MH / CL Decisions

Individual KPIs discriminate hard and are close to bimodal, but averaging all of them pulls every deal toward the middle. The ladder (A<1.75 B<2.5 C<3.25 D<4.0) was set for the full 1-5 scale, so A and F are close to unreachable however the book performs.

composite spans 0.00-0.00 (0% of the 4.00 the scale allows), median 0.00. KPI readings: 0% at band 1, 0% at band 5. Grades today: A:0 B:0 C:0 D:0 F:0

M-4 Responsiveness and M-5 Transparency are judgements with no pipeline that could ever answer them; /report-card/manager was built for exactly this and holds no entries. Their weight is not idle - it is redistributed onto the two KPIs that do answer, so the Manager score is really a reporting-punctuality score wearing a broader name.

Manager band: 0% of its weight is live. M-1 and M-2 band all 0 deals; M-3 M-4 M-5 M-6 M-7 band none. Hand entries recorded: 44

The rubric that is PRINTED and the rubric that COMPUTES are different documents today. A KPI with a weight and no data does not lower a grade - it hands its share to whatever was measured, so a deal is graded on a narrower rubric than the one on the page.

100% of the overall grade is weight that bands no deal at all. By component: Performance 100% dead | Returns & Cash Flow 100% dead | Debt 100% dead | Value 100% dead | Manager 100% dead

A portfolio member is graded as its own deal today, on its own statements. That is right for an operating comparison and arguably wrong for a capital one: the money was committed to the portfolio, the debt is often cross-collateralised, and a Reporting Entity row is a Col-K bucket rather than a property at all - it has no operations to grade. Grading each member separately can show four D grades where one asset is dragging three healthy ones, or hide that in an average.

0 Active rows carry member_class "Reporting Entity" across 0 deals; 2 deals are named as part of a portfolio or family. Largest families: (none)

CF-6 is built and live on invested capital, which is our money and is the CF-1 denominator, so the two cash-flow KPIs stay comparable. The DIT also carries `initial_equity` (pref + common equity in the property) off the same REO tab, and the two do not agree - the SAME KPI reads "nothing outstanding" or "a full year of pref owed" for most of the book depending on which is used. `initial_pref_equity`, the textbook base, survived on almost none of the REO tab and cannot carry it - see the measurement.

CF-6 bands 0 of 0 deals on invested capital. Measured both ways across the graded book: invested capital puts 64 at band 1 and 16 at band 4; initial_equity puts 15 at band 1 and 62 at band 4. `initial_pref_equity` is populated on 15 of 15 properties.

Distributions are not tagged pref vs return-of-capital, so something has to stand in. CF-6 counts OPERATING distributions only, which never understates the obligation. Most waterfalls do let a refinance or a sale clear accrued pref, so that is conservative rather than neutral.

operating only: {}. Adding capital returns: {1: 41, 2: 4, 3: 1, 4: 5, 5: 31} - it moves 17 deals from the worst band to the best

The rates come off the REO tab, not out of the documents. Where an OA is on file the recorded term is checkable; where it is not, CF-6 is grading on a number with no document behind it. And there are deals where we hold the OA and never read the term out of it.

CF-6 bands 0 of 0 deals. Cross-checked against the Documents app on 28 Aug 2026: of the 82 carrying a recorded pref, 36 had an OA or LPA on file and 46 did not; a further 19 deals held an OA and had no recorded pref at all.

CL asked for a first pass derived rather than typed: Strategy from BUDGETED CAPEX on the REO tab, Lifecycle from how much of it has been SPENT to date. The method is sound and the cut points are a judgement - the same kind as a KPI cut-point, and the same reason they are not decided in a resolver. A deal labelled core rather than value-add is being described differently to anyone reading the grade.

Budgeted capex over purchase price across the 15 properties that carry both: median n/a, p25 n/a, p75 n/a, p90 n/a, max n/a. Under the proposed cuts that splits 7 core / 8 value-add / 0 opportunistic. COVERAGE IS THE REAL CONSTRAINT: only 15 of 15 properties have a budgeted capex at all, 15 have an actual, and 6 carry units_to_renovate - so a derived Strategy can only speak for a fraction of the book however the bands are cut.

Below the floor no letter is issued and the deal reads "unrated". Set too high it hides deals that are genuinely gradeable; too low it publishes a letter drawn from half a scorecard.

0 of 0 deals are below the floor and read unrated. Average coverage across the book is 0%

Held at CL's instruction: "financial accuracy should perhaps be based on later restatements but i dont have a good way of identifying these right now". It carries a full ladder and no weight until there is a rule for spotting one.

M-3 bands 0 of 0 deals today

Two of the four cut-points are NEGATIVE and the direction is "lower is better". A deal cannot restate fewer than nought times, so on a raw count the top two bands are unreachable and every deal that ever files lands in C or worse — the ladder would grade the measure, not the manager. Negative cut-points spaced like these read as a z-score or a variance against the book average, which is a different question from the one the KPI name asks. This is the same trap M-1 was caught in, where the ladder was in DAYS while the measure had become months; the code refuses to emit M-1 until the two agree (`_m1_ladder_is_months`) precisely because a mismatched ladder does not raise, it just bands confidently and wrongly. Worth settling BEFORE M-3 is wired, because once it emits a value the ladder stops being obviously wrong and starts being quietly wrong.

M-3 ladder as it stands: [-0.653333, -0.453333, 0.546667, 2.546667], direction "lower" — 2 of its 4 cut-points are below zero, so on any non-negative count that many bands are unreachable and the best grade available is C. It bands 0 of 0 deals. WHAT IT WOULD BE CUT AGAINST, on the mechanical rule: of 86 period cells 1 were refiled but only 0 changed NOI — the other 1 were resubmissions of the same figures, so counting refilings would overstate this 0.0x. Per deal (restatements:deals) 0:11; 11 of 11 deals that file have never restated. As a COUNT that puts 100% of them in the best band. As a RATE it reads differently: the deal with the most restatements (0) sits at 0% of its own filed periods, and 0 deals are worse on that measure than it is.

MANAGER_ENTRY_KIND marks M-4 and M-5 as judgements — hand entry is their permanent home — and M-6 and M-7 as TRANSCRIBED, meaning the manager already publishes the number and we simply cannot read it yet. That distinction only means something if the reading eventually happens, and today neither can: the rent roll holds a move-out date on a small fraction of its units and no row pairs a move-out with the following move-in, because a rent-roll row describes the CURRENT tenant. So approving the queued rent rolls (see rent-roll-queue) will not answer M-6 either. There is no maintenance or work-order table at all for M-7. Meanwhile nothing has been hand-entered for either, so the "on loan" label is carrying weight it has never earned.

Rent roll: 1012 units over 7 deals, 0 carry a move-out date, 0 usable move-out/move-in turn pairs. Maintenance-shaped tables in the database: 0. Hand entries recorded for M-6/M-7: 22

Both columns are carried and empty - they are not held on the investment row. The DIT is the obvious home, and until one is chosen the columns print "TBD".

both columns empty on all 0 deals

They unlock the leasing and renewal KPIs, which band almost nobody today. They are gated deliberately: the source as-of dates were found to be wrong and were corrected on the way in, so the first batch is worth sampling rather than bulk approving.

0 rent-roll ingests pending across 0 deals; 7 deals hold live rent-roll units today

Excel cannot re-apply an AutoFilter when a CELL changes — only VBA can. So the panel writes an Include? column, the filter is saved against it, and the reader presses Ctrl+Alt+L. A six-line Worksheet_Change macro would remove that step, at the cost of shipping .xlsm: the recipient gets a trust prompt, some mail gateways strip macro workbooks, and every other pack in the Report Center is .xlsx.

AutoFilter is live on all 61 columns, the panel drives the Include? column and the filter is saved against it. Multi-select works today by typing "A; B" into a selector.

CL asked for the call in as many words: "add what to do for projected returns when there is no UW or it has expired". CF-4 measures the trailing-twelve actual cash yield against what we underwrote for the same twelve months. Today a deal with no plan is simply ABSENT from the KPI - it lowers coverage and never the grade, which is the right refusal but not necessarily the right answer. An EXPIRED plan is treated the same way, and that is the case worth deciding: a ten-year underwriting on a twelve-year hold is not missing, it is finished.

CF-4 bands 0 of 0 deals; 0 carry no underwriting at all

CL asked why so many properties are blank, and for the T-12 NOI and cap rate to be shown so the working is visible. The working is now on the tab - NOI, cap rate and basis are all columns. The blank itself is a real choice: the mark this band reads is the manager's own line, READ rather than derived, so it cannot disagree with what was reported. `re_valuation_service` would compute one from NOI and a cap rate plus overrides - a different number by construction, and one nobody filed.

0 of 0 deals have a T-12 NOI; 0 of those have no filed value mark

D-8 divides a guaranteed amount by net equity, and the guaranteed amount is `property_debt.guarantee_pct` times the balance, falling back to K-1 Item K. Neither source distinguishes a full-recourse loan from a non-recourse loan carrying the usual carve-outs — a carve-out guarantee leaves the loan NON-RECOURSE for tax purposes while we have still signed something. CL reads Elmwood Terrace and Naches as truly recourse and most of the rest as personal guarantees with standard carve-outs, which on his reading should read zero.

D-8 bands 0 deals; 0 carry a recourse figure at all

CL asked for it in as many words: "For every section, there should be a compared to average and mean of all other assets (including with use of filter)." A comparison needs a peer group, and the four candidates give materially different answers: the whole book, the filtered selection, the same asset sub-type, or the same manager. "All OTHER assets" also means leave-one-out — a deal must not be in its own benchmark, or a large position flatters the average it is measured against.

0 deals graded across 0 sub-types and 0 managers; median weighted score 0.00. A filtered average is one SUBTOTAL row; a leave-one-out mean is a second column per band.

CL asked directly: "Report Timeliness, and Completness - how determined? are we considering if they sent financials but we haven't filed it?" Yes on M-1, and on M-2 by a different route. M-1 takes the LATER of filed and received, which can only help a manager, so a document label we fail to parse costs them nothing. M-2 cannot use arrivals — only about seven in ten document period labels parse to a month, and a missed parse would read as a GAP and penalise exactly the manager it is meant to protect — so it moves the WINDOW to the deal's own latest filing instead.

M-1 bands 0 deals, M-2 bands 0. 0 deals hold a statement that has arrived and is not filed.

CL asked for the call: "GB-ED decision as to how to log issues - manual, report, separate report?" The Review band holds Last reviewed, By and Commentary today and nothing writes to them except the override form, so an issue noticed while reading the workbook has nowhere to go but a person's memory.

0 of 0 deals carry any review note today

CL: "I think that Report Card grades should flow into the Portfolio Dashboard / Asset Allocation report - that was a GB-ED decision there but lets do that now include. Include overall grade as well as grade by component". The obstacle is SCOPE, not plumbing: the Report Card grades the deals it can score — those with filed Actual NOI — and the dashboard covers the whole book. A grade column there would be blank on most rows unless it says why.

0 deals are scoreable and 0 carry a grade, against the whole Active book — so a dashboard grade column would be populated on a minority of rows and must say why on the rest.