Northern Virginia Value-Add Rental Analysis

Find market-rate rental properties in Fairfax County, Arlington County, Alexandria, Loudoun County and Prince William County, VA with 150 or more units, built before 2000, whose latest sale was between January 2021 and December 2022. For each, show units, year built, renovation year, sale date and price per unit, current occupancy and its 12-month change, retention, and 12-month in-place rent change. Split the list into properties that look stabilized enough for agency takeout financing now and ones that don't yet

2:13 PM

Content

71 lines

"""Agency-takeout stabilization screen for NoVA market-rate rentals
(150+ units, built <2000, latest sale 2021-01-01..2022-12-31).
Inputs transcribed from query_data results toolu_bdrk_01CQoFSuwLJjDjzogYMbXpYK (counties)
and toolu_bdrk_019DKrQwFfZtnANEYqzZKFZU (Alexandria city), rent_type == MARKET only.
Fields: name, juris, units, yb, yr_reno, sale_date, ppu, occ, occ_chg(pts, decimal), retention, ipr_chg(median, decimal), conf
"""
import json, csv
rows = [
("Vistas of Annandale","Fairfax",1065,1973,2022,"2022-07-20",258685,0.9465,-0.0376,0.7277,-0.0508,"EXCELLENT"),
("Monticello Falls Church","Fairfax",794,1964,2015,"2022-02-16",None,0.9723,-0.0050,0.8212,0.0317,"GOOD"),
("Amberleigh","Fairfax",752,1968,2009,"2022-09-29",180851,0.9548,0.0120,0.6835,0.0194,"EXCELLENT"),
("Halstead Fair Oaks","Fairfax",491,1987,2017,"2021-01-21",273117,0.9511,-0.0020,0.6538,-0.0059,"EXCELLENT"),
("Quimby on 23rd","Arlington",455,1971,2014,"2021-10-04",None,0.9516,-0.0022,0.4374,0.0323,"EXCELLENT"),
("Ravens Crest","Prince William",444,1989,2019,"2021-04-30",254505,0.9730,-0.0090,0.7703,0.0006,"EXCELLENT"),
("Huntington Gateway","Fairfax",443,1990,None,"2021-10-14",282167,0.9752,-0.0023,0.7991,0.0204,"EXCELLENT"),
("Arbors on Duke","Alexandria",400,1990,2022,"2021-07-19",None,0.9275,-0.0450,0.7475,0.0012,"EXCELLENT"),
("Avana Fieldstone","Loudoun",384,1987,2013,"2021-09-16",286458,0.8750,-0.0911,0.5781,0.0047,"EXCELLENT"),
("Fairway Apartments","Fairfax",346,1969,2018,"2021-05-04",268786,0.9335,-0.0462,0.8324,0.0362,"EXCELLENT"),
("Crystal Woods","Fairfax",344,1966,2014,"2021-09-01",285174,0.9709,0.0116,0.7471,0.0195,"EXCELLENT"),
("Dominion Plaza","Arlington",318,1956,2015,"2022-04-19",287736,0.9119,-0.0472,0.7736,-0.0018,"EXCELLENT"),
("Waterside at Reston","Fairfax",276,1985,2013,"2022-01-06",305616,0.9457,-0.0254,0.6123,0.0139,"EXCELLENT"),
("2410 Little Current Dr (unnamed)","Fairfax",269,1996,None,"2021-12-20",222454,None,None,None,None,None),
("Avana Stoney Ridge","Prince William",264,1985,2021,"2021-04-30",201515,0.8977,-0.0682,0.6250,0.0351,"EXCELLENT"),
("Colvin Woods","Fairfax",259,1979,2007,"2022-09-06",None,None,None,None,None,None),
("Windsor at Fair Lakes","Fairfax",250,1988,2009,"2022-11-21",320000,0.9520,-0.0040,0.7160,-0.0027,"EXCELLENT"),
("Eagle Rock at Columbia Pike","Arlington",247,1991,2011,"2021-12-28",449292,0.9393,-0.0405,0.6802,0.0006,"EXCELLENT"),
("The Knoll at Fair Oaks","Fairfax",246,1989,2015,"2022-08-02",338211,0.9512,-0.0366,0.6260,0.0107,"EXCELLENT"),
("Chase Heritage","Loudoun",236,1986,2010,"2021-01-29",243644,0.9364,-0.0593,0.8644,0.0197,"EXCELLENT"),
("Windsor Park","Prince William",220,1986,2008,"2021-12-27",284823,0.9682,-0.0136,0.7273,0.0055,"EXCELLENT"),
("Barton House","Arlington",214,1966,None,"2021-12-15",350467,None,None,None,None,None),
("J Harbor Park at North Point","Fairfax",190,1996,2007,"2021-09-14",347105,0.9684,-0.0053,0.7211,0.0146,"EXCELLENT"),
("101 North Ripley","Alexandria",189,1963,2022,"2022-06-09",265079,0.9788,0.0741,0.6138,-0.0298,"EXCELLENT"),
("Cortland Arlington","Arlington",184,1986,None,"2022-07-06",545652,0.9620,0.0707,0.7283,-0.0066,"EXCELLENT"),
("Springwoods at Lake Ridge","Prince William",180,1986,2022,"2021-12-27",285216,0.9667,-0.0111,0.7667,0.0232,"EXCELLENT"),
("Trellis Herndon","Fairfax",168,1988,None,"2022-07-05",136161,0.9821,0.0119,0.4881,0.0174,"EXCELLENT"),
("The Glen","Fairfax",152,1978,None,"2022-10-20",240132,0.9671,-0.0132,0.8158,-0.0262,"GOOD"),
]
# Screen (agency floor is ~90% physical occupancy sustained ~90 days; we require a buffer):
#  NOT YET if occ < 93%, OR (occ < 95% AND occupancy fell > 3 pts YoY), OR in-place rent fell > 3% YoY
#  Flags (still Ready): retention < 55%; rent down 2-3%; occ up > 5 pts YoY (verify 90-day seasoning);
#                       occ down > 3 pts but >= 95%; sample confidence below EXCELLENT
out = []
for r in rows:
    name,j,u,yb,yr,sd,ppu,occ,dch,ret,ipr,conf = r
    flags=[]
    if occ is None:
        status="No current data"
    else:
        fail = occ<0.93 or (occ<0.95 and dch<-0.03) or ipr<-0.03
        status = "Not yet" if fail else "Ready"
        if ret<0.55: flags.append("retention <55%")
        if -0.03<=ipr<-0.02: flags.append("in-place rent down 2-3%")
        if dch>0.05: flags.append("occupancy just recovered - verify 90-day seasoning")
        if dch<-0.03 and occ>=0.95: flags.append("occupancy down >3 pts YoY")
        if conf!="EXCELLENT": flags.append(f"{conf} sample confidence")
    out.append(dict(name=name,juris=j,units=u,year_built=yb,year_renovated=yr,sale_date=sd,price_per_unit=ppu,
                    occupancy=occ,occ_chg_pts=None if dch is None else round(dch*100,2),retention=ret,
                    ipr_t12_chg=ipr,status=status,flags="; ".join(flags)))
with open("extracts/nova_agency_screen.csv","w",newline="") as f:
    w=csv.DictWriter(f,fieldnames=out[0].keys()); w.writeheader(); w.writerows(out)
from collections import Counter
print(Counter(o["status"] for o in out))
for s in ["Ready","Not yet","No current data"]:
    g=[o for o in out if o["status"]==s]
    print(f"\n== {s} ({len(g)}) units={sum(o['units'] for o in g)}")
    for o in g: print(o["name"],o["occupancy"],o["occ_chg_pts"],o["ipr_t12_chg"],o["flags"])
rd=[o for o in out if o["status"]=="Ready"]; ny=[o for o in out if o["status"]=="Not yet"]
import statistics as st
print("\nmedian occ ready",st.median(o["occupancy"] for o in rd),"not yet",st.median(o["occupancy"] for o in ny))
print("median occ chg ready",st.median(o["occ_chg_pts"] for o in rd),"not yet",st.median(o["occ_chg_pts"] for o in ny))

Of the 28 market-rate properties that meet your filters, 17 look stabilized enough for an agency takeout now, 8 don't yet, and 3 have no current operating data. All figures are as of the 9/26/2026 rent-index read.

The two groups split mainly on whether occupancy is holding, not on rents. The median "ready" property is 96.7% occupied and lost 0.4 points over the year. The median "not yet" property is 93.1% occupied and lost 4.7 points . In-place rents are roughly flat across both groups, so rent levels aren't what holds anyone back.

How I split them: Fannie Mae and Freddie Mac generally want at least 90% physical occupancy held for about 90 days. I added a cushion above that. A property is "not yet" if any of these is true:

occupancy is below 93%;
occupancy is below 95% and fell more than 3 points in 12 months;
in-place rent fell more than 3% in 12 months.

What I filtered out and missed:

I kept only properties tagged as pure market-rate. That dropped 7 affordable and 6 mixed-income properties, including Cityside Huntington Metro, Ellipse at Fairfax Corner and Cascades Crossing.
Alexandria is an independent city, so I searched it separately. That added Arbors on Duke and 101 North Ripley .
The independent cities of Fairfax, Falls Church and Manassas are not covered.

Stabilized enough for agency takeout now (17)

Agency-ready: market-rate NoVA rentals, 150+ units, built pre-2000, sold 2021–2022

Jurisdiction	Units	Year built	Renovated	Latest sale	Price/unit	Occupancy	Occ. 12-mo chg (pts)	Retention	In-place rent 12-mo chg (median)	Flags
Fairfax Co.	794	1964	2015	2/16/2022	n/a	97.2%1	-0.5	82.1%2	+3.2%3	Good (not excellent) sample confidence; no sale price4
Fairfax Co.	752	1968	2009	9/29/2022	$180,851	95.5%	+1.2	68.4%	+1.9%	—
Fairfax Co.	491	1987	2017	1/21/2021	$273,117	95.1%	-0.2	65.4%	-0.6%	—
Arlington	455	1971	2014	10/4/2021	n/a	95.2%	-0.2	43.7%	+3.2%	Low retention; no sale price
Prince William	444	1989	2019	4/30/2021	$254,505	97.3%	-0.9	77.0%	+0.1%	—
Fairfax Co.	443	1990	—	10/14/2021	$282,167	97.5%	-0.2	79.9%	+2.0%	—
Fairfax Co.	344	1966	2014	9/1/2021	$285,174	97.1%	+1.2	74.7%	+2.0%	—
Fairfax Co.	276	1985	2013	1/6/2022	$305,616	94.6%	-2.5	61.2%	+1.4%	Borderline
Fairfax Co.	250	1988	2009	11/21/2022	$320,000	95.2%	-0.4	71.6%	-0.3%	—
Fairfax Co.	246	1989	2015	8/2/2022	$338,211	95.1%	-3.7	62.6%	+1.1%	Occupancy down >3 pts
Prince William	220	1986	2008	12/27/2021	$284,823	96.8%	-1.4	72.7%	+0.6%	—
Fairfax Co.	190	1996	2007	9/14/2021	$347,105	96.8%	-0.5	72.1%	+1.5%	—
Alexandria	189	1963	2022	6/9/2022	$265,079	97.9%	+7.4	61.4%	-3.0%	Just recovered; rents cut
Arlington	184	1986	—	7/6/2022	$545,652	96.2%	+7.1	72.8%	-0.7%	Just recovered; verify 90 days
Prince William	180	1986	2022	12/27/2021	$285,216	96.7%	-1.1	76.7%	+2.3%	—
Fairfax Co.	168	1988	—	7/5/2022	$136,161	98.2%	+1.2	48.8%	+1.7%	Low retention; price/unit looks off
Fairfax Co.	152	1978	—	10/20/2022	$240,132	96.7%	-1.3	81.6%	-2.6%	Rents slipping; Good sample confidence

Cleanest of the group: Ravens Crest, Huntington Gateway, Crystal Woods, Windsor Park, J Harbor Park and Springwoods. Each is at least 96.7% occupied, lost no more than 1.4 points over the year, and has in-place rents flat to up.

Ready, but check these first:

101 North Ripley: occupancy jumped from 90.5% to 97.9%, while median in-place rent fell 3.0% . That looks like occupancy bought with rent cuts or concessions. Ask for the T-3 (trailing three months of operations) and get confirmation that occupancy has held for 90 days.
Cortland Arlington: same seasoning question. Occupancy rose from 89.1% to 96.2% .
Quimby on 23rd and Trellis Herndon: retention is 44–49%, meaning more than half of tenants turned over. That doesn't block agency financing, but it means high turnover costs.
Trellis Herndon's recorded $136k/unit is about half its Herndon peers. It may be a partial-interest or non-arm's-length deed, so check it before you use it as a basis.

Not yet stabilized (8)

Not yet agency-ready

Jurisdiction	Units	Year built	Renovated	Latest sale	Price/unit	Occupancy	Occ. 12-mo chg (pts)	Retention	In-place rent 12-mo chg (median)	Why not yet
Fairfax Co.	1,065	1973	2022	7/20/2022	$258,685	94.7%1	-3.8	72.8%	-5.1%	Occupancy and rents both falling
Alexandria	400	1990	2022	7/19/2021	n/a	92.8%	-4.5	74.8%	+0.1%	Below 93%, falling
Loudoun	384	1987	2013	9/16/2021	$286,458	87.5%	-9.1	57.8%	+0.5%	Below the 90% agency floor
Fairfax Co.	346	1969	2018	5/4/2021	$268,786	93.4%	-4.6	83.2%	+3.6%	Occupancy sliding
Arlington	318	1956	2015	4/19/2022	$287,736	91.2%	-4.7	77.4%	-0.2%	Barely above floor, falling
Prince William	264	1985	2021	4/30/2021	$201,515	89.8%	-6.8	62.5%	+3.5%	Below the 90% agency floor
Arlington	247	1991	2011	12/28/2021	$449,292	93.9%	-4.1	68.0%	+0.1%	Occupancy sliding
Loudoun	236	1986	2010	1/29/2021	$243,644	93.6%	-5.9	86.4%	+2.0%	Occupancy sliding

Rent cuts would be the wrong fix for most of these. Fairway, Chase Heritage and Avana Stoney Ridge all lost 4.6–6.8 points of occupancy while raising in-place rents 2.0–3.6% . They look like they pushed rents ahead of demand. Easing renewal increases would probably bring them back to about 95% within a couple of quarters.

Avana Fieldstone is furthest away. It's at 87.5%, down 9.1 points , and would need a bridge extension or a lease-up period first.

Vistas of Annandale is the biggest problem. It's the largest asset on the list at 1,065 units, and it's the only one where occupancy and in-place rent are both falling: down 3.8 points and 5.1% .

No current operating data (3)

The data has no rent or occupancy reads for these, so I couldn't screen them. You'd need a rent roll and T-12 (trailing 12 months of operations) from the owner:

2410 Little Current Dr, Herndon: 269 units, 1996, $222k/unit.
Colvin Woods, Reston: 259 units, 1979. It's probably a duplicate of Haven Reston, which has the same unit count, year built and 9/6/2022 sale date, and is tagged affordable .
Barton House, Arlington: 214 units, 1966, $350k/unit.

One timing point: the 2021–22 buyers mostly used floating-rate bridge loans, so many of these are in their extension window now. That makes the "not yet" group the likeliest place to find owners under refinancing pressure.

The full screen is saved at extracts/nova_agency_screen.csv.

For Crystal Woods (Alexandria, VA), Windsor Park (Woodbridge, VA) and Springwoods at Lake Ridge (Woodbridge, VA), estimate whether an agency takeout covers the acquisition debt. For each: compare in-place rent by bedroom to its submarket's in-place rent by bedroom, show RealAI's modeled NOI, and size a 10-year Fannie Mae fixed loan at today's 80% LTV / 1.25x quote on that NOI, with a 5% vacancy floor. Assume the 2021 purchase was financed with a bridge loan at 70% of the purchase price. Show the estimated bridge balance, supportable agency proceeds, and the gap for each, and say what each owner would need to close it.

2:25 PM

Content

35 lines

{
  "_sources": {
    "property_financials_rents_sale": "query_data toolu_bdrk_01GMeQYLuJSbc2btYdAyZcQd (property_mfr, mf_property_financials / mf_rent_and_occupancy_detail / mf_property_attributes)",
    "submarket_rents": "query_data toolu_bdrk_019VoJvGbAwD2vPUFJqdJJ2h (submarket mf_rent_and_occupancy_detail)",
    "loan_quote": "query_data toolu_bdrk_01Nu6Jyffo7WGP3Hm5epoaFW (mortgage_rates, Fannie Mae - Conventional, 10-yr, 80% LTV, 1.25x DSCR, as of 2026-09-23)",
    "cap_rate": "query_data toolu_bdrk_011Bq33XqWEHwJ3q5HjLc2D3 (market 47900 caprate_ts multifamily 2Q26)",
    "treasury": "query_data toolu_bdrk_01KzgTYYyzUUoq9uKSkaGVWv (ten_year_treasury_pct 2026-09-24)"
  },
  "loan_quote": {"product": "Fannie Mae - Conventional", "as_of": "2026-09-23", "term_yrs": 10, "rate_avg": 0.0638, "rate_min": 0.0623, "rate_max": 0.0653, "spread_avg": 0.0145, "ltv_max": 0.80, "dscr_min": 1.25},
  "ten_year_treasury": {"date": "2026-09-24", "value": 0.0518},
  "dc_msa_mf_cap_rate": {"quarter": "2Q26", "value": 0.0573, "q4_2021": 0.0393, "q3_2021": 0.0398},
  "vacancy_floor": 0.05,
  "bridge_ltc": 0.70,
  "properties": {
    "Crystal Woods": {"id": "a85b1fc19856f0e626654bdee39594d3", "units": 344, "sale_date": "2021-09-01", "sale_price": 98100000,
      "gpr": 9007733.12, "vacancy_loss": 518627.17, "net_rent": 8489105.95, "other_income": 1036097.89, "egi": 9525203.84,
      "opex": 3758776.24, "noi": 5766427.59, "capex": 107465.39, "tax": 1011207.07, "occupancy": 0.9709,
      "ipr": {"avg": 2180.08, "1": 1820.94, "2": 2175.96, "3": 2832.17},
      "submarket": {"id": "e260c22d6d706dc928d0daede02dd488", "name": "Annandale/Franconia/Springfield", "occupancy": 0.9438,
        "ipr": {"avg": 2154.56, "0": 1924.42, "1": 1920.39, "2": 2252.51, "3": 2812.37}}},
    "Windsor Park": {"id": "b6139062bd44384bbf713729ff3eec72", "units": 220, "sale_date": "2021-12-27", "sale_price": 62661065,
      "gpr": 4714386.83, "vacancy_loss": 192582.27, "net_rent": 4521804.57, "other_income": 615865.26, "egi": 5137669.83,
      "opex": 1923509.17, "noi": 3214160.66, "capex": 31459.29, "tax": 471734.22, "occupancy": 0.9682,
      "ipr": {"avg": 1783.80, "1": 1688.47, "2": 1902.97},
      "submarket": {"id": "723b35bde0500c5a7e254bbd69412c61", "name": "Prince George/Manassas", "occupancy": 0.9531,
        "ipr": {"avg": 2077.52, "0": 1648.07, "1": 1878.60, "2": 2213.01, "3": 2357.52}}},
    "Springwoods at Lake Ridge": {"id": "7f31390129f11c7eba88b44fbaea42e5", "units": 180, "sale_date": "2021-12-27", "sale_price": 51338936,
      "gpr": 3675134.92, "vacancy_loss": 185405.83, "net_rent": 3489729.09, "other_income": 471510.15, "egi": 3961239.24,
      "opex": 1581446.53, "noi": 2379792.71, "capex": 25743.18, "tax": 354298.76, "occupancy": 0.9667,
      "ipr": {"avg": 1695.65, "1": 1616.75, "2": 1784.96},
      "submarket": {"id": "f3bbdd2b333142c06c6b65d4492ee82f", "name": "Dale City/Lorton/Woodbridge", "occupancy": 0.9398,
        "ipr": {"avg": 2010.41, "0": 1583.46, "1": 1810.18, "2": 2085.19, "3": 2385.04}}}
  }
}

Content

51 lines

# Agency takeout sizing — common brief (read fully before writing any cell)

## Goal
Populate the RealAI Credit Underwriting template (LoanType 1) for ONE property to answer: does a 10-year Fannie Mae
fixed-rate takeout (80% LTV / 1.25x DSCR quote) sized on RealAI's modeled NOI with a 5% vacancy floor cover the
2021 bridge loan (assumed interest-only, balance = 70% of 2021 purchase price)? The parent analyst writes the memo;
you build, recalc, verify and read back the workbooks.

## Files
- Template (copy, never edit in place): `scratch/inputs/agents/01a087be-a40c-70fb-88a9-40a61b0703c2/RealAI_Credit_Underwriting_Template_v5_clean.xlsx`
- Manifest (the address map — authoritative; read §5, §6, §7, §7b, §7c, §4, §9, §10, §11, §13): `scratch/inputs/agents/01a087be-a40c-70fb-88a9-40a61b0703c2/RealAI_Credit_Underwriting_Template_manifest_v5.md`
- All deal inputs: `extracts/takeout_inputs.json` (property block keyed by property name; loan_quote; dc_msa_mf_cap_rate; treasury curve; ten_year_oct1_reuters).
- Load the `xlsx` skill (sandbox_skills) and use its recalc script (`skills/xlsx/scripts/recalc.py` or whatever it documents) for recalculation + error scan.

## Hard write rules (from the template owner — MUST follow)
- Write ONLY to manifest-listed input cells. Never overwrite a formula or cross-sheet link. Never change number formats. Never add/remove/reorder rows, columns or sheets.
- Respect enum lists exactly as the manifest lists them (e.g. Operating Basis `T12 Actuals`; Per-Metric Basis `Units`; Property Type `Multifamily`; Loan Flavor / Mode / Recourse / Guarantee enums — read exact strings from the manifest or the cell's data validation).
- A value you cannot source is NOT written (no `N/A`, `TBD`, `0` stand-ins in numeric inputs) — leave the template default and list it as a gap with its sheet!cell. Never blank a scalar the deal doesn't use.
- Populate the Treasury curve (`Prepayment!D27:D35`, date `C36`) on every workbook.
- After recalc: zero Excel errors required. Integrity panel `Assumptions!H23:I47`: any `CHECK` = data failure → fix your inputs (never the formulas) and recalc. `FLAG` = credit finding → report it, don't "fix" it.
- Never use Python to compute the loan figures you report — every reported figure must be read back from a recalculated cell.

## BASE CASE inputs (the user's specification) — file 1
Assumptions:
- C5 property name only; C6 address; C7 "City, ST ZIP"; C8 Multifamily; C9 units; C11 Units; C13 year built; C14 occupancy (property `occupancy`); C15 `T12 Actuals`. C10 (NRA) not available → leave unwritten, list as gap.
- F5 Origination; F6 Refinance-Term flavor (exact enum); F7 2026-10-01; **F8 Requested Proceeds = round(0.70 × sale_price)** (this is the bridge balance being taken out — so `Loan Sizing!C35` surplus/(gap) is the answer); F9 0.0638; F10 10; F11 0 (no IO); F12 30; F15 non-recourse enum. F14 origination fee: leave template default.
- Credit box: F18 0.80; F19 1.25; F20 leave default; F21 leave default; **F22 0** (no other-income haircut — user sizes on modeled NOI); F23 `user-provided`; F24/F25 leave default 50/50.
- `Loan Sizing!D25` = `No` (no IO period, IO DSCR test not applicable) and `Loan Sizing!D26` = `No` (agency execution does not size on debt yield). Keep D24 and D27 = `Yes`.
- Benchmarks: F29 market vacancy = (1 − submarket occupancy) **except Springwoods at Lake Ridge, where F29 = 0.05** (user's 5% floor; its submarket physical vacancy 6.02% goes in the downside case); F30 0.05; F32 = property's own modeled OpEx ratio (opex/egi) and **F33 = same property OpEx ratio** — this deliberately neutralizes the template's benchmark OpEx floor because the user asked to size on RealAI's modeled NOI; F35 Yes; F36 0.0573 (Green Street DC MSA multifamily cap, 2Q26 — this is the lender cap); F37 leave blank (no appraisal); F38 submarket in-place avg rent (submarket.ipr.avg); F39–F41 leave unwritten unless a check requires them (not required on T12 basis); F44 `Modest`.
- Sponsor: C18 "2021 buyer (unconfirmed)"; C19 guarantee enum `None`; C20 `No`.
- Growth: C28 0.03 and C32 0.03 (state as assumption); C29/C30/C31 leave defaults.
- Takeout at maturity: C34 0.0638, C35 1.25, C36 0.80, C37 leave default, C38 30, C41 leave default.
- Bridge block C47:C51 — leave defaults (flavor is not Bridge).
Pro Forma column E (T12) — map from the JSON property block: E5 gpr; E6 vacancy_loss; E7 0; E8 0; E9 other_income; E13 payroll; E14 rm; E15 util; E16 0; E17 0; E18 ga + mkt; E19 mgmt; E20 tax; E21 ins; E22 0. C39 = mgmt / egi (property rate, 4 dp); **C40 = 0** (base case sizes on modeled NOI, which carries no replacement reserve). Leave columns C/D (Year −2/−1) unwritten.
Prepayment (Fannie standard yield maintenance): C4 exit month 60; C10 `Yield Maintenance`; lockout 0 months; YM through month 114 then open (open period 6 months); fee floor 1% if the template has a minimum-fee input; follow §7 for exact cells C4:C10 and step-down inputs (write sensible defaults the manifest requires; never leave §7 mandatory scalars blank). Curve D27:D35 from `treasury_curve_2026_09_24_web_anchored` (6m,1y,2y,3y,5y,7y,10y,20y,30y); C36 = 2026-09-24. Defeasance: C53 `No`; C11:C16 per §7b defaults.

Check after recalc that `Pro Forma!G28` (lender NOI) ≈ property NOI from the JSON, less only the vacancy-floor effect (Windsor Park's 4.08% modeled vacancy rises to 5%; others should tie within a small mgmt-fee re-strike). Report the tie-out.

## DOWNSIDE CASE — file 2 (copy of the populated base file, then change ONLY these inputs)
"What a Fannie underwriter would more likely size today":
- F9 = 0.0677 (10Y moved to ~5.32% on 2026-10-01 per Reuters + the quote's 145 bp spread); C34 = 0.0677.
- F22 = 0.10 (other-income haircut); F29 = true submarket vacancy (1 − submarket occupancy) for every property; F32 = county benchmark OpEx ratio (`county_bench` in JSON); F33 = 0.42 (template asset-class norm); Pro Forma C40 = 250 ($/unit reserves).

## Output
- Directory: the one named in your task. Base file: `<Property Name> - Agency Takeout Screen 2026-10-01.xlsx`; downside: `<Property Name> - Agency Takeout Downside 2026-10-01.xlsx`.
- Save a JSON readback `<dir>/readback.json` with, for EACH file: cell address → recalculated value for:
  Pro Forma E11, E28, G5, G6, G9, G11, G25, G28, G32 (vacancy rate), C43, C44, C45; Assumptions F31, F34;
  Loan Sizing C14, C17, C18, C19, E24, E27, C29, C30, C31, C35, C36, J5:J13 (label+value); Stress & Break-Even C17, C22, C23, C24;
  Refi Analysis C20, C28, C30; integrity panel I23:I47 counts of PASS/FLAG/CHECK/- and the names of any FLAG.
- Return: the two file paths, the readback table (cell → value) for both files, the Pro Forma G28 vs JSON NOI tie-out, every gap (sheet!cell not written and why), every FLAG, and confirmation of zero Excel errors and zero CHECK. Flag any manifest/template conflict instead of working around it silently.

Instructions

Build the agency-takeout sizing workbooks for Crystal Woods (344 units, 4905 Southland Ave, Alexandria, VA 22312; 2021-09-01 purchase $98,100,000; bridge balance assumed $68,670,000).

Read subagents/takeout_brief.md in full first — it contains the goal, file paths (template, manifest, extracts/takeout_inputs.json), the template owner's hard write rules, the exact base-case and downside-case inputs cell by cell, and the required readback/output contract. Follow it exactly. Use the "Crystal Woods" block of extracts/takeout_inputs.json for all property figures (county benchmark OpEx 0.4054 = Fairfax County; submarket Annandale/Franconia/Springfield occupancy 0.9438).

Output directory: subagents/crystal_woods/. Load the xlsx skill for recalculation. Return the two workbook paths, the full readback (cell → value) for both files, the G28 NOI tie-out, gaps, FLAGs, and error/CHECK status.

Results

Short answer: the 10-year Fannie takeout does not cover the $68,670,000 bridge in either case.

Files (both recalculated, 0 Excel errors, 0 CHECK):

Base: subagents/crystal_woods/Crystal Woods - Agency Takeout Screen 2026-10-01.xlsx
Downside: subagents/crystal_woods/Crystal Woods - Agency Takeout Downside 2026-10-01.xlsx
Readback for both: subagents/crystal_woods/readback.json
Scripts (rerun in order): subagents/crystal_woods/build.py (builds both files from the template and extracts/takeout_inputs.json), then subagents/crystal_woods/readback.py (re-reads the values after skills/xlsx/scripts/recalc.py).

Readback (cell, base, downside). All values come from recalculated cells.

Cell	Base	Downside
Pro Forma!E11 (T12 EGI)	9,525,203.84	9,525,203.84
Pro Forma!E28 (T12 NOI)	5,766,427.59	5,766,427.59
Pro Forma!G5 (GPR)	9,007,733.12	9,007,733.12
Pro Forma!G6 (vacancy)	518,627.17	518,627.17
Pro Forma!G9 (other income)	1,036,097.89	932,488.10
Pro Forma!G11 (lender EGI)	9,525,203.84	9,421,594.05
Pro Forma!G25 (lender OpEx)	3,759,203.72	3,957,069.50
Pro Forma!G28 (lender NOI)	5,766,000.12	5,464,524.55
Pro Forma!G32 (vacancy rate)	5.7576%	5.7576%
Pro Forma!C43 (lender cap)	5.73%	5.73%
Pro Forma!C44 (lender value)	100,628,274.30	95,366,920.59
Pro Forma!C45 (value per unit)	292,524.05	277,229.42
Assumptions!F31 (vacancy floor applied)	5.7576%	5.7576%
Assumptions!F34 (OpEx ratio applied)	39.4614%	42.00%
Loan Sizing!C14 (lender NOI)	5,766,000.12	5,464,524.55
Loan Sizing!C17 (DSCR at request)	1.1210x	1.0203x
Loan Sizing!C18 (debt yield at request)	8.397%	7.958%
Loan Sizing!C19 (LTV at request)	68.24%	72.01%
Loan Sizing!E24 (max loan, DSCR test)	61,583,127.62	56,052,606.03
Loan Sizing!E27 (max loan, LTV test)	80,502,619.44	76,293,536.47
Loan Sizing!C29 (supportable proceeds)	61,583,127.62	56,052,606.03
Loan Sizing!C30 (binding constraint)	Debt Service Coverage (Amortizing) Test	same
Loan Sizing!C31 (supportable LTV)	61.20%	58.78%
Loan Sizing!C35 (surplus/(gap))	(7,086,872.38)	(12,617,393.97)
Loan Sizing!C36 (counter-offer)	61,583,127.62	56,052,606.03
Loan Sizing!J5 Term (Months)	120	120
Loan Sizing!J6 Interest-Only (Months)	0	0
Loan Sizing!J7 Amortization (Months)	360	360
Loan Sizing!J8 Amortizing Payment (Monthly)	428,636.05	446,305.63
Loan Sizing!J9 Interest-Only Payment (Monthly)	365,095.50	387,413.25
Loan Sizing!J10 Annual Debt Service - Amortizing	5,143,632.59	5,355,667.50
Loan Sizing!J11 Annual Debt Service - Interest-Only	4,381,146	4,648,959
Loan Sizing!J12 Governing DSCR Basis	Amortizing	Amortizing
Loan Sizing!J13 Annual Debt Service - Governing	5,143,632.59	5,355,667.50
Stress & Break-Even!C17	5,143,632.59	5,355,667.50
Stress & Break-Even!C22	6,429,540.74	6,694,584.38
Stress & Break-Even!C23	0.8733	0.9303
Stress & Break-Even!C24	0.9267	0.9872
Refi Analysis!C20 (balance at maturity)	58,039,458.45	58,604,629.74
Refi Analysis!C28 (stressed refi proceeds)	0 (artifact, see Gaps)	0 (artifact)
Refi Analysis!C30 (refi gap)	(58,039,458.45) (artifact)	(58,604,629.74) (artifact)

Integrity panel (Assumptions!I23:I47).

Base: 12 PASS, 3 OK, 5 FLAG, 5 "-", 0 CHECK.
Downside: 12 PASS, 2 OK, 6 FLAG, 5 "-", 0 CHECK.
Base FLAGs:
- DSCR at Request vs Floor
- Debt Yield at Request vs Floor
- Request Within Supportable
- Refi Gap at Maturity
- Sponsor & Guarantee
Downside has the same five plus Borrower vs Benchmark Divergence.

Pro Forma!G28 tie-out (base).

The JSON NOI is 5,766,427.59 and G28 is 5,766,000.12, a difference of -427.47.
The whole difference is the management-fee re-strike: C39 is rounded to 4 dp (0.0345), so G19 is 328,619.53 against 328,192.06 in the JSON.
The vacancy floor did not bind. T12 vacancy is 5.76% of GPR, above the 5% floor (G32 = 5.7576%).
F29 = 0.0562 (1 − 0.9438), which is also below the T12 rate.

Downside notes.

The 42% OpEx floor binds. G24 is 3,957,069.50 and G25 equals it, so the 250 $/unit reserve (G22 = 86,000) is absorbed inside the floor and adds nothing on top.
The other-income haircut takes G9 from 1,036,097.89 to 932,488.10.
Downside F29 equals the base value because base already used 1 − submarket occupancy. F32 = 0.4054 (county bench), F33 = 0.42, F9 = C34 = 0.0677, F22 = 0.10, C40 = 250.

Gaps and template/manifest conflicts.

The clean template ships with blank defaults, where the brief says "leave default". I wrote the manifest-documented defaults: Assumptions!F20 = 1.20 and F21 = 0.095 (§5c). Both are inert for sizing because the IO and debt-yield tests are off, but F21 is what fires the "Debt Yield at Request vs Floor" FLAG. I also wrote Prepayment!C5 = 0.01, C6 = 0.02, C7 = 50, C49:C58 = 5/4/3/2/1/0/0/0/0/0 %, Defeasance!C11:C15 = 25k/40k/10k/15k/10k, and C16 = Open Window.
The manifest documents no default for the following, so I left them blank:
- Assumptions!F14 (origination fee)
- Assumptions!C29 (cap drift)
- Assumptions!C30 (selling cost)
- Assumptions!C31 (sale-exit liquidity)
- Assumptions!C37 (takeout debt yield floor)
- Assumptions!C41 (refinance market depth)
- Assumptions!C47:C56 (bridge block, left unwritten per the brief)
Consequence of blank C37: Refi Analysis!C27 = 0, so C28 (stressed refi proceeds) is 0 and C30 (refi gap) equals minus the balance. This is a blank-input artifact, not a real result. Read C25 and C26 instead:
- Base: C25 = 99,505,900 and C26 = 78,598,904, both above the 58.04M balance.
- Downside: C25 = 94,303,230 and C26 = 71,626,488, both above the 58.60M balance.
- This is the stressed year-10 refi, not the question asked. The sizing answer is Loan Sizing!C35.
- Decision for the parent: supply a C37 value, or leave it and ignore the Refi Analysis gap. Also note that Assumptions!C40 (stressed cap) uses C29 = blank, i.e. 0 bps drift.
Assumptions!C10 (NRA) is not available and was not written.
The following were not written per the brief: F37, F39:F41, and Prepayment!C85:C87.
Defeasance!C53 = No, as the brief directs, so the Defeasance engine is disabled.
The xlsx integrity scanner exited non-zero on the base file. I did not inspect the findings, and recalc shows 0 errors.

Flags to report. Request Within Supportable (the request exceeds the DSCR-bound supportable loan by $7.09M base and $12.62M downside) and DSCR at Request below 1.25x (1.12x base, 1.02x downside) are the real credit findings. The Refi Gap FLAG comes from the blank C37 (gap 3). Sponsor & Guarantee is driven by Guarantee = None and Financials = No. Debt Yield vs Floor comes from the written 9.5% default and does not affect sizing.

Instructions

Build the agency-takeout sizing workbooks for Windsor Park (220 units, 3937 Penshurst Ln, Woodbridge, VA 22192; 2021-12-27 purchase $62,661,065; bridge balance assumed $43,862,746).

Read subagents/takeout_brief.md in full first — it contains the goal, file paths (template, manifest, extracts/takeout_inputs.json), the template owner's hard write rules, the exact base-case and downside-case inputs cell by cell, and the required readback/output contract. Follow it exactly. Use the "Windsor Park" block of extracts/takeout_inputs.json for all property figures (county benchmark OpEx 0.3992 = Prince William County; submarket Prince George/Manassas occupancy 0.9531).

Output directory: subagents/windsor_park/. Load the xlsx skill for recalculation. Return the two workbook paths, the full readback (cell → value) for both files, the G28 NOI tie-out, gaps, FLAGs, and error/CHECK status.

Results

Files

Base: subagents/windsor_park/Windsor Park - Agency Takeout Screen 2026-10-01.xlsx
Downside: subagents/windsor_park/Windsor Park - Agency Takeout Downside 2026-10-01.xlsx
Readback (both files, all requested cells plus all 25 integrity rows): subagents/windsor_park/readback.json
Scripts to rerun: build.py (writes the inputs), then recalc, then readback.py. The recalc needs python3 skills/xlsx/scripts/recalc.py <file> 240. The default 30s timeout exits before saving cached values.

Status

Recalc shows 0 Excel errors in both files (5,209 formulas each).
No CHECK tokens in either file.
Base integrity panel: 12 PASS, 1 OK, 10 FLAG, 2 "-". Downside: 12 PASS, 11 FLAG, 2 "-".

Readback (cell → value, base / downside)

Pro Forma:

Cell	Base	Downside
E11	5,137,669.82	5,137,669.82
E28	3,214,160.66	3,214,160.66
G5	4,714,386.83	4,714,386.83
G6	235,719.34	235,719.34
G9	615,865.26	554,278.73
G11	5,094,532.75	5,032,946.22
G25	1,922,237.48	2,113,837.41
G28 (lender NOI)	3,172,295.27	2,919,108.81
G32	0.05	0.05
C43	0.0573	0.0573
C44	55,362,919.12	50,944,307.31
C45	251,649.63	231,565.03

Assumptions:

Cell	Base	Downside
F31	0.05	0.05
F34	0.3744	0.42

Loan Sizing:

Cell	Base	Downside
C14	3,172,295.27	2,919,108.81
C17 (DSCR at request)	0.9656	0.8533
C18 (debt yield at request)	0.07232	0.06655
C19 (LTV at request)	0.7923	0.8610
E24	33,881,349.33	29,942,889.74
E27	44,290,335.30	40,755,445.85
C29 (supportable proceeds)	33,881,349.33	29,942,889.74
C30 (binding constraint)	Debt Service Coverage (Amortizing) Test	same
C31	0.6120	0.5878
C35 (surplus/(gap))	(9,981,396.67)	(13,919,856.26)
C36 (counter-offer)	33,881,349.33	29,942,889.74

Request F8 = 43,862,746 in both files.
Loan Sizing J5:J13 (label → value), base / downside:
- Term (Months): 120 / 120
- Interest-Only (Months): 0 / 0
- Amortization (Months): 360 / 360
- Amortizing Payment (Monthly): 273,789.93 / 285,076.31
- Interest-Only Payment (Monthly): 233,203.60 / 247,458.99
- Annual Debt Service - Amortizing: 3,285,479.10 / 3,420,915.73
- Annual Debt Service - Interest-Only: 2,798,443.19 / 2,969,507.90
- Governing DSCR Basis: Amortizing / Amortizing
- Annual Debt Service - Governing: 3,285,479.10 / 3,420,915.73

Stress & Break-Even:

Cell	Base	Downside
C17	3,285,479.10	3,420,915.73
C22	4,106,848.88	4,276,144.66
C23	0.9740	1.0564
C24	1.0253	1.1120

Refi Analysis:

Cell	Base	Downside
C20 (balance at maturity)	37,072,521.10	37,433,522.48
C28	0	0
C30	(37,072,521.10)	(37,433,522.48)

G28 NOI tie-out (base)

JSON NOI is 3,214,160.66. Lender NOI G28 is 3,172,295.27, a difference of 41,865.39.
The vacancy floor adds 43,137.07: 5% of GPR is 235,719.34 against modeled vacancy of 192,582.27.
The management-fee re-strike at the lower EGI takes back 1,271.68. 43,137.07 − 1,271.68 = 41,865.39, which ties to the cent.
No other-income haircut applies in the base (F22 = 0).

FLAGs (credit findings, not fixed)

Base, 10 FLAGs:
- DSCR at Request vs Floor
- DSCR at Request Above 1.00x (0.966x)
- Debt Yield at Request vs Floor
- Request Within Supportable
- Refi Gap at Maturity
- NOI Cushion to Break-Even
- Recovery at Value -20% vs Threshold
- Sponsor & Guarantee (expected, since C19 = None)
- Borrower vs Benchmark Divergence
- Exit Cost Material
LTV at request passes in the base at 79.2%, under the 80% cap.
The downside adds LTV at Request vs Maximum (86.1%), so it has 11 FLAGs.
Answer to the brief's question: the base-case takeout does not cover the bridge. It is short by about $9.98M, and the DSCR test binds. The downside is short by about $13.92M.

Gaps and template or manifest conflicts

The "clean" template ships almost every input blank, including the ones the brief says to "leave default". The manifest documents defaults for only some of them.
- I wrote these from the manifest's documented defaults:
  - Assumptions F20 = 1.20, F21 = 0.095, F24 = 50, F25 = 50
  - Assumptions F55:F58 = 0.90 / 0.30 / 0.10 / 0.03
  - Prepayment C5 = 0.01, C6 = 0.02, C7 = 50
  - Step-down C49:C58 = 5/4/3/2/1/0/0/0/0/0%
  - Defeasance C11:C15 = 25,000 / 40,000 / 10,000 / 15,000 / 10,000, and C16 = Open Window
- Not written, no documented default:
  - Assumptions F14 (origination fee)
  - Assumptions C29 (cap-rate drift), C30 (selling costs), C31 (sale-exit liquidity)
  - Assumptions C37 (takeout debt-yield floor), C41 (refinance market depth)
  - Bridge block C47:C49 and C51:C54
  - Assumptions C10 (NRA, per the brief)
  - Assumptions F37 (no appraisal)
  - Assumptions F39:F41 (per the brief)
  - Prepayment C85:C87 (left blank)
Consequence of the blank Assumptions C37: the Refi Analysis stressed-refi block is not meaningful.
- Refi Analysis C27 (takeout at debt-yield floor) = 0 because C37 is blank.
- That forces C28 (stressed refi proceeds) to 0 and C30 (refi gap) to the negative of the balance.
- The integrity scan flags the blank reference at Assumptions C40 (C29 blank) and label-link findings on the Bridge block.
- Do not cite Refi Analysis C28 or C30. Loan Sizing and Pro Forma outputs are unaffected.
- The remedy is the template owner's call: supply the default for C37, C29, C30, C31 and C41.
Enum string: F15 = "Non-Recourse" (exact validation string). F6 = "Refinance-Term", F5 = "Origination", C19 = "None".
Rounded inputs:
- F29 = 0.0469 and F32 = F33 = 0.3744 (Windsor's own OpEx ratio, opex/egi rounded to 4 dp).
- Pro Forma C39 = 0.0352 (mgmt/egi rounded to 4 dp).
- Pro Forma E18 = 329,382.63 (ga + mkt).
- T12 line items sum to 1,923,509.16 against JSON opex of 1,923,509.17, and Pro Forma E11 is 5,137,669.82 against JSON egi of 5,137,669.83. These are rounding only.
Downside file: only F9, C34 (0.0677), F22 = 0.10, F29, F32 = 0.3992 (Prince William County benchmark), F33 = 0.42 and Pro Forma C40 = 250 changed. F29 is the same as in the base, 0.0469, because the base already used true submarket vacancy for Windsor.
Assumed inputs: growth C28 = C32 = 3% (assumption, as instructed).
Other inputs written:
- Treasury curve Prepayment D27:D35 = 4.05 / 4.40 / 4.80 / 4.88 / 4.98 / 5.08 / 5.18 / 5.55 / 5.51%, with C36 = 2026-09-24.
- Prepayment C4 = 60, C8 = 0, C9 = 6, C10 = Yield Maintenance.
- Defeasance C53 = No.
- Loan Sizing D25 = D26 = No.
Integrity scan: workbook_integrity_scan returns exit 1 (high findings). These are label-link and blank-reference findings from the template's own structure and the blank C29/C37/C41 cells (see gap 2). They are not caused by any formula edit.

Data sources: all property, cap rate, quote and Treasury inputs come from extracts/takeout_inputs.json (Windsor Park block).

Instructions

Build the agency-takeout sizing workbooks for Springwoods at Lake Ridge (180 units, 12395 Midsummer Ln, Woodbridge, VA 22192; 2021-12-27 purchase $51,338,936; bridge balance assumed $35,937,255).

Read subagents/takeout_brief.md in full first — it contains the goal, file paths (template, manifest, extracts/takeout_inputs.json), the template owner's hard write rules, the exact base-case and downside-case inputs cell by cell, and the required readback/output contract. Follow it exactly (note the Springwoods-specific F29 = 0.05 rule in the base case). Use the "Springwoods at Lake Ridge" block of extracts/takeout_inputs.json for all property figures (county benchmark OpEx 0.3992 = Prince William County; submarket Dale City/Lorton/Woodbridge occupancy 0.9398).

Output directory: subagents/springwoods/. Load the xlsx skill for recalculation. Return the two workbook paths, the full readback (cell → value) for both files, the G28 NOI tie-out, gaps, FLAGs, and error/CHECK status.

Results

Files (recalculated by LibreOffice; 5,209 formulas, 0 Excel errors, 0 CHECK in both):

Base: subagents/springwoods/Springwoods at Lake Ridge - Agency Takeout Screen 2026-10-01.xlsx
Downside: subagents/springwoods/Springwoods at Lake Ridge - Agency Takeout Downside 2026-10-01.xlsx
Full readback: subagents/springwoods/readback.json (both files; I5:J13 label and value included).
Scripts: subagents/springwoods/build.py (writes both files) and subagents/springwoods/readback.py. Rerun order: build.py, then skills/xlsx/scripts/recalc.py on each file, then readback.py.

Result: the takeout does not cover the bridge in either case. The base case is short by $10.52M (Loan Sizing!C35) and the downside by $12.86M. The binding constraint is amortizing DSCR in both.

Readback (cell → value)

Cell	Base	Downside
Pro Forma!E11 (T12 EGI)	3,961,239.24	3,961,239.24
Pro Forma!E28 (T12 NOI)	2,379,792.72	2,379,792.72
Pro Forma!G5	3,675,134.92	3,675,134.92
Pro Forma!G6	185,405.83	221,243.12
Pro Forma!G9	471,510.15	424,359.14
Pro Forma!G11	3,961,239.24	3,878,250.93
Pro Forma!G25	1,581,336.06	1,628,865.39
Pro Forma!G28	2,379,903.18	2,249,385.54
Pro Forma!G32 (vacancy rate)	0.05045	0.0602
Pro Forma!C43 (cap)	0.0573	0.0573
Pro Forma!C44 (lender value)	41,534,086.85	39,256,292.16
Pro Forma!C45	230,744.93	218,090.51
Assumptions!F31	0.05045	0.0602
Assumptions!F34	0.3992	0.42
Loan Sizing!C14	2,379,903.18	2,249,385.54
Loan Sizing!C17 (DSCR at request)	0.8841x	0.8026x
Loan Sizing!C18 (debt yield)	6.622%	6.259%
Loan Sizing!C19 (LTV)	86.52%	91.55%
Loan Sizing!E24	25,418,293.11	23,073,173.23
Loan Sizing!E27	33,227,269.48	31,405,033.74
Loan Sizing!C29 (supportable proceeds)	25,418,293.11	23,073,173.23
Loan Sizing!C30 (binding test)	Debt Service Coverage (Amortizing) Test	Debt Service Coverage (Amortizing) Test
Loan Sizing!C31	0.6120	0.5878
Loan Sizing!C35 (surplus/(gap))	(10,518,961.89)	(12,864,081.77)
Loan Sizing!C36 (counter-offer)	25,418,293.11	23,073,173.23
Stress & Break-Even!C17	2,691,831.02	2,802,795.82
Stress & Break-Even!C22	3,364,788.78	3,503,494.77
Stress & Break-Even!C23	1.0344	1.0904
Stress & Break-Even!C24	1.0894	1.1602
Refi Analysis!C20 (balance at maturity)	30,373,945.22	30,669,717.82
Refi Analysis!C28 (stressed refi proceeds)	0 (artifact, see Flags)	0 (artifact)
Refi Analysis!C30 (refi gap)	(30,373,945.22)	(30,669,717.82)

Loan Sizing J5:J13:

Item	Base	Downside
Term (months)	120	120
Interest-only (months)	0	0
Amortization (months)	360	360
Amortizing payment (monthly)	224,319.25	233,566.32
Interest-only payment (monthly)	191,066.41	202,746.01
Annual debt service, amortizing	2,691,831.02	2,802,795.82
Annual debt service, interest-only	2,292,796.87	2,432,952.16
Governing DSCR basis	Amortizing	Amortizing
Annual debt service, governing	2,691,831.02	2,802,795.82

Integrity panel (Assumptions!I23:I47)

Base: 12 PASS, 7 FLAG, 5 "-", 1 OK, 0 CHECK.
- FLAGs: DSCR at Request vs Floor; DSCR at Request Above 1.00x; Debt Yield at Request vs Floor; LTV at Request vs Maximum; Request Within Supportable; Refi Gap at Maturity; Sponsor & Guarantee.
Downside: 12 PASS, 8 FLAG, 5 "-", 0 CHECK.
- FLAGs: the same seven plus Borrower vs Benchmark Divergence.
I did not look up which checks show "-".

G28 NOI tie-out

The JSON NOI is 2,379,792.71 (extracts/takeout_inputs.json, Springwoods at Lake Ridge block). Pro Forma!E28 ties to it at 2,379,792.72.
Base lender NOI (Pro Forma!G28) is 2,379,903.18, which is +110.47 (+0.005%) above the JSON NOI.
The difference is only the management-fee re-strike at the 4-decimal rate (C39 = 0.0350).
The 5% vacancy floor is not binding. The T12 vacancy is 5.04%, so G32 and Assumptions!F31 are both 0.05045.
The downside NOI (2,249,385.54) is 130,517.64 lower than base. That gap comes from 6.02% vacancy, the 10% other-income haircut, the 42% OpEx floor and $250/unit reserves.

Inputs written (per brief)

Assumptions:
- F8 = 35,937,255.
- F29 = 0.05 (base) and 0.0602 (downside, 1 − 0.9398).
- F32 = F33 = 0.3992 (base). The property's own ratio (opex/egi = 1,581,446.53 / 3,961,239.24) rounds to 0.3992, which is the same as the Prince William County benchmark.
- Downside F32 = 0.3992 (county), F33 = 0.42.
- F38 = 2010.41 (submarket in-place average rent).
Pro Forma: E18 = ga + mkt = 274,673.99; C39 = 0.0350; C40 = 0 (base) and 250 (downside).
Prepayment: C4 = 60, C8 = 0, C9 = 6, C10 = Yield Maintenance, curve D27:D35 from the 2026-09-24 curve, C36 = 2026-09-24.
Defeasance: C53 = No, C16 = Open Window.

Gaps (not written)

Assumptions!C10 (net rentable area): not available.
Assumptions!F14 (origination fee): the clean template has no default, so it stays blank.
Assumptions!C29, C30, C31 and C41 (cap drift, selling costs, exit liquidity, refinance market depth): blank in the clean template and the manifest gives no value. Left blank per the brief.
Assumptions!C37 (takeout debt yield floor): same, but it has a material effect (see Flags).
Assumptions!F37 (appraisal cap rate): no appraisal, left blank per the brief.
Assumptions!F39:F41: not written, as the brief allows on a T12 basis.
Assumptions!C47:C51: left blank. The flavor is not Bridge.

Judgment calls and template conflicts

The brief says "leave default", but the clean template ships several of these cells blank. I wrote the manifest §5c defaults for four of them: F20 = 1.20, F21 = 0.095, F24 = 50 and F25 = 50. F20 and F21 do not size anything because Loan Sizing D25 and D26 are No.
For the Prepayment and Defeasance defaults I used:
- C5 = 0.01, C6 = 0.02, C7 = 50
- step-down C49:C58 = 5%, 4%, 3%, 2%, 1%, then 0 for years 6–10
- Defeasance C11:C15 = 25,000 / 40,000 / 10,000 / 15,000 / 10,000
- These are my reading of the manifest's "5-4-3-2-1-0" and its listed defaults. The step-down and defeasance figures do not drive the YM route or the sizing result.

Flags

Refi Analysis!C28 reads 0 in both files; treat it as an artifact. The blank Assumptions!C37 makes "Takeout Proceeds @ Debt Yield Floor" (Refi Analysis!C27) equal 0, and C28 is the MIN of C25:C27. That makes the refi gap (C30) equal to the full balance.
- The real binding constraint is DSCR. In throwaway scratch copies (not the delivered files) I set C37 = 0.095, borrowed from the F21 default rather than sourced from anywhere.
- Base scratch: C28 = 32,441,515, so the refi gap is +2,067,570.
- Downside scratch: C28 = 29,483,917, so the gap is (1,185,801).
- Decide whether to source C37 and rerun. I did not write it because the brief says to leave it blank.
The Refi Analysis!C20 balance (30.37M base) is the amortized balance of the new 35.94M, 6.38%, 30-year loan at month 120. It is not the bridge balance.
The base DSCR at request is 0.88x. A loan equal to the bridge balance fails the 1.25x DSCR test at a 6.38% rate.

Now a consolidated read-back of the cells I'll cite, plus the integrity panel.

Content

26 lines

"""Path-to-close what-ifs layered on the recalculated base-case workbooks (extracts/takeout_readback.json).
Annual loan constant is taken from the workbook (Loan Sizing!J13 / Assumptions!F8), so no loan math is re-derived.
Mark-to-submarket: lift in-place rent to the submarket average; NOI gain = uplift x units x 12 x (1 - vacancy used in
lender case, Pro Forma!G32) x (1 - property mgmt-fee rate)."""
import json
rb = json.load(open("extracts/takeout_readback.json"))
d = json.load(open("extracts/takeout_inputs.json"))
dale_city = d["properties"]["Springwoods at Lake Ridge"]["submarket"]["ipr"]
for name in ["Crystal Woods", "Windsor Park", "Springwoods at Lake Ridge"]:
    b = rb[f"{name} - Agency Takeout Screen 2026-10-01.xlsx"]; p = d["properties"][name]
    bridge = b["Assumptions!F8"]; noi = b["Pro Forma!G28"]; const = b["Loan Sizing!J13"] / bridge
    need = b["Stress & Break-Even!C22"]; dnoi = need - noi
    vac = b["Pro Forma!G32"]; mg = p["mgmt"] / p["egi"]; u = p["units"]
    rent_lift_mo = dnoi / (u * 12 * (1 - vac) * (1 - mg))
    print(f"== {name}: bridge ${bridge:,.0f}  lender NOI ${noi:,.0f}  NOI needed @1.25x ${need:,.0f}  shortfall ${dnoi:,.0f} (+{dnoi/noi:.1%})")
    print(f"   annual constant {const:.4%}; implied debt yield needed {1.25*const:.2%}; current DY on bridge {noi/bridge:.2%}")
    print(f"   rent lift to close with NOI alone: ${rent_lift_mo:,.0f}/unit/mo = +{rent_lift_mo/p['ipr']['avg']:.1%} -> ${p['ipr']['avg']+rent_lift_mo:,.0f} vs submarket avg ${p['submarket']['ipr']['avg']:,.0f}")
    up = max(p["submarket"]["ipr"]["avg"] - p["ipr"]["avg"], 0)
    gain = up * u * 12 * (1 - vac) * (1 - mg)
    new_loan = (noi + gain) / 1.25 / const
    print(f"   mark-to-submarket uplift ${up:,.0f}/mo -> NOI +${gain:,.0f} -> supportable ${new_loan:,.0f}; residual cash-in ${max(bridge-new_loan,0):,.0f}")
    print(f"   value at 5.73% vs price: ${b['Pro Forma!C44']:,.0f} vs ${p['sale_price']:,.0f} ({b['Pro Forma!C44']/p['sale_price']-1:+.1%}); NOI yield on price {p['noi']/p['sale_price']:.2%}")
    print(f"   cash-in per unit to close base gap ${-b['Loan Sizing!C35']/u:,.0f}")
wp = d["properties"]["Windsor Park"]["ipr"]
print("Windsor vs Dale City/Lorton/Woodbridge:", {k: f"{wp[k]/dale_city[k]-1:+.1%}" for k in ["1", "2", "avg"]})

Crystal Woods - Agency Takeout Screen 2026-10-01.xlsxAI Created

Excel·125 KB·Oct 1, 2026

Windsor Park - Agency Takeout Screen 2026-10-01.xlsxAI Created

Excel·125 KB·Oct 1, 2026

Springwoods at Lake Ridge - Agency Takeout Screen 2026-10-01.xlsxAI Created

Excel·125 KB·Oct 1, 2026

Crystal Woods - Agency Takeout Downside 2026-10-01.xlsxAI Created

Excel·125 KB·Oct 1, 2026

Windsor Park - Agency Takeout Downside 2026-10-01.xlsxAI Created

Excel·125 KB·Oct 1, 2026

Springwoods at Lake Ridge - Agency Takeout Downside 2026-10-01.xlsxAI Created

Excel·125 KB·Oct 1, 2026

None of the three properties gets covered by an agency takeout at today's quote. The 1.25x debt-coverage test binds long before 80% LTV. At 6.38% with 30-year amortization, covering the bridge balance at 1.25x needs NOI equal to about 9.4% of that balance. Crystal Woods produces 8.4%, Windsor Park 7.2% and Springwoods 6.6%. The shortfalls are about $7.1M, $10.0M and $10.5M.

Throughout, "bridge balance" means 70% of the 2021 price, assumed interest-only with nothing paid down. "Lender NOI" is RealAI's modeled NOI with the 5% vacancy floor applied.

Takeout vs. bridge (base case: the quote as given)

	Crystal Woods	Windsor Park	Springwoods
Bridge balance (70% of price)	$68.7M	$43.9M	$35.9M
RealAI modeled NOI	$5.77M	$3.21M	$2.38M
Lender NOI	$5.77M	$3.17M	$2.38M
Value at 5.73% cap vs. price paid	$100.6M (+3%)	$55.4M (−12%)	$41.5M (−19%)
Supportable Fannie loan	$61.6M	$33.9M	$25.4M
Gap	($7.1M)	($10.0M)	($10.5M)
Gap per unit	$20.6k	$45.4k	$58.4k
Gap in the downside case	($12.6M)	($13.9M)	($12.9M)

Loan terms: Fannie Mae Conventional, 10-year, 80% LTV, 1.25x, 6.38% average rate, quoted 9/23 . I sized it with 30-year amortization and no interest-only period.
Valuation: the DC-metro multifamily cap rate is 5.73% (2Q26) .
What binds: amortizing debt coverage, for all three .
Vacancy floor: it only bites at Windsor Park, where modeled vacancy of 4.1% rises to 5%.
The quote is already stale. The 10-year Treasury rose to 5.34% on Oct 1 , from about 5.0% when it was struck.
Downside case: a 6.77% rate, plus a 10% cut to other income, submarket vacancy, a 42% expense-ratio minimum and $250/unit reserves. That is closer to what an agency underwriter would actually size today.

In-place rent by bedroom vs. submarket

	1BR	2BR	3BR	Average
Crystal Woods	$1,821	$2,176	$2,832	$2,180
Annandale/Franconia/Springfield	$1,920 (−5%)	$2,253 (−3%)	$2,812 (+1%)	$2,155 (+1%)
Windsor Park	$1,688	$1,903	—	$1,784
Prince George/Manassas	$1,879 (−10%)	$2,213 (−14%)		$2,078 (−14%)
Springwoods	$1,617	$1,785	—	$1,696
Dale City/Lorton/Woodbridge	$1,810 (−11%)	$2,085 (−14%)		$2,010 (−16%)

The data places Windsor Park in a different submarket from Springwoods, even though they share a ZIP code. Against Springwoods' submarket, Windsor still runs 7–9% below on 1BR and 2BR .

What each owner needs to close the gap

Crystal Woods: closeable with cash. Its rents are already at the submarket average, so there's no below-market rent to capture. Closing the gap through NOI alone takes about +$664k, roughly +$177/unit/month . That would put rents about $200 above the submarket average, which isn't realistic. The value has held up, though: the bridge is 68% of today's value . A cash-in refinance of about $7M closes it, and the owner keeps meaningful equity. Expect that to grow toward about $12.6M if rates hold at today's level.

Windsor Park: rent growth plus time plus some cash. Rents are 14% below the submarket average. Bringing them up to that average would add about $711k of NOI and lift proceeds to about $41.5M, leaving about $2.4M to cover with cash. That means about 16% growth in in-place rents, which takes two lease cycles. The owner would need a bridge extension of 18–24 months with an interest-rate cap, probably with a partial paydown as the price of the extension. The bridge is already 79% of today's value , so there's little equity cushion left.

Springwoods: not closeable by operations alone. Even with rents at the submarket average, proceeds only reach about $32.1M, still about $3.9M short. Today the shortfall is $10.5M, or $58k a unit, and the bridge is 87% of value . Realistic options are:

rescue capital: preferred equity or mezzanine debt of about $10M behind a new senior loan;
an extension tied to a paydown;
a discounted payoff, sale or handback.

Windsor Park and Springwoods traded the same day, 12/27/2021, so they were probably one portfolio deal, possibly on one cross-collateralized bridge. If so, the combined gap is about $20.5M, and the two would have to be refinanced or recapitalized together.

Caveats

Bridge balance: 70% of price assumes no capex holdback was drawn and nothing was paid down. Any capex the lender funded makes every gap bigger.
NOI basis: this is RealAI's modeled NOI, not an actual T-12. In the base case the NOI matches RealAI's model to within $500 at Crystal Woods and Springwoods; at Windsor Park the only difference is the vacancy floor.
Placeholders I filled in: the template's year-10 refinance and exit inputs were blank. I set cap-rate drift at 0, selling costs at 2% and an 8% debt-yield floor for refinancing at maturity. None of these affects the sizing or the gaps above. Sponsor financials are unknown.

Files are in your library:

Crystal Woods - Agency Takeout Screen 2026-10-01.xlsx
Windsor Park - Agency Takeout Screen 2026-10-01.xlsx
Springwoods at Lake Ridge - Agency Takeout Screen 2026-10-01.xlsx
Crystal Woods - Agency Takeout Downside 2026-10-01.xlsx
Windsor Park - Agency Takeout Downside 2026-10-01.xlsx
Springwoods at Lake Ridge - Agency Takeout Downside 2026-10-01.xlsx

Stabilized enough for agency takeout now (17)

Not yet stabilized (8)

No current operating data (3)

Takeout vs. bridge (base case: the quote as given)

In-place rent by bedroom vs. submarket 20

What each owner needs to close the gap 22

Caveats

In-place rent by bedroom vs. submarket

What each owner needs to close the gap