Watchline NYC — Accountability infrastructure for NYC housing enforcement
Case studies

The layers in practice

Four worked examples of the owner-identity layer doing its job — merging owners the record splits (a typo; a shared deed), and refusing to merge parties it only appears to connect (a shared office; a shared manager). The answer, in data, to "doesn't it just merge everything?"
Leads, not verdicts. Ownership and control links below are algorithmic inferences from public records — investigative leads to verify, not determinations of legal ownership. Recorded owners, deeds, court records, and managing agents are directly sourced; who controls or owns across them is inferred. Where a case merges parties (Escobar, AXL), that a shared origin means one present owner is a lead an investigator confirms — never a verdict. All names are drawn from public NYC filings.

Watchline separates three questions the record often conflates — who owns a building, what operational network it runs through, and who manages it. These cases show each boundary holding. (See also the flagship Steven Croman case file.)

1 · One owner, split by a typo merge

Who Owns What splits landlord Ramon Escobar's Bronx portfolio into two portfolios that carry the identical name and business address — because that one office (2432 Grand Concourse #504) is recorded in the registration data a dozen inconsistent ways, and Who Owns What links on the exact address string. Watchline keeps all 26 buildings together as one owner.

GroupingBuildingsLabel
WatchlineNYC1 owner26one Bronx portfolio
Who Owns What2 portfolios2 + 24both "RAMON ESCOBAR @ 2432 GRAND CONCOURSE #504"
One office, ~a dozen keys. Every form below is the same 2432 Grand Concourse #504 as filed in HPD — and because Who Owns What links on the exact string, each distinct one peels a building off: GRAND CONCOURSE 504 GRAND CONCOURS 504 GRAND COURSE 504 GRAND COCNOURSE 504 GRAND CONCOURSE 105 SHERMAN AVE (blank) …plus varying city and ZIP. Watchline reunites them by resolved identity, never the string.

Three of Escobar's Creston Avenue buildings — 2031, 2064 and 2070 — show the split cleanly: same owner, same office, yet Who Owns What files 2064 with the main portfolio and strands 2031 and 2070 in a separate one. Not on the typo — only 2070 carries the GRAND COURSE misspelling; 2031 has the correct street and splits off purely because its standardized ZIP was blanked (it was filed under a rejected out-of-area 10607), and WoW's name edge is ZIP-gated, so a blank ZIP can't rejoin the valid-10458 node. Watchline reunites all three on evidence that never reads the ZIP — the ACRIS held-deed signal sourced and Splink name-anchored resolution. Layer: owner-identity de-fragmentation. Explore the interactive map (toggle Watchline's one portfolio ↔ Who Owns What's two — the split-off dots are the two buildings whose ZIP standardization blanked), or read the full node-fragmentation analysis (why the raw table holds five nodes).

→ ① Who Owns What's edges only WOW PORTFOLIO · 24 BUILDINGS RAMON ESCOBAR ZIP 10458 · 24 buildings WOW PORTFOLIO · 2 BUILDINGS RAMON ESCOBAR blank ZIP · 1 bldg RAMON ESCOBAR blank ZIP · 1 bldg "GRAND COURSE" typo 2 portfolios (24 + 2) ② + identity edges (Splink · deed) WATCHLINE OWNER GROUP OG-93013 · 26 RAMON ESCOBAR ZIP 10458 · 24 buildings RAMON ESCOBAR blank ZIP · 1 bldg RAMON ESCOBAR blank ZIP · 1 bldg "GRAND COURSE" typo 1 owner (26 buildings) CONNECTED_BY_NAME / _ADDRESS — Who Owns What's edges (gated on exact ZIP) CONNECTED_BY_SPLINK / _DEED — identity edges that reunite the split (never read ZIP)
The same case, in the graph. With only Who Owns What's name/address edges — both gated on an exact ZIP — the blank-ZIP pair can't reach the ZIP-10458 node: two portfolios. Add the identity edges (Splink · registered-LLC · deed), which never read the ZIP, and it's one owner. Links are inferred — leads, not verdicts.

2 · One office, seven owners un-merge · address

Who Owns What groups 27 buildings as a single owner (its portfolio #183, labeled "Nathan Obstfeld") — but they are really ~10 different owners who merely share one registration office at 235 River Ave, Lakewood NJ. Watchline separates them into the seven distinct owners they actually are.

GroupingWhat connects them
Who Owns What1 portfolio · 27 bldgsa shared registration office (an address)
WatchlineNYC7 distinct ownersnothing — different people at a shared address

The office is a shared suite, not a registered agent and not one operator — so "same office" is not "same owner." Layer: operational nexus ≠ ownership. Explore the interactive map (toggle Watchline ↔ Who Owns What), or read the full over-merge analysis (the shared office, and why the seven owners stay apart).

This is the direct answer to "doesn't it just merge everything?" — here it does the opposite, correctly.

3 · One manager, three separate owners un-merge · manager

Five buildings whose one constant is the managing agent — Anya Levitov / Verus Real Estate, named on every registration. The recorded owners and apparent controllers all differ, and the owners even file from different offices (33 West 60th St; 240 Riverside Blvd CU2) — so the shared thread is management, not ownership.

BuildingRecorded owner sourcedApparent controller inferredManager sourced
349 West 53 StGates Overseas NYCMichael SchwarzVerus Real Estate
418 MacDonough StMichael SchwarzMichael SchwarzVerus Real Estate
124 South 2 StBollingen LLCOleg EvdokimenkoVerus Real Estate
1239 Putnam AveMulholland Rev. TrustDmitry SokolovVerus Real Estate
1111 Jefferson AveCem OlcerDmitry SokolovVerus Real Estate

Watchline puts none of these in one owner group; it splits them into three portfolios grouped by their apparent owners, and keeps the shared operator only in the management layer. If it had merged them, it would almost certainly have been wrong — the deeds name five unrelated parties. Layer: management ≠ ownership.

One manager, five buildings — three owners Verus / Levitov cluster · the managing agent is the only common thread; the owner-identity layer abstains Edges by layer MANAGED_BY — management (sourced) APPARENT_CONTROL — control (inferred) CONNECTED_BY_ADDRESS — shared office IN_OWNER_GROUP — identity (absent) Portfolio PF-…-67385 Portfolio PF-…-11 Portfolio PF-…-73027 shared office · 33 W 60th St shared office · 240 Riverside CU2 VERUS REAL ESTATE managing agent (Anya Levitov) · MANAGED_BY Michael Schwarz apparent controller Dmitry Sokolov apparent controller Oleg Evdokimenko apparent controller 349 W 53 StManhattan 418 MacDonoughBrooklyn 1239 Putnam AveBrooklyn 1111 JeffersonBrooklyn 124 South 2 StBrooklyn ⊘ IN_OWNER_GROUP — no owner group spans the five · the identity layer abstains Management connects all five buildings. Owner-identity connects none. Three separate owners behind one managing agent — the graph refuses to let “same manager” become “same owner.”
Every edge is drawn from the graph: MANAGED_BY ties all five to Verus; CONNECTED_BY_ADDRESS glues each owner's own office; no IN_OWNER_GROUP spans them. Control/ownership links are inferred — leads, not verdicts.

4 · Two houses, two owners, one deed merge · deed

Two adjacent row houses on one Flushing block — 43-58 and 43-60 164th Street (Queens) — bought together on a single deed. Today they register to two different people at two different addresses, filing two differently-named single-purpose LLCs, so Who Owns What files them under two unrelated owners — nothing it keys on is shared. Watchline reunites them into one owner group, on evidence Who Owns What never reads.

GroupingWhat connects them
Who Owns What2 unrelated portfolios (Brian Lin · Xianglian Wang)nothing it can see
WatchlineNYC1 owner group · both housesa single 2015 ACRIS deed

The tie is a transaction. In 2015 AXL HOME LLC bought both houses from the Cherot family estate on one deed sourced for $1.2M. Four years later, on a single day in 2019, AXL re-deeded each house into its own single-purpose shell — BRIDGEWOOD DEVELOPMENT LLC and HONG LI GROUP LLC — both at $0 sourced. Watchline's CONNECTED_BY_DEED veil-pierce inferred recovers the common origin the re-titling obscured, admitting it precisely because the onward transfers are nominal ($0), not sales. Layer: owner identity, read from the deed record. Explore the interactive map — two next-door houses that Who Owns What paints as two owners (toggle Watchline's one owner ↔ WoW's two).

Verified against real Who Owns What: the two houses sit in two distinct JustFix portfolios (#14133 and #55695), neither an aggregator lump — so this is a genuine WoW split the deed reunites, not an artifact of Watchline's own layer.

Here Who Owns What isn't wrong — it's blind: the only tie is a transaction, and it reads no deeds. Two different registrants after a $0 split could also be a partition between partners; the single-grantor chain and nominal price are strong common-control signals — a lead to verify, not proof of one present beneficial owner.
One message across all four: the owner-identity layer is precision-first. It keeps who owns cleanly separate from who operates through this office and who manages the building — merging a true owner the record fragments (even when the only trace is a deed), and declining to merge parties the record only appears to connect.

Deeper dives — full analyses

Longer, data-first write-ups behind three owner-identity results — each tracing the exact records, Who Owns What portfolios, and graph edges involved.

AnalysisWhat it shows
Steven CromanWho Owns What's 6 portfolios reunite into one Watchline portfolio — the same 128 buildings, none added (de-fragmentation).
Ramon EscobarWhy one operator becomes 5 nodes in landlords_with_connections — the node-level fragmentation behind the Escobar case above.
The Miller officeThe reverse: Who Owns What fuses 7 owners into one portfolio on a shared Lakewood office (degree 45); Watchline declines the aggregator-address merge — an anatomy of an over-merge.

And the mechanism beneath all of them: how the owner-identity edge layer works — the six edge types, their guards, and which layer consumes each.