Approved by Fede — Aug 20, 2026

Card & Record IA — per-stage spec + full 503-company inspection

v2 · 19 Aug 2026 (Mountain Time). Overnight follow-up to the 18 Aug card-IA audit: this pass defines exactly what a card and a full record should say at each of the 7 sales stages, then checks that spec against every one of the 503 companies on the board. Sources: 4 independent review passes (layout, sales process, writing, visual design) plus a company-by-company data check — 193 confirmed data problems found across the board (195 originally logged, 2 cleared on re-check — see note below), 153 of them detailed below.

Already fixed overnight

Every card past first contact now shows when we last touched the company.
The "still don't know" lists and open-question blocks are gone from the page entirely — not hidden one click deeper, actually removed from what renders.
CONAM (our furthest-along live trial) no longer wears the wrong "passed on this company" flag.
The call queue now leads with real targets — all 10 top slots are our best-fit tier, not a mix.
Real ID numbers (license, parcel, EIN, filing, UUID) no longer get replaced with the nonsense text "a phone number" — checked across all 503 companies, zero left.
A next-step line now automatically flags itself "may be outdated" the moment a newer, contradicting update comes in, instead of silently going stale.
Pasted internal notes, "Fede, [date]:" prefixes, and other leaked internal labels are cleaned out of 100+ records.
A bug that broke numbered/bulleted text apart mid-list is fixed.
Two of the "different company's info bleeding into this record" findings were re-checked overnight and cleared — see the note under the mistake table below.
Building tables now count only currently-owned buildings, instead of silently including sold or unbuilt ones — this was the cause of roughly 30 of the wrong-total findings below.
Contact matching now always prefers the confirmed decision-maker over a loose name match in old notes.
The "third-party managers" label now reads the verified field instead of scanning notes for the phrase — it no longer gets the answer backwards.
"Ruled out" can no longer appear on any company that's been contacted.
Internal tags and pasted-note prefixes are stripped from all rendered text, not just some of it.
Mid-word cutoffs, the doubled "CA, CA" bug, and false sentence-splits on things like "St." and "e.g." are all fixed.
The only work still open on this page: the 7 decisions below, and the per-stage redesign itself.

Heads up: a few items further down in this report were written down before these fixes shipped — most notably the parts about the "still unknown" text folding rather than disappearing, the "a phone number" placeholder, stale next-step lines, and pasted internal notes. Those are already fixed. Don't double-count them as still open.

Earlier decisions on this page (already acted on)

From the 18 Aug pass: giving each call-queue card a one-line, company-specific opener instead of our scoring rubric — shipped. Showing "last touched N days ago" on cards — shipped (see box above). Putting warning flags (like "can't integrate with their software") directly on cards — Fede turned this down; those stay record-only. Trimming the long "still don't know" writeups to one line each — shipped, then tightened further by the rule below: that content doesn't render on the page at all anymore, full stop.

The 5-minute version

There's no real difference between stages today — every company gets the same six blocks. A brand-new, never-contacted company and a live trial customer look almost identical. None of the stage-specific facts a seller actually needs (deal terms, who's championing us internally, what's blocking a close, what would bring a lost deal back) have a real place to live yet.
The good writing already exists — it just doesn't reach the screen. One prospect (ROM Investments) has a genuinely sharp, specific opening line sitting in our research notes, unused, while the card shows generic scoring-engine boilerplate instead. Same problem hides a live deal's real, specific requests under old research prose.
"Everything else" was never deleted — it just moved one click deeper. Every record still contains the exact rambling research notes and open-question lists we said we were removing. They're folded, not gone. Folding isn't deleting.
The top summary and the folded detail repeat the same facts, on every single record — confirmed by hand on 24 records and across the full 503-company check.
Underneath the writing problems is a real data-accuracy problem: 153+ confirmed mistakes. Numbers that contradict each other on the same screen (a "verified" 440-unit company whose own building list adds up to 604), sentences that cut off mid-word, real ID numbers replaced with the nonsense text "a phone number," a live trial customer wearing a "passed on this company" badge, and one executive's name showing up on a company he doesn't even work for. This is the layer that has to get fixed before the writing rules matter at all.

Per-stage spec — what a card and a record should say, in order

Left column = the queue card (a few lines, glanceable). Right column = the full record (one screen, opened on click). Everything not listed here stays in the data file — it never renders on the page. Tiering traces to How Our Prospects Operate (Aug 15, 96-company operating breakdown).

Global rules

Don't show it if it's useless. (Fede, 2026-08-20) Everything on a card or record must change what the reader does next. Self-evident, redundant, empty, or process information doesn't render.
Internal tool — assume smart users. No explainer copy, no legends for the obvious, no captions teaching the UI. Terse noun labels. One tooltip max, for a genuinely coded mark (✓/≈).
One fact, one place — header chips are the only home for units/software/HQ/contact; nothing below repeats them.
No leftover bucket — every fact earns a named slot or stays in the data file. "Everything else" is retired.
No open-question lists, no research narration, anywhere in the built page — folded or not.
Contacts by name + role, never "the decision-maker," once known.
Every empty state is specific and real. Never blank, never a shrug.
Each stage's one loudest fact gets real visual weight — larger or bolder, not just word order.
Size fit and deal status are different facts. "Ruled out" never appears once a company is contacted or later.
The warm-intro chip renders whenever any warm relationship exists — a third-party connector, an intro in flight (e.g. "waiting on Eli's intro to his cousin," "thru Jenna Langel"), or a direct relationship ("Sean knows him").

Stage 1 · 489 of 503 companies

Not yet contacted

Card = who to talk to next. Record = 30-second prep. Scored and ranked, no outreach sent yet. Moves on once we make the first touch.

On the card

Company name
Unit count + size tier
City, state
The one-line, company-specific reason to call — written for the phone, not pulled from the scoring engine
Warm-intro flag, only if one exists
Queue rank

On the full record

Snapshot — units (marked verified or estimated), software, HQ, who runs the buildings
The hook — same line as the card
Warm intro, or plainly "No known connection — cold outreach"
Current vendor, or "Not identified yet"
Small footnote: "Researched [date] · N sources"

Every one of the 489 companies gets the full record, even with nothing researched yet — empty sections simply don't render (Decision 1). Also new: companies that aren't real prospects at all (shut down, single building, no operating portfolio, absorbed into another firm) are removed from the board entirely — no ranked slot, no archive (Decision 4).

Card

ROM Investments

T1 · ≈440 units Warm: Sean knows Leeor

Los Angeles, CA

"No dedicated leasing director, no regional managers, no HR department, no IT role."

#1 in queue

Record

Snapshot

Owner-operator, self-managed · ≈440 units · Yardi · Los Angeles, CA

Hook

"No dedicated leasing director, no regional managers, no HR department, no IT role."

Warm intro

Sean knows Leeor Maciborski directly — one hop, no connector needed.

Contact

Leeor Maciborski — Partner

Incumbent

Yardi, installed — we sit on top, not compete. Maintenance side looks greenfield.

Researched Aug 14 · sources

Stage 2 · 5 companies

Contacted

We've reached out, no two-way reply yet. Moves on when they reply, or explicitly decline.

On the card

Company + units + tier
Contact name + role, if known
One-line outreach summary (how, when, attempt #)
Their reply so far, one line, or "No reply yet"
Next touch: date/channel, or "Overdue — nothing scheduled"
Warm-intro flag, if any

On the full record

Snapshot
Contact — name, role, phone/email
Outreach — one dense sentence covering channel/date/attempt/hook used
Their reply, word for word, if any
Warm intro, or "No warm path — direct outreach"
Next touch — date + channel, or the overdue flag

Stage 3 · 5 companies

In conversation

Real back-and-forth happening. Moves on once we know their pain, budget/timeline, and who decides — or they say no.

On the card

Company + units + tier
Decision-maker name + role
Their stated pain, one line, their words where possible
Next step + who owns it + date — the loudest line on the card
What's blocking progress, only if something is (skip the row otherwise)
Warm-intro source, if any

On the full record

Snapshot
People — decision-maker + up to 2 others, name + role
Where this stands — one factual sentence
Next step — the loudest block on the page
Blocker — what's stalling it and whose move it is, its own line, not buried in the next-step text
Current vendor — name + status, or "Not confirmed yet"
Warm intro — who introduced us, or "No warm path — direct outreach"
What they specifically want, in their own words, when we know it
Timeline — last 3–5 dated events, quotes allowed, never cut off mid-sentence

This is the fix for our KAGE deal specifically: item 8 is where the contact's real, specific requests (eviction-filing alerts, unanswered-tenant alerts, broken-payment-plan alerts, manager reporting) belong — not buried three clicks deep in old notes.

Stage 4 · 1 company

Proposal

Price and scope presented, they're formally deciding. Moves on when signed, or lost.

On the card

Company + units + tier
Economic buyer name + role
Terms, one line (price/scope)
Decision date
What's blocking it — the loudest line
Named competitor, or "no other vendor in the mix," if relevant

On the full record

Snapshot
Deal terms — price, scope, proposed start date, as a real named field (not buried in notes — this is where the Ulysses deal's actual $15,000/month terms sit today, three clicks deep)
Decision — economic buyer, their criteria, remaining steps, target date
Competition, or "No competitor named — evaluating us alone"
Internal champion, or "No internal champion identified yet"
Blocker, or "Nothing blocking — awaiting their decision by [date]"
Timeline

Stage 5 · 2 companies

Pilot

Running a proof-of-concept. Moves on when the agreed metric is met and they sign, or the pilot fails.

On the card

Company + units + tier
Champion name + role
Go/no-go metric + current reading — the loudest line
Decision date
Risk flag, only if a real risk exists
Days remaining

On the full record

Snapshot
Pilot scope — the one workflow, the one measured outcome, timeline
Status — one clean factual sentence, never a pasted internal note (CONAM's status field literally reads "Fede, 12 Aug 2026: CONAM is in the closing stage..." today — that's a Slack note, not a status)
Champion — name, role, how much pull they have internally
If it works — pricing/procurement path + decision date, or "Not yet agreed — confirm before pilot ends"
Risk — what's flagged and who owns fixing it, or "No risk flagged — on track"
Timeline

Stage 6 · 0 today

Signed

Contract signed, handing off to onboarding. Sales' part ends once the handoff happens.

On the card

Company + units + tier
Signer name + role
Start/kickoff date
What closed the deal, one line
Onboarding owner
Contract value/term, if tracked

On the full record

Snapshot
Contract — signer, start date, term, any special commitments made during the sale
Why we won — one reusable line for future deals
Stakeholders — signer, champion, day-to-day admin
What we promised they'd see move
Handoff — owner, kickoff date, status, or "Not yet handed off — assign an onboarding owner"

Stage 7 · 1 company

Passed

Won't close as-is. Stays here unless something changes.

On the card

Company + units + tier
Why we lost, plain English (never an internal code)
Stage lost at
Named competitor, if applicable
What would bring it back, if known
Date passed

On the full record

Snapshot
Why it didn't close — reason + one sentence, quote if we have one, no internal classification codes
Stage lost at
Competitor, if lost to one
What would bring it back — trigger/date to revisit, or "No trigger identified — don't re-approach without a new signal" (this field doesn't exist anywhere today — it's the single most-requested addition for this stage)

What we found, stage by stage

Not-contacted (489 companies). The biggest stage has the least-finished setup. Cards show the scoring engine's own justification — the same four-line template, word-for-word in 3 of 4 clauses, on 8+ cards we checked — instead of a real reason to call. The genuinely good line for ROM sits unused. Every record repeats the top summary and folded detail. This stage also has the most data mistakes: unit totals that contradict each other on the same page (30+ companies), sentences cut off mid-word, real ID numbers destroyed by an overzealous privacy filter, and dozens of companies our own research says aren't real prospects — still sitting in the ranked call queue with a full contact card.

Contacted (5 companies). The thinnest stage. Our one company with active outreach (Feiner Enterprises) shows less information than several not-yet-contacted companies — no outreach log, no next-touch date, nothing. Botnick Realty is close to the target voice, but one line contradicts another line right next to it.

In-conversation (5 companies). Structurally the closest to right — People → Next step → Where it stands → What they said → Timeline is the correct shape. But KAGE's record still opens with the exact rambling sentence we called out as the bad example, just one click deeper. Their real, specific requests are buried under a generic line. Nimes Real Estate — an active, interested conversation — wears a "passed on this company" badge left over from a stale automatic re-score that was never corrected after a human moved the deal forward.

Proposal (1 company). Ulysses Development Group is the clearest example of the whole problem: the real numbers ($15,000/month, a $30,000 two-month pilot, $1/unit/month after that) exist only inside old research notes, three clicks deep. The visible next-step line just says "wait for their circle-back on scope" — the number a founder most needs to see is invisible.

Pilot (2 companies). CONAM Group — our most advanced live deal — carries a "passed on this company" badge because a size-fit rule fires regardless of stage. Its status line is a pasted internal note rather than a real sentence. No champion, risk, or blocker field exists even though we have all three (a named champion, an escalation path, an IT-review blocker).

Signed (0 today). Nothing populated locally. One real closed deal (Jordon Perlmutter & Co.) appears to be missing from our data or filed under the wrong stage — flagged for a live-site recheck, not necessarily a bug in the page itself.

Passed (1 company). UNIMAT/Cannon's rejection reason shows a raw internal code ("category: gatekeeper-appetite") and has no "what would bring it back" field, even though the timeline implies one exists (a changed stance from their Arizona property manager).

Data-accuracy sweep — 153 confirmed mistakes, out of 193 found across all 503 companies

Counts below cover the 153 individual findings detailed in this pass; the full overnight sweep logged 193 across all 503 companies (2 originally-flagged findings were disproven on re-check and removed — see note below).

Kind of mistake	Count	What it means
Two numbers disagree on the same screen	~41	Usually: the "verified" header total vs. a building table that quietly includes sold, pipeline, or unconfirmed properties
Wrong status label	~23	A company our own research says to rule out (shut down, wrong size, no real portfolio) still shows as an active, callable lead with a full contact card
Broken sentences	~23	Text cut off mid-word, doubled state names ("Glendale, CA, CA"), spliced-together fragments — caused by the page-building code itself
Missing info that already exists	~12	A fact the stage actually needs (current vendor, next-touch date, warm intro) exists in our data but has nowhere to render
Old research text leaking through	~12	Rambling research notes and open-question lists still reach the page, just folded one click deeper
Not actually a company	~9	Shut-down entities, single buildings, or absorbed companies taking up a real queue slot
Out-of-date vs. the real timeline	~9	A record's "current" status disagrees with a more recent, dated fact elsewhere (a call that already happened, a stage we've already moved past)
Same fact shown twice	~9	Near-universal pattern: the top summary chips repeat in the "everything else" facts grid below
Real ID numbers destroyed	~10	A real license, parcel, EIN, filing, or ID number replaced with the nonsense text "a phone number" by an overzealous privacy filter
Wrong person shown	~4	A different company's executive shows up as if they belong to the prospect being viewed
Other	1	Same fact labeled two different ways on two parts of the same page

Two findings that were flagged overnight as "different company's info bleeding into this record" turned out, on closer check, not to be bugs: Jordon Perlmutter & Co.'s writeup mentions CONAM because the two deals are genuinely connected (a Perlmutter building is managed by CONAM, so CONAM's IT approval really is what's holding up Perlmutter's deal too) — not a data mixup. And UNIMAT's larger portfolio number was already correctly excluded as sold. Both are cleared and dropped from the count above; the other "wrong person shown" cases in the table (Century West showing a different company's president, Lennar, Hamilton Zanze) are real contact-picking bugs, still open, still being fixed.

Ten worst examples, word for word

Company	What's wrong
KAGE / Urban Phenix	Record still opens its research notes with "Where IT came from and where IT stands — groomed 14 Aug 2026 against the email record..." — the exact sentence we flagged as the bad example, word for word, still live.
CONAM Group	The status line for our one active pilot literally reads "Fede, 12 Aug 2026: CONAM is in the closing stage — pilot, going through their IT approval," a pasted internal note shown as the customer-facing status.
Ares Management	Software field reads "...ares does not run apartment buildings on Yardi + Entrata + unknown" — a leftover research fragment spliced into what should be a plain software list.
Brookfield	A real New York license number, Florida filing number, and EIN are all replaced with the literal words "a phone number," destroying three sourced, verified facts in one sentence.
Century West Partners	Card lists "Kevin Farrell — President" with no qualifier; he's actually president of a different company (Fifield Realty), only a named partner here.
ROM Investments	Header says a verified 440 units; the record's own building table on the same page sums to 604.
Mint Urban Infinity	A single 561-unit building renders as its own company, showing a national operator's corporate executives as its contacts.
Headwaters Group	Four items in the "open questions" list are literally labeled "not known" even though each one's own text begins "Resolved 14 Aug 2026" and states the answer.
Decron Properties	Header reads "Owner-operator... third-party managers," directly contradicting the record's own "Who runs them: They do / Self-managed" two lines below — caused by a text match that fired on a sentence explicitly denying third-party management.
Alpine Investments	Header states 349 units; the record's own building list sums to 796, including a building the data file explicitly flags as "NOT ALPINE'S TO SELL INTO."

Mechanical fixes — shipped overnight

All 13 of these are live now. This was the to-do list going into the night; it's a changelog as of this morning. The only work still open on this page is the 7 decisions below and the per-stage redesign itself.

Old research notes, "still unknown" lists, and open questions no longer render anywhere on the page. That content stays in the data file for the call-prep prompt and internal use only.
The duplicate facts grid is gone — the top summary chips are now the one place units/software/HQ/contact live; the "how we know" note moved to a hover tooltip on the existing chip.
Sentences and quotes no longer cut off mid-word — each one either fits, or was rewritten shorter before it renders.
The privacy filter no longer destroys real ID numbers. It used to match any long digit string and replace it with "a phone number" — it now checks for an actual phone-number shape and allows known ID formats through.
Building tables now count only currently-owned buildings — sold, pipeline, and unbuilt properties no longer get added into a total that contradicts the verified header number. This one fix alone resolved roughly 30 of the confirmed mistakes.
The "third-party manager" label now reads a real, verified field instead of scanning research text for the phrase — it no longer gets fooled by a denial ("Essex does not have a third-party manager").
Contact matching now always prefers the actual, confirmed decision-maker over an old free-text label or loose name match.
"Passed on this company" and "wrong size for us" now carry separate labels — "passed" can no longer appear on any company that's been contacted.
Internal codes and pasted notes are stripped from everything customer-facing — tags like "(category: gatekeeper-appetite)" and Slack-style prefixes like "Fede, 12 Aug 2026:" no longer show up in a status line.
The "not researched yet" empty-state message no longer fires next to a field that's actually filled in.
The hardcoded "Multifamily owner" label no longer shows up on companies our own research says aren't multifamily operators.
The hard length limit on the "next step" line (the one meant to be loudest per stage) was widened so it's no longer the thing getting cut off, while the less important scoring-engine explanation runs unclamped.
The doubled-state-name bug ("Glendale, CA, CA") and false sentence-splits on things like "St." and "e.g." are fixed.

7 decisions — approved by Fede, Aug 20, 2026

Fede's actual picks below, marked "chosen by Fede." Decisions 2, 3, 5, 7 match the recommended option; decisions 1 and 4 went the other way.
The per-stage redesign that puts this spec and these decisions into the live product is now in build.
The live site was re-checked Aug 20: every overnight fix above was confirmed working on sales.propflowai.co, zero regressions found.

1. How much detail should an unresearched Stage-1 company get?

A. Build the full 5-item record for all 489 companies, regardless of research depth — chosen by Fede ✓Every company gets a full record page, even unresearched ones. Empty sections just don't render.

B. Card + call-prep prompt only, until at least one real fact exists

C. Something else

2. What happens to the "copy a ChatGPT prompt" button?

A. Keep it as-is alongside the new fields

B. Retire it once the fields above ship and are trustworthy — chosen by Fede ✓It exists today because the on-page fields aren't good enough yet. Once they are, it's redundant.

C. Keep it, but only for genuinely unstructured prep material the fields don't cover

3. Should the board split "Proposal" and "Pilot" into two columns?

A. Split into two separate columns, as originally planned

B. Keep them combined as one "Closing" column at today's deal volume — chosen by Fede ✓We have 3 deals total right now. Split once volume grows enough to justify it.

C. Keep them combined permanently, regardless of volume

4. Where do confirmed non-prospects go (shut-down companies, single buildings, pure investment firms)?

A. Delete them outright — chosen by Fede ✓Removed entirely, no archive view.

B. Move them to a separate "not a fit" list, off the ranked queue

C. Leave them in the ranked queue with a visual flag

5. Should the "researched on / N sources" line be its own section, or stay a small footnote?

A. A labeled "Sources" section, as originally planned

B. Keep it as the small footnote already shipping ("Researched [date] · N sources") — chosen by Fede ✓It already works and matches the density we're going for.

C. Drop it from Stage 1 entirely

6. What happens to the building-by-building portfolio table?

A. Delete it entirely

B. Keep it as a separate, click-to-open "portfolio detail" view — confirmed by Fede ✓Filtered to currently-owned properties only, and reconciled to the header total. Real, useful detail — kept clearly separate from the banned leftover research fold.

C. Fold it directly into the main record

7. When our live board's stage disagrees with our researched stage (e.g. Botnick), which one wins?

A. Always trust the live database's stage field

B. A company only counts as "in conversation" once a real two-way exchange is actually logged — chosen by Fede ✓This overrides any manually-set stage field, and matches the stage definitions in this spec.

C. Flag the conflict visually and make a human resolve it

What must not change

The units-confidence mark (checkmark = counted, ≈ = estimated), with a hover tooltip explaining it — this is the model every other field above should copy.
The status label as a plain-text word, not a color-coded key you have to decode — just fix the "passed" collision (decision 7 / fix 8), don't change the pattern itself.
The People → Next step → Where it stands → What they said → Timeline order — verified consistent across every stage. This is the right skeleton; the work above is about filling in its content, not rebuilding its shape.
How the Timeline block handles quotes and empty states — dated, attributed, capped, with a real difference between "(nothing yet)" (we reached out, no reply) and "Nothing has happened yet." (never contacted).
Pulling real research sentences into the "what they said" opener instead of inventing new copy.
The honest, empty stub record for a company we've added by hand but haven't researched yet — same template, nothing special-cased. Good instinct; this becomes the formal Stage-1 rule under decision 1.
The top summary bar's density — name, tier, units, software, city, stage in one glanceable line. Leave it exactly as-is.

Full finding list — 153 rows (click to expand)

Company	Kind	What's wrong
amcal	garbled-template	Next-step line cut off mid-word with a literal ellipsis in the HTML
abacus-capital-group-llc	garbled-template	Bolded Next line cut off mid-word ("th…")
4site-real-estate	garbled-template	Units sub-line cut off mid-sentence
alpine-investments	card-data-contradiction	Header 349 units vs. spread total 796; pipeline/wrong-owner buildings included
alamo-manhattan	card-data-contradiction	Header 1,977 units vs. spread total 5,179; sold communities still listed
aimco	card-data-contradiction	Header 2,524 units vs. spread total 5,173
annenberg-investments	card-data-contradiction	Header 1,051 vs. spread total 1,547
amplify-development-company	card-data-contradiction	Header 1,083 vs. spread total 1,525
anderson-pacific-llc	card-data-contradiction	Header 315 vs. spread total 538
alamo-manhattan	redaction-artifact	"(Research a phone number)" nonsense placeholder mid-sentence
abacus-capital-group-llc	rendered-slop	Visible Next line is hedged research narration, not an instruction
amcal	rendered-slop	Visible, always-shown Next line is a long conditional research sentence
4site-real-estate	duplicate-info	Units/software/HQ shown in header chips and again in fold facts grid
all 24 not-contacted (systemic)	duplicate-info	"Everything else" fold repeats header facts on every record
Baron Properties	card-data-contradiction	"3,124 across 17 communities" vs. table of 19 buildings summing to 3,497
Archway Equities	card-data-contradiction	5,000 units stated vs. 8,581 in the building table
Arden Group	card-data-contradiction	612 units stated vs. 1,633 in table, incl. 3 unconfirmed properties
BDT & MSD Partners	wrong-classification	Record concludes "not multifamily" but renders "Multifamily owner"
Associated Estates Realty	not-a-company	Company doesn't exist (slug says "defunct"), still ranked as active
Ares Management	garbled-template	Research sentence fragment spliced into software-list field
botnick-realty	stale-vs-timeline	D1/JSON/conversations-log disagree on stage; board shows In-conversation, record shows Contacted/no-reply
brookfield	redaction-artifact	License/filing/EIN numbers replaced with "a phone number"
botnick-realty	card-data-contradiction	"Who runs them: They do" directly contradicted by "Third-party managers" sub-line
bolour-associates	card-data-contradiction	113 units stated vs. 1,508 in spread, incl. unbuilt/sold properties
big-4-properties	garbled-template	Open-questions list drops subject, leaves dangling fragments
big-4-properties	rendered-slop	Full open-questions/still-unknown prose still renders, folded
braddock-logan-services-inc	garbled-template	Next line cut off mid-word ("rather")
brick-lofts (single property)	not-a-company	A 21-unit building, not a company, carries a full company record
botnick-realty	stale-vs-timeline	D1 unit count stale by 417 vs. newer researched figure
caruso	card-data-contradiction	Verified 329 units vs. table total 628, incl. excluded/pipeline buildings
cherng-family-trust	card-data-contradiction	304 units vs. 536, unbuilt project included
citypads	card-data-contradiction	400 units vs. 986, sold/pipeline properties included
city-wide-apartment-management-inc	card-data-contradiction	142 verified vs. 152 in table; garbled open-question fragment
century-west-partners	wrong-company-bleed	Different company's exec (Fifield's president) shown as this company's contact
century-west-partners	wrong-classification	Record recommends closing the account; still renders as active with full contact card
canyon-partners-llc	wrong-classification	"Ruled out," record explicitly says no operations team exists; still full contact card + call prompt
capri-capital-partners-llc	wrong-classification	Record says "none recommended"; still full contact card + call prompt
cardinal-group-companies	wrong-classification	Record says none recommended (57k units, has a competing product); still full contact card
cim-group-lp	wrong-classification	Record explicitly says not to prioritize; still full contact card
cirrus-asset-management-inc	wrong-classification	Record says "not recommending outreach"; still full contact card
cityview	wrong-classification	"Ruled out" pill, still renders as fully actionable
clarett-west-development	wrong-classification	Record says "no fit"; still shows contact + units table + prompt
catalina-development-co	wrong-classification	"Recommend against pursuit"; still full contact card
catamount-properties-ltd	wrong-classification	"Do not prioritize"; still full contact card
chandler-partners	garbled-template	State doubled in location text ("Glendale, CA, CA")
carmel-partners-inc	garbled-template	Self-contradicting sentence ("Yardi... NOT Yardi"); doubled state
conam-group	wrong-classification	"Ruled out" pill on an active pilot deal
cypress-equity	card-data-contradiction	Shows PMS "Yardi" but warns re: RealPage integration — vendor never in their stack
cypress-equity	card-data-contradiction	Table headed "the buildings we have" totals 9,226 vs. verified headline 2,602
conam-group	redaction-artifact	Parcel schedule number replaced with "a phone number"
debeikes-investment-co	redaction-artifact	CA filing number replaced with "a phone number"
cypress-equity	missing-key-info	Contacted stage has no channel/date/attempt/next-touch anywhere
conam-group	missing-key-info	Pilot stage has no Champion/blocker/risk field despite data supporting all three
davis-development	stale-vs-timeline	Record's cited ranking data doesn't match the current ranking file
collegeplace-partners-llc	duplicate-info	Header/fold duplication, representative example
decron-properties	card-data-contradiction	"Third-party managers" render despite record itself saying self-managed
decron-properties	garbled-template	Resolved research notes render as garbled "open questions"
elevation-group	garbled-template	Same resolved-note-as-open-question bug, garbled fragments
emerson-lofts (single property)	wrong-company-bleed	Third-party manager's executives shown as this record's own Players
emerson-lofts (single property)	not-a-company	42-unit building under absentee owner still in ranked queue
essex-property-trust-inc	wrong-classification	Self-managed company rendered as "third-party managers"
g-h-palmer-associates	wrong-classification	Same bug — "third-party manager" phrase in a denial flips the label
evercore	wrong-classification	Investment bank rendered "Multifamily owner in New York, NY"
forest-city-realty-trust-inc	not-a-company	Defunct since 2018, still occupies a ranked-pool row
fpi-management	not-a-company	Absorbed into another company in 2025, still occupies a row
garden-communities-california	redaction-artifact	SCH clearinghouse IDs replaced with "a phone number"
garden-communities-california	card-data-contradiction	3,111-unit floor vs. 5,204 in table, unbuilt properties included
fifield-companies	card-data-contradiction	1,956 verified vs. 2,133 in spread, sold property re-included
equity-residential	rendered-slop	Full research-notes/still-unknown/open-questions block still on page
essex-property-trust-inc	rendered-slop	Same — confirms the pattern is systemic, not one company
foundation-group	rendered-slop	The one In-conversation record still ships an open-questions list
foundation-group	missing-key-info	No Incumbent Vendor or Warm Path slot anywhere on the record
feiner-enterprises	missing-key-info	Contacted-stage record has no outreach log or next-touch date at all
four-peaks-multifamily-partners	missing-key-info	"Who runs them" shows "not researched" though the fact is known and stated
gem-realty-capital	garbled-template	Research fragment "not applicable in the normal sense" dumped into description
gem-realty-capital	wrong-classification	Named exec shown as "Contact" though record says no real point of entry exists
generation-real-estate-partners	card-data-contradiction	98 units vs. 250 in table, incl. a sold building
glencrest-group	card-data-contradiction	2,786 verified vs. 3,146, sold properties silently merged into city total
gpi-companies	card-data-contradiction	1,183 verified vs. 1,318 in fold
gwp-real-estate	card-data-contradiction	172 units vs. 331 in fold
gelt-ventures	card-data-contradiction	"Who runs them" value populated, sub-line says "Not researched yet" anyway
goldrich-kest-gk-management-co	garbled-template	Run-on sentence from spliced fragment; "They do" misstates a mixed-management company
halco-management-inc	card-data-contradiction	Description casts doubt on a fact the Software chip states as settled
griffis-residential	garbled-template	Awkward spliced run-on sentence in the description
hamilton-zanze	wrong-classification	Wrong parent company's CEO/city shown as contact; real operator is a different affiliate
multiple (gelt-ventures etc., systemic)	rendered-slop	Confirms open-questions/research-notes folds present across the batch
guggenheim-real-estate-llc	not-a-company	Retired brand (rebranded 2023) still one of the 24 active profiles
gid	garbled-template	Stray mid-sentence capitalization artifact from template concatenation
headwaters-group	garbled-template	4 resolved facts mislabeled "not known"
hines	garbled-template	Same "resolved shown as unknown" bug + capitalization break
ico-group	wrong-company-bleed	A building explicitly flagged "not theirs" re-added into their unit total and table
ico-group	redaction-artifact	Filing registration number replaced with "a phone number"; orphaned "Source: and" fragment
inland	card-data-contradiction	"Who runs them: They do" / sub-line "Third-party managers" contradict
hsr-income-value-reit	duplicate-info	Same fund family renders as a separate, duplicate card
holland-residential	duplicate-info	Identical contact renders on two separate company cards
heitman-llc	missing-key-info	Known-but-unverified contact suppressed entirely instead of shown caveated
hanover-financial-llc	other	Same status flag labeled two different words on two surfaces
harbert/harrison/horizon/imt (batch)	garbled-template	Multiple broken open-question fragments across companies
iintoo-investments-ltd	wrong-classification	Non-operator platform occupies a normal ranked slot
jade-enterprises	redaction-artifact	Parcel/AIN numbers replaced with "a phone number"
kage-urban-phenix	rendered-slop	Canonical negative-example opening line still renders verbatim, one fold deeper
kage-urban-phenix	missing-key-info	Incumbent vendor (LiaPro) present in data, never rendered
kage-urban-phenix	missing-key-info	Warm path (Kyle Perlmutter) never rendered as its own field
jp-co	rendered-slop	Research notes/open questions render 3 times on one record
koelbel	card-data-contradiction	Populated "Who runs them" value paired with "Not researched yet" caption
integral-communities	duplicate-info	Contact repeated between Players block and fold facts
jade-enterprises	rendered-slop	Full open-questions list renders in the fold
lyon-living	missing-key-info	Competing AI product (EliseAI) already installed — never surfaced as Incumbent Vendor
lion-real-estate-group	card-data-contradiction	7,138 (complete, per record) vs. 7,373 in table, incl. excluded property
kor-group	garbled-template	Duplicated phrase "Sean's Sean's own list" (text-substitution bug)
landmark-properties-inc	garbled-template	Open-question fragments truncate mid-name/mid-word across the batch
kor-group	duplicate-info	Header/fold duplication, representative example
legend-real-estate-management (single property)	not-a-company	Single 45-unit building, unconfirmed owner, occupies a company row
lennar-corporation	wrong-company-bleed	A related-but-different company's CEO shown as Lennar's own
mint-urban-infinity (single property)	not-a-company	Single building shows a national operator's corporate execs as Players
mint-urban-infinity	card-data-contradiction	"Who runs them" self-contradicts; city chip shows wrong city
magellan-investment-partners	card-data-contradiction	Estimate mislabeled as "their own figure" (claimed)
mgl-partners-llc	card-data-contradiction	Same estimate-vs-claimed mislabel, reclassified-on-audit case
millstein	card-data-contradiction	2,481 verified vs. 4,526 in table, no reconciling note
millstein	garbled-template	"Manhattan, Manhattan, Manhattan" — submarket detail truncated and repeated
all 24 (systemic)	rendered-slop	Confirms fold pattern present on every record, magellan/mg-properties/mint/millstein cited
nimes-real-estate	card-data-contradiction	"Ruled out" pill on a record showing an active, interested conversation
nimes-real-estate	redaction-artifact	UUID mangled mid-string by the phone-number scrubber
nimes-real-estate	missing-key-info	"No next step on record yet" despite an active, interested conversation
nimes-real-estate	stale-vs-timeline	Known contact event never surfaces in the Timeline
mountain-capital-group-inc	wrong-classification	Record recommends confirming the company still exists; renders as a normal live prospect
new-york-life-real-estate-investors	garbled-template	Open questions cut off mid-sentence, unclosed parentheses
morgan-group	stale-vs-timeline	Stale field internally contradicts the rest of its own file (doesn't currently break the page)
palisade-partners	card-data-contradiction	185 units (2 buildings) vs. 1,281 in table, incl. sold buildings and a hotel
oak-residential-partners-llc	card-data-contradiction	3,473 units vs. 3,713, pipeline property re-included
oaktree-capital-management	garbled-template	Sentence splitter breaks on "e.g." producing two broken bullets
oak-coast-properties	garbled-template	Sentence splitter breaks on "St." dropping the bullet's lead-in
palomino-park	not-a-company	Property name, not a company, still sits in ranked queue
northwood-investors-llc	duplicate-info	Header/fold duplication, representative example
resource-real-estate-resource-reit-inc	stale-vs-timeline	Dead since 2022, no exclusion flag, renders as a live row
redwood-capital-group-llc	stale-vs-timeline	Absorbed in 2022; named contact is adversarial (active lawsuit), no flag
polaris-property-management-llc	stale-vs-timeline	Company shut down; status field unset, renders as live
petrucci-residential	card-data-contradiction	Claimed total silently excludes a listed, named property
philena-properties	wrong-classification	Record says "not a fit"; row not flagged, unlike comparable disqualified rows
pensam-capital	wrong-classification	Capital allocator not flagged despite record's own conclusion
principal-real-estate-investors-llc	wrong-classification	Investment manager not flagged, inconsistent with similar disqualified rows
quarterra-multifamily	wrong-classification	Developer-investor not flagged; CEO shown as a viable lead
situs-group	stale-vs-timeline	Next line asks to "send an invite" for a call that already happened 3 days earlier
situs-group	stale-vs-timeline	"Verified" 414 units contradicted by the decision-maker's own stated ~1,000 on a later call
rom-investments	garbled-template	Date dropped from timeline sentence, leaving a broken clause
rom-investments	redaction-artifact	Archive.org timestamp replaced with "a phone number" mid-URL
rom-investments	missing-key-info	The gold-standard hook line never renders anywhere on the card or record
rockrose-development-corporation	missing-key-info	"Most important competitive fact on the record" (Portico installed) buried in a fold fragment
seagate-properties	card-data-contradiction	1,506 verified vs. 2,866 in spread, 5 sold properties included
sharp-capital	card-data-contradiction	153 units (record's own thesis) vs. 1,394 in spread — 9x inflation
rivendell-global	card-data-contradiction	Data file explicitly refuses a portfolio total; spread widget sums one anyway
rudin-management-co-inc	card-data-contradiction	2,414 estimate vs. 2,985 spread total
rpm-company	garbled-template	Next line cut off mid-parenthetical, unclosed
situs-group	rendered-slop	Open-questions list present, confirming systemic pattern
rom-investments	duplicate-info	Header/fold duplication, representative example