SFR Acquisition

Key

SFR Assumptions — live model

The sliders below are the SFR pro forma. Every parcel reprices live as you drag — pins re-color, the deal card refreshes, Market Data aggregates re-tally. Move the ROC chip to widen or narrow the visible pool.

Default filter posture on load: Status = Active · ROC ≥ 20% · Lot ≥ 5K SF · Lot W ≥ 40 ft · Lot D ≥ 100 ft · Acquisition ≤ $2M · Exit ≥ $1,200/SF · Land basis ≤ $500/SF · Slope ≤ 20%. Click Reset all in the header chip bar to return to this posture.

Map Guide

PIN COLORS (Return on Cost)

0–15% ROC

15–30% ROC

30–50% ROC

50–100% ROC

100%+ ROC (burn-lot territory)

Ghost pin — negative ROC at asking price

DEAL STATUS DOTS

Screening

Analyzing

LOI Submitted

Under Contract

Closed

Passed

OTHER MARKERS

Pulsing pin = listed within 7 days

Switch to the Assumptions tab above to tune the SFR pro forma — every parcel reprices live as you drag. The map opens on viable deals only (ROC ≥ 0% at asking price); use the ROC filter to widen.

SFR Acquisition — Strategy

We identify single-family-zoned LA County parcels — teardown or vacant — where the spread between land cost and a new-construction $/SF exit is large enough to pencil. We acquire, build one modern modular SFR per lot, and sell it. Baseline economics: $500/SF all-in ($425 hard + $75 soft, configurable) against a $1,000/SF+ target exit. Build-to-sell only. One home per lot for v1.

Market

This app covers LA County only (City + all incorporated cities). All listings, comps, and zoning data are LA-specific. Vacant Land + Single Family Residential are both admitted to the acquisition layer; vacant lots are the preferred build-to-sell target and run the pro forma off defaultHomeSf with demoCost = 0.

VHFHSZ: Very High Fire Hazard Severity Zone is an informational flag only in the SFR app, NOT a knockout. The Jan-2025 Palisades burn footprint is a priority target — post-fire vacant lots are surfaced via the burn-zone overlay (toggle in the filter bar) and the 🔥 badge on deal cards. Insurance + WUI code surcharge belong in the per-deal pro forma, not the eligibility gate.

Listing status: Captured at scrape time and refreshed on every rebuild. Redfin's STATUS field maps to a user-facing label: Active → “Active” (no pill), Contingent → “In Contract” (amber pill), Pre On-Market → “Coming Soon” (gray pill). Filter via the Listing Status segmented control under Filters; pill renders on the deal card header for non-Active deals.

Zoning coverage: 4-endpoint cascade (City of LA / Santa Monica / Malibu / LA County DRP) confirms zone for ~80% of slice parcels. Parcels in uncovered cities (Beverly Hills, WeHo, Culver City, Inglewood, Pasadena, Glendale, Burbank, Long Beach) fall back to the Redfin property-type mapping with a zoningUnconfirmed flag on the card.

Pro Forma Model

Every parcel runs through the SFR pro forma (see src/models/proforma.js):

Sell value = homeSf × exitPsf. sellingCosts = sellValue × sellingCostPct (default 4%).

Build cost = (hardPsf + softPsf + siteWorkPsf) × homeSf + demoCost. Defaults: $425 hard + $75 soft, $0 site / demo per Appendix C.

Financing cost = loanAmount × (rate÷12) × holdMonths × drawFactor. loanAmount = LTC × (land + build); default LTC 70%, rate 10%, hold 14 mo, drawFactor 0.60.

Total cost = landCost + buildCost + financingCost. Gross profit = sellValue − totalCost − sellingCosts.

Returns: ROC = grossProfit ÷ totalCost, where grossProfit = netSaleProceeds − totalCost and netSaleProceeds = sellValue × (1 − sellingCostPct). Selling costs reduce proceeds before profit is measured — never added back in the denominator. Margin = grossProfit ÷ sellValue.

Two Land Solvers

The acquisition decision always turns on “max land price.” The model exposes both forms:

(a) Simple — target gross return on net sales (transparent, default heatmap metric):
maxLand = netSales − buildCost − financingCost − (netSales × targetGrossReturnPct)
Default target 20%. landPctOfValue = maxLand ÷ sellValue is the gut-check ratio (Palisades-class baseline ~25.8%).

(b) Residual — target return on cost, time-aware:

maxLand = ((sellValue − sellingCosts) − allInBuildPsf×homeSf×(1 + roc×holdMonths÷12)) / (1 + roc×holdMonths÷12)

Default target ROC 25% annualized. Plug the residual maxLand back as landCost and the deal yields exactly the target ROC.

Deal signal = maxLand vs. asking. The gap is the opportunity.

Financing (debt inputs only)

Debt sizing: loanAmount = LTC × (land + build); default LTC 70% (range 65–80). Rate 10%. Origination fee 1.25%. These inputs drive financingCost (interest draw + origination) which flows into totalCost.

Why no equity / net-to-owner: The build-to-sell single-owner frame expresses returns as ROC (profit ÷ total cost) and Margin (profit ÷ sale price). Loan payoff and equity splits are intentionally omitted — they belong in a per-deal LP/GP structure that this tool doesn’t model.

Map Tools

Draw polygon — click the pencil icon to draw a custom boundary. Only listings inside the polygon are shown. Click × to clear.

Favorites — star a deal in the deal card to save it. Toggle “Favorites only” in the filter bar to isolate saved deals. Persists across sessions.

Deal status — use the dropdown in the deal card header to track workflow (Screening → Analyzing → LOI → Under Contract → Closed / Passed). Status dots appear on pins.

URL sharing — all active filters encode into the URL hash. Copy the URL to share a filtered view.

Layer Toggles

On-Market — active listings colored by Return on Cost

Sale $/SF — sold comp dots + neighborhood grid (median exit $/SF per 0.01° cell)

Fire / Flood / Quake — hazard overlay layers from state GIS data

🏛 Historic — Toggles HPOZ (purple shaded areas) and HCM (purple dots) overlays across the full map. Historic designations are hard exclusions for build-to-sell SFR (demo + new construction is precluded). Any listing within an HPOZ or HCM boundary is flagged ‘🏛 Historic — UNQUAL’ and removed from the viable deal pool.

Deal Card (LOCKED layout)

Header badges — left to right: DOM pill (days on market, gray; omitted if the listing has no DOM field), zone pill (R1 / LAND / R2 / R3), Vacant/Teardown · N-home pill, and a comp-confidence badge (High / Moderate / Low · N comps). Trailing micro-tags surface non-Active status (In Contract / Coming Soon) and Build B1 snapshot-diff signals (New, status changed, price ↓/↑) when the underlying fields are present. Edit lives at the right.

Four hero tiles (row 1) — Asking · Exit $/SF · ROC · Profit. ROC is shown in green, Profit tile uses a green wash. Asking and Exit $/SF flip to the reddish editable style only in edit mode.

Deal-math equation — one monospace line:
Land $X + Build $Y ($/SF × homeSf) + Demo $Z + Financing $F = $T all-in → Sell $S − Txn cost $C → $P profit. Demo is broken out of Build so the equation foots to Total cost. Financing now includes the loan origination fee (P1-6 fix: computeCostStack rolls loanAmount × loanFeePct into financingCost so totalCost foots).

Zoning strip — sits directly below the deal-math line. Driven by far_rules.json v2.1 (89 jurisdictions, 282-entry neighborhood map) — the JSON is the single source of truth; src/data/zoning-rules.js is a thin loader. Resolution chain: l.city.toLowerCase() → neighborhoodMap (e.g. “Woodland Hills” → los-angeles-city; “Topanga” → los-angeles-county-unincorporated) → jurisdictions[key] → merge zones[normalizeZone(l.zone)] over jurisdiction.default. Any miss falls to advisoryDefault {method:'verify', far:0.50}. Post-resolution demotion: LA-City listings whose zone has an H suffix OR carry the listing coastal flag demote method to verify (advisory, never clamps). Non-LA coastalCDP:true jurisdictions (Redondo, Manhattan Beach, Malibu, …) keep their method but never clamp.

4 maxBuildable branches (returns {sqft, branch, provenance, clamp}):

far — far × lotSf
coverage — effFarOverride × lotSf for per-floor-cap cities (e.g. Santa Monica — 30% ground + 20% second floor → eff. FAR 0.50). Otherwise coverage × lotSf × stories (stories from JSON; fallback floor(maxHeightFt/10)).
envelope — (lw − 2·sideFt) × (depth − frontFt − rearFt) × floors. depth uses surveyed l.ld when present, else lotSf / lw. floors = stories || floor(maxHeightFt/10). Width missing or footprint ≤ 0 → falls to fallback. Used for jurisdictions with sourced setbacks + height but no codified FAR (LA County unincorp, Title 22 §22.20).
fallback — 0.50 × lotSf. Used when jurisdiction is unmapped, method demoted to verify, or envelope branch is unrenderable.

Provenance label matrix (flag-never-fake in UI):

provenance:'computed' + no flags → normal mono numeral; CLAMP badge iff clamp:true.
'computed' + zoningUnconfirmed → advisory tier “~{n} SF · zone unconfirmed”, no CLAMP badge.
'est-dims' (sourced rule, but lotShape==='est' — synthetic 2.5:1 dims, envelope branch only) → advisory “~{n} SF · est. dims”.
'fallback' → verify-gray “~{n} SF · ~0.50 assumed”. Never styled like a computed number.

Coverage never renders as a bare FAR — the strip shows e.g. “cov 50% × 2 → eff. FAR ~1.00” or “cov 30%·2F → eff. FAR 0.50” for effFarOverride cities.

CLAMP eligibility (hard cap on Home SF): method === 'far' && confidence === 'high' && !l.zoningUnconfirmed && !demoted && !coastalCDP. In practice only LA-City non-H non-coastal confirmed-zone listings clamp. Everything else is advisory: 75 long-tail cities are honest verify stubs (advisory 0.50, low confidence), LA County unincorp is envelope-advisory (footprint computable but never clamps), Santa Monica / Long Beach / Burbank are coverage-advisory, Malibu / Manhattan Beach / Redondo are coastal-advisory. Header “Zoning” button opens the right drawer with the full per-jurisdiction table (method, ratio basis, setbacks, height, confidence, clamp eligibility, source) populated at runtime from the JSON.

Second tile row — Lot (W′ × D′) · Lot SF · Land $/SF · Home SF. Both tile rows are pinned to the same height via fixed tile padding + a fixed .dcp3-tv line-height (smaller-font values use .dcp3-tv-sm, same line-height). Land $/SF = asking ÷ lot SF (broker frame). Missing data renders as “—”. Home SF is a plain static tile by default; only in edit mode does it flip to the reddish editable input (bounds 400–12,000).

Expand chevron — boxed disclosure: 24×24 rounded square housing a chevron. Rotates 90° and turns green when Full underwriting is open.

▸ Full underwriting — collapsed by default. Two columns:

Deal economics (left) — dollar waterfall: Acquisition → Demo → Build (with $/SF × homeSf annotation) → Financing cost (with LTC·rate·hold annotation; includes origination fee) → Total cost. Then Sale → − Txn cost → Net sale proceeds. Then Profit → ROC → Margin. ROC = profit ÷ total cost. Margin = profit ÷ sale price. Acquisition + Demo + Build + Financing (incl. origination fee) sum exactly to Total cost.
Sensitivity (right top) — moved out of the collapsed view. Base + 3 downside rows stepping SENSITIVITY_DELTA_PSF ($100) below the base exit $/SF. Each row re-runs calculateProForma (P1-5 fix: resolved as window.calculateProForma || window.Models?.ProForma?.calculateProForma so the lookup never fails on load order). Viable badge green when ROC ≥ 0.
Max land offer (right bottom) — lead line: “The most you can pay for the land and still clear each target.” @ 25% ROC (residual, time-aware) and @ 20% margin (simple gross-return) solves, plus a room line vs asking.

Removed from card and exports — ROE, Owner equity, Loan @ LTC, Loan payoff, Net-to-owner, Land/value. The build-to-sell single-owner thesis surfaces ROC and Margin; the LP/equity strip is intentional dead weight cut from both the LOCKED layout and the bulk CSV export.

Footer — Sale comps · Maps · Redfin ↗ · Zillow ↗ · ★ Save · ✕ Unqual. Comp table opens via Sale comps.

Address Search

Address search — type any California address to jump to it. If it matches an active listing, the deal card opens.

No match — the map pans to the geocoded point and a brief popup confirms the address is not in the current on-market set.

Edit Mode

Enter / Exit — click “✎ Edit” in the deal card header. “✓ Done” saves overrides to localStorage. “↺ Reset” clears all overrides. Press Escape to revert to pre-session values.

Editable fields — ALL editable fields share the same reddish/orange style and surface ONLY in edit mode: Asking Price + Exit $/SF in row 1, Home SF in row 2, and Hard $/SF + Soft $/SF beneath the Deal-economics ledger when Full Underwriting is open. Out of edit mode every tile (including Home SF) is the plain neutral style — the always-on inline Home SF knob is gone.

Home SF wiring — the Home SF tile holds _editInput('homeSf', ...) which routes to _onEditInput, writes _editMode.values.homeSf, and persists to _overrideHomeSf on Done. Drives both build cost (hard+soft $/SF × homeSf) AND sale (homeSf × exitPsf); re-renders tiles, math line, Zoning strip cap pill, all sensitivity rows, and the expanded ledger from the single-source pro forma.

Validation — Home SF clamped to 400–12,000 (LOCKED bounds). Out-of-range values are silently rejected; the card never breaks. Other numeric fields parse digits-only and re-format on blur.

Live recalc — every keystroke (debounced 150ms) re-runs the full pro forma. Hero tiles, the deal-math equation, the 4-row sensitivity table, and the expanded ledger (Deal economics + Max land offer) all repaint in real time.

Persistence — overrides saved per-AIN in localStorage (sfr_overrides_<AIN>). An “Edited” badge appears on cards with saved overrides. Internally they write _overrideHardPsf / _overrideSoftPsf / _overrideHomeSf — the same keys resolveDealInputs consumes, so card edits actually apply.

Bulk CSV export — the “Export” button in the sidebar generates a CSV of all visible listings. When overrides are active, exports use the override values. Columns: listing facts (address, zone, AIN, city, asking, lot SF, lot width, slope, DOM, home SF, exit $/SF, comp count/confidence) → deal economics (acquisition, demo cost, build cost, financing cost, total cost, sale price, selling costs, net sale proceeds, gross profit, ROC %, margin %) → notes + URL. No equity, loan payoff, or net-to-owner columns.

Filters

Page-load defaults: ROC ≥ 0%, Lot Width ≥ 50 ft, Lot Depth — any, Exit $/SF ≥ $1200, Slope ≤ 20%, Asking ≤ $2M, Land Basis ≤ $500/SF, Zoning = R1 + LAND + R2/R3 vacant.

ROC — range slider + preset chips (50%/30%/0%). Gross profit ÷ total cost at asking price. Drives pin color and is the SFR deal signal. Defaults to ≥ 0% (viable deals only).

Lot Width — defaults to ≥ 50 ft. Parcels with no width data pass through.

Lot Depth — defaults to no minimum (any depth). Set a floor to filter out flag lots and oddly-shallow parcels. Parcels with no depth data pass through. Data field: l.ld.

Acquisition price ≤ / Exit $/SF ≥ — bound the asking-price and comp-derived exit ranges. Exit $/SF defaults to ≥ $975 (the lower bound of the new-build $1,000/SF+ market thesis).

Land basis $/SF ≤ / Slope ≤ — defaults to ≤ $1000/SF and ≤ 20% slope. USGS 1m LiDAR grades are stamped on every parcel via fetch_slopes.py; drag the slider to widen. Lots with slope ≥15% show a cost-uplift warning.

Lot Type (Vacant / Teardown) — segmented control: All (default) / Vacant / Teardown. Wired to l.isVacantLand. Persists across page reloads. Vacant lots carry $0 demo cost; teardowns add demo cost from the Assumptions slider.

Listing Status (Active / In Contract) — segmented control: All (default) / Active / In Contract. Reads l.status from Redfin (Active, Contingent → “In Contract”, Pre On-Market → “Coming Soon”). Captured at scrape time; refreshed on every rebuild. Persists across page reloads.

New since last refresh — checkbox (default off). When ON, restricts the visible set to listings stamped newSinceLast=true by the snapshot diff (i.e. absent from the prior pipeline run). Persists per browser. See New & Changed Listings for the underlying detection.

Zoning — defaults to R1 + LAND + R2/R3 vacant (single-family build-to-sell only). R4/MU off by default.

Burn-zone parcels only — OFF by default. When ON, restricts the view to parcels inside the Palisades fire perimeter. Burn-zone parcels are never auto-hidden — they are TARGETS for build-to-sell SFR.

VHFHSZ, Burn-Zone & Coastal Zone (Phase 2c)

VHFHSZ is a FLAG, not a filter. Stamped via local PIP against CAL FIRE GeoJSON. Surfaced on the card as ⚠ VHFHSZ flag; never drops a parcel.

Burn-zone = l.burnZone = "Palisades". Displayed as 🔥 BURN-ZONE TARGET. Surfaced, never suppressed.

Coastal Zone = l.coastal = true for parcels inside the CA Coastal Commission boundary. Displayed as 🌊 Coastal Zone — CDP required on the deal card. Informational only — never excludes. The Coastal Development Permit (CDP) adds 3–6+ months to entitlement. Toggle the coastal zone overlay in the filter panel to visualize the CCC boundary on the map.

Home SF per parcel — Home SF lives in the deal card’s second tile row. It is a plain static tile by default; clicking ✎ Edit in the card header flips it to the reddish editable input (400–12,000 SF). Typing a number updates _editMode.values.homeSf live, drives the FAR cap pill + pro forma, and persists to _overrideHomeSf on Done. Overrides the global Home Size slider for that one deal.

Comp Confidence (Phase 2c)

The confidence badge in the deal card header (High / Moderate / Low · N comps) reflects how reliable the cluster-derived Exit $/SF is: High = ≥10 comps within 0.5 mi; Moderate = ≥5 comps within 1.5 mi; Low = <5 comps OR radius >1.5 mi. Confidence is purely informational — it never changes the exit value or the ranking.

When confidence is Low, the badge turns red. No pin color or rank change occurs. Use it to flag deals for deeper comp research before LOI.

Visual Encoding

Pin size — larger pins = bigger lots (≥15K SF = large, ≥10K SF = medium, <10K SF = small)

Land basis $/SF — green < $120, yellow $120–$160, red > $160

Lot efficiency — buildable SF ÷ lot SF. Higher = more home per square foot of land.

Heatmap Color Scale (Phase 3b-PREP)

Rescaled to the realized SFR exit-$/SF distribution observed in the SM+Palisades slice (~$840 P5 → $3,540 P95). The band discriminates inside the $1.5K–$2.5K/SF meat of the market.

$840 — blue (cool / value)

~$1.5K — cyan

~$2.2K — green (mid Westside)

~$2.9K — yellow (luxury pocket)

$3,540+ — red (ultra-luxury / Palisades-cliff)

Listing Data Freshness

Listing data is a dated snapshot from the most recent pipeline run. Individual listings may have gone under contract, been withdrawn, or had price changes since the snapshot was taken. Always verify a listing is still active on Redfin before acting. Use the Redfin ↗ link in the card footer to confirm live status.

A footnote at the bottom of each deal card shows “Listing data as of [date]”. The text turns amber when the snapshot is more than 3 days old.

New & Changed Listings (Build B1 — snapshot diff)

Every pipeline run compares today’s Redfin pull against a committed rolling snapshot (data/listings_snapshot.json, keyed by Redfin URL) and stamps three fields on each listing:

firstSeen — ISO date the listing was first observed by the pipeline. Carries forward on every run.

newSinceLast — true when the listing key was absent from the previous snapshot. Surfaces in the card header as a green “New” badge and is filterable via the New since last refresh toggle in Filters.

statusChanged — {from, to} when Redfin status moved between runs (e.g. Active → Contingent). Surfaces as an amber “→ In Contract” micro-tag in the badges row.

priceChanged — signed integer dollar delta vs. the prior snapshot. Negative values surface as a green “Price ↓ $X” micro-tag; positive as a red “Price ↑ $X”.

Gone listings (in prior snapshot, absent today) drop from listings.js naturally and are written to data/new_gone_<YYYY-MM-DD>.json for the (future) Build B2 email pass.

First-run guard: on the very first pipeline run after this feature ships there is no prior snapshot — the run establishes the baseline and newSinceLast=false for every listing. Real diffs start on run 2. Legacy listings missing these fields render as if all three were unset (no badges, no filter hits).

Data Age Indicators

Colored dots in the header bar show when each dataset was last refreshed:

< 3 days old

3–7 days old

> 7 days old

Age unknown

Datasets tracked: Listings, Comps, Parcels.

Access

Invite-only access. Contact matt@lucidresi.com.

How Exit $/SF Is Calculated

Exit $/SF is the estimated sale price per square foot for a completed new-construction single-family home on this parcel. The comp universe is restricted to detached single-family residences only (no condos, no townhomes, no multi-family) and to sales within the last 24 months. Comps are classified into three condition tiers, and the exit price prefers the truest comp pool (T1-New) and falls back gracefully when that pool is thin.

prefer T1-New regression → T1-Reno regression → T1-New median → T1-Reno median → T2 regression × (1 + uplift) → T2 median × (1 + uplift)
uplift = 10% (canonical new-construction premium; T2-fallback paths only — exits reprice at next data refresh)

No floor. No ceiling. The model output is the number. Each cluster reports its exitSrc (which tier the price came from) so confidence can be judged at a glance.

SFR Comp Universe & Three-Tier Classification

Comps are spatially clustered (0.005° ≈ 0.35 mi grid cells) and bucketed into one of three tiers:

T1-New — Built within the last 5 years — the truest comp for a new-build exit; preferred when ≥3 prints exist in the cluster

T1-Reno — Near-new finish proxy: a mid-recency build (yb 2000–2020) with non-negative residual, OR an older build (yb ≥1900) selling >$100/SF above the neighborhood median (indicating a full remodel)

T2-Existing — Older / unrenovated baseline. Used as exit anchor only when T1-New and T1-Reno are both insufficient, and only with a +10% condition uplift (canonical new-construction premium)

A cluster needs at least 3 comps in a tier to fit a weighted size-regression. Below that, the tier median is used. T1-Reno requires no premium — recently renovated SFR finishes are comparable to new construction in $/SF terms; the gap shows up only in T2.

Sale $/SF Color Scale

The Sale $/SF grid layer and comp popup colors use a 14-bucket cold-to-hot gradient:

$1,500+

$1,400–1,500

$1,300–1,400

$1,200–1,300

$1,100–1,200

$1,000–1,100

$950–1,000

$900–950

$850–900

$800–850

$750–800

$700–750

$650–700

<$650

Size Curve

Within each cluster and each tier (with ≥3 comps), the pipeline fits a recency-weighted linear regression of price = intercept + slope × sqft across comp sale price and square footage, then predicts the price at the 3,500 SF target home size. Output is converted back to $/SF for display. Comps outside 1,200–6,000 SF are excluded from the curve fit; comps with extreme slopes (<$200 or >$1,500/SF) are rejected as bad fits.

The 3,500 SF anchor matches our default new-build SFR home size. Larger or smaller targets are interpolated from the same curve via the KEY-panel home-size override.

Recency Weights

0–6 months ago 1.00

6–12 months ago 0.85

12–18 months ago 0.70

18–24 months ago 0.50

Older than 24 months excluded

Comp Confidence

The model requires at least 5 comps. If fewer qualify, the search radius expands (up to 3.0 miles) and lower product tiers are added until 5 are found. Confidence is scored 0–100 based on comp count and method quality:

High (75+) — 10+ scored comps within search radius

Medium (50–74) — 5–9 scored comps

Low (25–49) — fewer than 5 high-quality comps; treat exit $/SF with caution

Very Low (<25) — sparse data; exit $/SF is unreliable

Outlier Rejection

The pipeline rejects regression slopes outside [$200, $1,500]/SF as bad curve fits, and excludes comps whose $/SF falls outside this same band. Comps outside 1,200–6,000 SF are excluded from the size-curve fit. Trimming is performed at the curve-fit stage, not as a post-hoc IQR pass.

T2 Condition Uplift (When Used)

When a cluster lacks enough T1-New or T1-Reno prints to anchor an exit, the model falls back to the T2 (older / unrenovated) pool with a +10% condition uplift (the canonical new-construction premium). This adjusts for the empirical gap between an existing-condition resale and what a new SFR build in the same submarket trades for. Exits reprice at the next data refresh.

The uplift is applied only when exitSrc is existing+uplift, existing-median+uplift, or blended+uplift. T1 paths (T1-New and T1-Reno) carry no premium. T1-Reno prints (recently renovated) are treated as comparable to new construction in finish level.

Exit Source Codes (Confidence Signal)

Each cluster exposes the path the exit number was derived from, so confidence can be judged at a glance:

new — Strongest: ≥3 T1-New comps in the cluster supported a size-regression fit

reno — Strong: ≥3 T1-Reno comps anchored the regression (recent renovation ≈ new finish)

new-median / reno-median — Moderate: ≥3 T1 prints exist but the regression did not pass sanity checks; tier median used

existing+uplift / existing-median+uplift — Limited: no T1 prints in the cluster; T2 baseline used with +10% condition uplift (canonical new-construction premium)

blended+uplift — Weakest: insufficient comps in any tier; blended cluster median used with +5% uplift (half the canonical premium). Treat exit with caution.

Comp Exclusion (Hide Toggle)

Hiding comps filters the comp table for inspection; exit $/SF stays at the pipeline value (full re-pricing on the remaining pool is a future enhancement). Excluded comps remain visible at reduced opacity and are flagged (EXCLUDED: TRUE) in CSV exports.

Exclusions are per-deal and per-session — they reset on page reload. To override exit $/SF, use edit mode on the deal card.

Zone Eligibility

Eligible parcels are single-family-zoned and buildable. In-scope zones: A, RA, RE, RS, R1, RU, RZ, RW1, and LAND (vacant residential).

R2 / R3 vacant land is also admitted when isVacantLand === true — these are buildable single-SFR lots within a multi-family zone. Improved or teardown R2/R3 (existing multi-family buildings) remains excluded due to RSO / tenant risk. In the current dataset this adds ~73 R2 and ~52 R3 vacant lots to the eligible pool.

Hard Exclusions

Two automatic disqualifiers — no override:

Historic Resource — Parcel listed or eligible under LA HCM, HPOZ (contributing), or CA Register of Historical Resources. Cannot build new SFR on a contributing historic resource. Flagged ‘🏛 Historic — UNQUAL’ on the deal card and removed from the viable pool.

Open Space / Protected Area — Conservation easement or similar restriction on development rights.

Informational Flags (never exclusions)

🔥 VHFHSZ — Very High Fire Hazard Severity Zone. Flagged on the deal card as an insurance cost input — not a knockout. Burn-zone and Palisades-perimeter parcels are acquisition targets for new-build SFR; insurance cost realities are surfaced on the deal card.

🌊 Coastal Zone — May trigger CEQA / California Coastal Commission review. Flagged informational only.

? Zoning Unconfirmed — Zoning inferred from Redfin property type rather than confirmed via city API (affects Beverly Hills, WeHo, Culver City, Inglewood, Pasadena, Glendale, Burbank, Long Beach). Verify with the city before acting on these parcels.

Lot Type

Corner Lot — Parcel touches 2+ named streets (OSM). Corner setbacks may differ; check local code. Shown as a blue tag on the deal card.

Alley Lot — Rear/side alley access detected (OSM highway=service, service=alley). May affect utility easements and rear setbacks. Shown as a blue tag on the deal card.

Standard — Single street frontage, no alley. No tag shown (default).

Note: Classification from OpenStreetMap. Verify against assessor maps.

Site Flags

⚡ Power Lines — Overhead power infrastructure detected within 80m via OpenStreetMap. May affect rear/side setback requirements or require utility relocation. Shown as an amber tag on the deal card. Verify with utility company and local building dept.

🏛 Historic Zones — Toggles HPOZ (purple shaded areas) and HCM (purple dots) overlays across the full map. Listings within these boundaries are hard-excluded and flagged on the deal card.

Deal Status