before anything is produced. The three outputs come from the same table, so the LP report, your dashboard and the board pages cannot disagree with each other. The register and the data requests show you where the gaps are and who to ask; those answers feed next year’s cycle.
How the week ran
A blind scout first, then four signed gates
Step 1
Scoping
22 scoping decisions, the week scaffolded, a sandbox set up
Step 2
Scout run
A full library built with no client data: register, engine, 21 checks
Gate 1
G1 spec signed
Charter of 15 rules, 20 assumptions, ten owner rulings
Gate 2
G2 results signed
Intake, D1 to D5, two reviewers, all must-fix findings closed
Gate 3
G3 frozen
Process log, hand-off notes, frozen (60 files)
Gate 4
Final signed
Open items closed, Scope 3 and PCAF built, all outputs reissued as v2
67
decisions, recorded in the owner’s words as they were made
4 gates
each signed by the engagement owner and frozen
Weeks 2 to 5
read from frozen outputs, not rebuilt
Week 6 readout · Portfolio edition · 5
Slide 5 notes. The week ran in two halves. First, a scout: a separate sandbox that built a complete disclosure library for a generic private-equity client, with no Northwood data in it. Second, the gated build: the spec gate, the results gate with two independent reviewers, and the hand-off gate. A fourth, final gate followed once every open item was closed or handed to you: the per-fund split, PCAF financed emissions and the Scope 3 estimate were built, and every output was reissued as version 2. All 67 decisions are in the decision log in the engagement owner's words. Emissions, EBITDA effects and physical risk were read from the frozen Week 3 to 5 outputs, so nothing you saw in earlier weeks was recalculated.
Why this strategy
The library had to stand on its own
Build the library blind, then run Northwood through it under gates.
Hybrid option C, chosen by the engagement owner.
It proves the asset is reusable
A library built with Northwood’s data in view tends to fit only Northwood. The scout saw none, so what it produced works for the next client too.
It puts review effort where judgment sits
The gated build reviewed only what differed from the scout, the lesson from Week 4. Reviewers spent their time on Northwood’s facts.
It tests every assumption against real files
The scout made 23 guesses about client evidence. Real data confirmed some and broke others, and each break became a rule or a check.
Week 6 readout · Portfolio edition · 6
Slide 6 notes. Why a scout? The brief’s point was that the library is the product and the report is a by-product. A library written while looking at Northwood’s files would quietly bake in Northwood’s quirks. So the scout built it for a generic US mid-market manager with no client data at all. Then the gated build applied it to Northwood, and the reviewers looked only at what changed from the scout, which is where the judgment is. The third benefit is that the scout wrote down 23 guesses about what client evidence would look like. Real files broke several: no company has a target or a governance body, sector labels alone identify companies in this portfolio, and the data room names no EU investor. Each break became a rule, a check or a recorded decision.
Key decisions
The calls that shaped what you received
Question
What we chose
Why it matters to Northwood
Who sees names
Anonymised LP report; named dashboard for you
Your protocol bars company names in LP disclosures
Descriptors
Groups of at least three: Fund I or Fund II company
Four sectors hold one company each, so a sector label would identify it
Coverage
All ten companies through one pipeline
Data existed for all ten; one pipeline made the extra six cheap
Frameworks
Global library, each Tier 1 framework labelled as asked today or library coverage
No EU or UK LP is named today; you are ready if one asks
Your tools
Static dashboard plus query pack, Option B
Works with no Kith access and no special software
Governance
Names you as responsible executive; formal confirmation is your action
LPs and buyers ask who owns climate risk
Close-out
Fund split, PCAF and a Scope 3 estimate built; pricing option A
LPs commit by fund; financed emissions and Scope 3 are what LPs ask next
Week 6 readout · Portfolio edition · 7
Slide 7 notes. These are the decisions you will feel when you use the deliverables. Each has a D-6 number in the decision log with the alternative that was rejected. Two need your attention. First, the governance section names you as the executive responsible for climate disclosure. That was the engagement owner’s decision, not a Northwood one, and the report says your formal confirmation is on Northwood’s list of actions. Second, anonymity: because four of the sectors in this portfolio hold only one company each, describing a company as, say, an environmental services business would identify it to an LP. So the report uses only two fund labels, a Fund I company and a Fund II company, each covering at least three companies. At close-out the fund split, PCAF financed emissions and a spend-based Scope 3 estimate were built, and on pricing, option A, an annual refresh retainer, is recommended; Northwood has not accepted it and no fee is set.
Quality assurance
How you know the numbers are right
22
quality checks on the LP report, each with a test that proves it can fail
All
final output files rebuilt byte for byte from the pinned inputs alone
40 of 41
reviewer findings fixed over two gates, all 12 must-fix among them; one carried
2
cold tests of the query pack in separate Claude sessions, as you would use it
19
framework checks against official text; three scout errors corrected
30
questionnaire answers, each quote checked word for word by script
Week 6 readout · Portfolio edition · 8
Slide 8 notes. If an LP asks how you know the report is right, this is the answer. Twenty-two checks run on every version of the report, and each one has a test that deliberately breaks it, so we know the check is not passing by accident. A clean re-run copies only the pinned inputs into an empty folder and rebuilds everything: every final file came out identical. At two gates, independent reviewers read the work, one checking that every figure traces to a source and one reading it as you would. They raised 41 findings, 12 of them must-fix, and all 12 were fixed; at the final gate they recomputed the Scope 3, PCAF and fund-split figures by hand. The one finding carried is that the pricing reflection names no fee, which is a commercial choice rather than an error. Two separate Claude sessions, given only the query pack, tried to answer questions the way you would. They found nine gaps and one real inconsistency in how the fund’s data grades were weighted. Framework facts were checked against official sources, with the date of each check recorded. The report itself is not externally assured, and it says so.
What real data exposed
Each break became a permanent check
What broke on Northwood’s data
What the system now does
The engine added the fund total to the company rows: Scope 1 would have read 23,899 t, not 11,599 t
Check Q22: a headline equals the fund roll-up or the company sum, never both
The LP draft printed internal company codes that the scout’s name check missed
A tighter name-leak scan; the builder refuses to copy any file with a leak
Sector labels singled out four companies
Fund labels, minimum group of three, tested by check Q21
The scout assumed EU investors; the data room names none
EU frameworks labelled library coverage until an LP asks (DR6-11)
Week 4’s “carbon” column was mostly a customer-data line, not a carbon cost
Board pages leave it out; one example is recomputed by hand before any frozen column is quoted
Synthetic data passed every check. Real data exposed two defects as soon as the report was drafted. That is why the gated build ran on Northwood’s own files before anything was signed.
Week 6 readout · Portfolio edition · 9
Slide 9 notes. We want to be open about what went wrong, because it is the reason you can trust what came out. The scout’s code had only ever run on synthetic data, and it passed. On Northwood’s data, two defects appeared as soon as the report was drafted. The engine added the fund roll-up row to the company rows, which would have nearly doubled Scope 1. And the report printed internal company codes, which an LP could have used to identify a company. Both were fixed, and both now have a permanent check with a test that proves the check fails when it should. The anonymity rule and the framework labels came from the same kind of collision with real facts. The Week 4 carbon column was caught by recomputing one company by hand: an 18 per cent of EBITDA carbon cost on about 5,100 tonnes did not make sense, and it turned out to be mostly the customer carbon-data line.
Brief against delivery
The brief asked for five things; each is delivered
The brief asked for
What was delivered
Where
Process reference
Engagement playbook: eight stages, client inputs, automation and human split, QA protocol, version rules
D1
Standard references: TCFD primary, CSRD stretch
IFRS S2 in TCFD structure as primary; ESRS E1 (CSRD) mapped requirement by requirement; plus SFDR PAI, ILPA EDCI, six Tier 2 and one Tier 3. Each file dated, with its verification status and a re-check date
D1
Methodology references
Materiality scoring, the A/B/C data-quality rubric, narrative voice, and an anonymisation convention
D1
The report
TCFD-structured LP report from the frozen Weeks 3 to 5 data; every figure drawn from the data table; 22 checks run: 18 pass, 4 not applicable, none fail
D3
Process log
Process log, a Week 1 promise-by-promise reconciliation, and a pricing reflection
D6
Also built
D2 register and data requests, D4 dashboard and query pack, D5 ten board pages
Week 6 readout · Portfolio edition · 10
Slide 10 notes. This is the line-by-line answer to the Week 6 brief. It set five engagement items: a process reference, standards references with TCFD as the primary and CSRD as the stretch, methodology references, the LP report, and a process log. Each is delivered, and the right-hand column tells you where to find it in the portal. On standards, we went past the brief: IFRS S2 carries the TCFD structure, since the ISSB took over TCFD’s monitoring, and ESRS E1 is mapped to the 2026 simplified standard. Every framework file records when and against which official source it was checked, and when it must be checked again. Beyond the brief we built three more deliverables, because your own protocol and your LPs need them: the register that maps one disclosure to many frameworks, your dashboard, and the board pages.
Slide 11 notes. The brief gave an eleven-item QA checklist and asked that it run on every draft, not from memory. Q01 to Q11 are those eleven items, one for one, and they run as a script every time the report is built, with every run logged against the report’s fingerprint. Four come back not applicable this year (Q04, Q08, Q10 and Q17), honestly rather than as a silent pass: there is no prior report to reconcile to (Q04, Q10), no company has a target (Q08), and the case is a client case, not synthetic data (Q17). Q09, the governance check, passes because the report names you as the responsible executive; if you do not confirm that, the check will need a different answer. The eleven on the right were added because Northwood’s own protocol needs them, such as the verbatim qualification sentence and no company names, or because real data showed they were needed, such as Q21 and Q22.
The brief’s reporting process
The automation carries the structure; people carry the judgment
1 Intake
2 Materiality
3 References
4 Data
5 Interviews
6 Draft
7 QA
8 Sign-off
The automation carried
Framework mapping: 32 concepts to four Tier 1 frameworks
Drafting from data-table figures, so no figure is typed
Interview prep packs written from the gaps
The record: decisions, content indices, versions
All 22 checks on every run, logged
People carry
Scope cuts: 67 owner decisions, each recorded
Conversations with company CFOs and operating leads
Interpretation: why the absences are the finding
Sign-off: four gates signed by the engagement owner
Release: your confirmation before anything goes out
Week 6 readout · Portfolio edition · 12
Slide 12 notes. The brief described a seven-stage reporting process and split the work between what the automation carries and what people carry. The playbook keeps all seven stages and adds materiality as a stage of its own, because your Week 4 materiality work already exists and should be converted, not redone. On the automation side, everything the brief listed is in place: framework mapping, drafting from data, interview prep packs, documentation and QA on every run. On the people side, nothing was automated away. Scope and materiality cuts were the engagement owner’s decisions, recorded in his own words. The conversations with your company CFOs and operations leads are still to come: they are the 16 data requests, which have not been sent. And release is yours.
Where we chose a different route
Every departure is stated, with its reason
The brief or the Week 1 proposal
What we did
Why
CSRD as the stretch framework
Mapped in full; not claimed in the report
No EU LP is named, and CSRD scope for a PE group is a legal question now with counsel
The report drafted by week’s end
Final and passing all checks; not yet sent
Your governance confirmation and Investor Relations review come first
A summary per company
A board page per company instead
Your protocol keeps company data out of LP reports; its own per-company vehicle is the board review
Four priority companies
All ten, same depth
One pipeline made the other six far cheaper than separate builds
Hours per phase in the process log
Counts only; effort hours not measured
The record does not hold time spent; we would not invent hours
Scope 3 screen for two companies
A spend-based estimate for all ten, plus PCAF
Owner rulings D-6-61, D-6-65 to D-6-67; graded C and labelled an estimate
Week 6 readout · Portfolio edition · 13
Slide 13 notes. The engagement owner’s instruction was that departing from a Week 1 promise is fine as long as the reason is explained, and that applies to the brief too. So here is every departure. CSRD is mapped requirement by requirement in the library and the content index, but the report claims nothing about CSRD, because no EU investor is named and whether CSRD reaches a US private-equity structure is a legal question; six questions are drafted for counsel. The report is final, but nothing is sent until you confirm the governance arrangements and Investor Relations has reviewed it. Per-company LP summaries became board pages because your protocol keeps company data out of LP disclosures. All ten companies were covered rather than four. The process log reports counts rather than effort hours, because the record does not hold effort, and we will not make those numbers up. Scope 3 went further than promised: a spend-based estimate for all ten companies, with PCAF financed emissions.
The brief’s two questions
Our answers, on the evidence we have
What process would you stand behind?
The eight-stage playbook, run under signed gates. Every figure comes from a sourced, graded row. Twenty-two checks run on every draft. Independent reviewers read the results, and a clean re-run proves the whole set rebuilds.
How long would it take?
Not measured: effort hours and time with company contacts; the record holds counts, not time spent. Under the annual refresh retainer, the first real cycle will be timed, and that sets the number we quote.
Week 6 readout · Portfolio edition · 14
Slide 14 notes. The brief closed with two questions and said they were the week’s real outputs. The process we would stand behind is the one you have seen: the playbook, gates signed by a named owner, sourced and graded data, scripted checks with tests that prove they work, independent review, and a re-run that proves it can be repeated. On how long, we will only give what was measured, and effort was not. The time your company contacts will spend answering data requests also happens outside our record. Pricing: option A, an annual refresh retainer, is recommended; Northwood has not accepted it and no fee is set. Its first cycle will be timed. The playbook carries an hours table from the scout; it is an estimate and should not be quoted. The first real annual cycle, timed, will give a defensible number.
For you, Sarah
Which deliverable to reach for, and when
When you need to
Open
Keep in mind
Report to LPs
D3 LP report and its framework indices
After you confirm the governance arrangements; Investor Relations reviews first
Answer an LP or buyer question
D4 dashboard, or the query pack in your Claude
Every answer cites its row; names stay internal
Brief a company board
D5 that company’s board page
That company’s data only; the deal lead sets the meeting
Chase missing data
D2 data requests
16 requests, one pack per recipient; none sent yet
Check whether a rule applies
D2 applicability, D1 framework notes
“Unknown” means a fact is missing, never no; legal questions go to counsel
Update next year
Re-run note and the D1 library
An analyst runs nine steps; you review the result
One rule across all of them: fund totals count two part-year companies for the months owned, company rows are full-year, and Scope 3 overlaps Scope 1 and 2. Never add across them.
Week 6 readout · Portfolio edition · 15
Slide 15 notes. From here on, this is about using what you have. Think of it by the job in front of you. When you report to LPs, use the LP report and its indices, and only once you have confirmed the governance arrangements; your protocol has Investor Relations review it before it reaches an LP. When an LP or a buyer asks a follow-up, open the dashboard, or load the query pack into your own Claude; it answers only from the files and cites the row behind every answer, and it will not use company names in anything meant for LPs. For a company board, use that company’s page and nothing else. For gaps, the data requests say who to ask and why it matters. For applicability, an unknown means a fact is missing, not that the rule does not apply. And for next year, the re-run note. One rule to keep: the fund total and the company rows are on different bases, so never add them.
Your annual cycle
Next year runs on your protocol’s calendar
January
Send templates
Collection files go to the ten companies, with this year’s data requests folded in
Mid-March
Data back
Energy, fuel and climate questionnaires returned
Mid-April
Re-run
An analyst runs the nine steps; every check must print PASS
Your review
Decide
Headlines against last year, the release list, which requests to send
Q4
Board reviews
Each company’s board page at its annual climate review
Before each cycle
Re-check any framework past its re-check date. First due: California SB 253, on or after 11 November 2026 (as at the 8 November report date; not re-checked for this edition).
Week 6 readout · Portfolio edition · 16
Slide 16 notes. The system fits the calendar your own climate data protocol already sets. In January the collection templates go out, and this is the natural moment to fold in the data requests. Companies return their energy data and questionnaires by mid-March, and the portfolio footprint is due in mid-April; that is when an analyst runs the nine re-run steps and confirms every check prints PASS. Your review then takes a short sitting: the headline numbers against last year’s, the release list at the top of the report, and which data requests to send. In the fourth quarter, each company’s board page goes to its annual board climate review. One standing task: before each cycle, re-check any framework whose re-check date has passed. The first is California SB 253, whose proposed first reporting date is 10 November 2026, so it should be checked on or after 11 November (as at the 8 November report date; not re-checked for this edition).
The leave-behind
Each deliverable ships with all it needs to be redone
The output
Report, dashboard, board pages and register, all built from one data table
The method
Playbook, methodology notes, and a guide in each folder saying how every figure is made
The reference data
Framework files, applicability rules and evidence rules as dated, structured tables
The re-run note
Nine steps in order, what each makes, and what to check at the end
The test
Could Northwood update this next year with no Kith access? Yes, with one analyst or IT colleague.
The caveat, stated plainly
The nine steps are commands, not spreadsheets, so a technical colleague runs them. Your part is the review. The clean re-run check has already rebuilt every final file from the inputs alone.
Week 6 readout · Portfolio edition · 17
Slide 17 notes. Leave-behind independence was a standing rule for this engagement: every substantive deliverable ships with its output, the method behind it, the reference data in a reusable form, and a plain-language note on how to run it again. The test we hold ourselves to is whether Northwood could update the disclosure next year with no access to Kith. The answer is yes, with one condition we want to be straight about. The nine re-run steps are programs, so someone technical, an analyst or an IT colleague, has to run them. Your part is to review the result. The clean re-run check has already shown the steps work: given only the inputs, it rebuilt every final output file exactly. The re-run note and the hand-off notes in the portal are the two documents to give that colleague first.
Handed to you
What is now yours
Action
Owner
Formally confirm the governance arrangements: responsible executive, committee oversight, the VP of ESG seat (DR6-01)
You
Investor Relations review of the LP report, then sending
Jennifer Park, then you
Choose which of the 16 data requests to send, folded into the January collection
You
Put the six CSRD and California questions to counsel
You, with a lawyer
Agree the Keystone data-limit wording before its board sees it
Nadia Fernandez (Northwood deal lead for Keystone), with Keystone management
Re-check California SB 253 on or after 11 November 2026 (as at the 8 November report date; not re-checked for this edition)
Kith, under the retainer
Kith’s items are closed: report date, source wording, the fund split, PCAF, the Scope 3 estimate, and pricing (option A, an annual refresh retainer, is recommended; Northwood has not accepted it and no fee is set).
Week 6 readout · Portfolio edition · 18
Slide 18 notes. Everything Kith could close is closed: the report date is set, the source wording approved, the exposure is split by fund, financed emissions follow PCAF, and Scope 3 is estimated from spend. What remains needs people at Northwood. You confirm the governance arrangements; the report names you as responsible executive and says that formal confirmation is Northwood’s action. Then Investor Relations reviews it under your protocol, and you decide when it goes. You choose which data requests to send; the new one, supplier spend by category, would turn the Scope 3 estimate into category-level figures. The legal questions need a lawyer. Keystone’s management should see its board page wording first. The SB 253 re-check falls under the annual refresh retainer.
In one line
The report is the by-product. The system is what Northwood keeps.
After the brief’s own framing: “The draft is a byproduct. The library is the product.”
Next step: confirm the governance arrangements, then the report goes to Investor Relations.
Slide 19 notes. To close. The Week 6 brief framed it this way: the draft is a byproduct, the library is the product. For Northwood that means the LP report is this year’s output, and the system behind it, the library, the register, your dashboard and the re-run steps, is what you keep and use again. The one step that unlocks the report is yours: confirm the governance arrangements. After that it goes to Investor Relations under your protocol. Everything is in the portal, in its final version.