DealRadar — System Architecture & Business Proposal
FYXIE AI · CONFIDENTIAL
ORLANDO, FLORIDA
Florida Real Estate Intelligence Engine

Autonomous Lead
Acquisition for
Wholesale Real Estate

DealRadar is a fully autonomous pipeline that scrapes distressed properties across all 67 Florida counties, scores them with AI, traces owners, and deploys conversational SMS outreach — with zero manual prospecting.

1,464FL Zip Codes
67Counties
$10KAvg Deal Fee
46×ROI on Op Cost
$340Starter/mo

What DealRadar Does

A production-grade intelligence and outreach engine replacing the entire manual prospecting workflow for Florida real estate wholesalers.

Problem

Manual Prospecting is Broken

Most wholesalers spend 80% of their time cold-calling lists from spreadsheets. No scoring, no automation, no AI. Deal velocity is limited by human bandwidth.

Solution

Fully Autonomous Pipeline

DealRadar automatically scrapes, scores, skip-traces, and outreaches to distressed property owners 24/7 across all of Florida — no manual intervention.

Outcome

26–46× Return on Op Cost

Each closed deal generates $10K in assignment fees. Operating cost per closed deal: $220–$382. One deal/month at Starter tier covers 30× operating cost.

Market Scope

  • Florida ONLY to start — prove ROI, then scale nationally
  • 1,464 zip codes · 67 counties · 158M+ property records via ATTOM
  • Target: distressed owners with equity ready to negotiate
  • Distress signals: pre-foreclosure, tax liens, code violations, high DOM, price reductions

Core Tech Stack

  • n8n Cloud — Workflow orchestration & scheduling
  • Supabase PostgreSQL + PostGIS — Core data layer (4 schemas)
  • Apify + TinyFish + ATTOM — Multi-source data ingestion
  • Claude 3.5 Sonnet — Lead scoring + conversational SMS AI
  • GoHighLevel + Twilio — CRM & outreach delivery

Wholesalers Are Flying Blind

The wholesale real estate industry is one of the last high-revenue verticals still running on manual cold-calling and spreadsheets.

80%

of a wholesaler's time spent on manual prospecting activities

2–5%

typical contact rate on cold lists without skip trace + scoring

$0

invested in AI or automation by the average independent wholesaler

3–7

days between a lis pendens filing and a competitor making first contact

The Core Problem Loop

Step 1Find List
Step 2Manual Filter
Step 3Skip Trace
Step 4Cold Call
Step 5Chase Responses
Result~1 Deal/mo

Each step is done manually, in a different tool, with no shared data layer. Lead quality is unknown until hours into a call sequence.

Full Stack Architecture

Eight horizontal layers from raw data sources to business intelligence, fully orchestrated by n8n with a Supabase data spine.

← Scroll to explore full diagram →
LAYER 1 · DATA SOURCES ZillowListings + Zestimates RedfinListings + DOM PropwireOff-market data 67 FL County ClerksLis Pendens · Foreclosures ATTOM Data API158M+ properties Municipal PortalsCode violations (future) Probate / DivorceCourt filings (future) LAYER 2 · INGESTION Apify ActorsZillow · Redfin · Propwire scrapers — $0.20–0.30/CU TinyFish AgentsStealth browser · SSE streaming · $0.012/step ATTOM REST APIPre-foreclosure · Liens · Auctions n8n Webhook TriggersInbound GHL replies · API callbacks LAYER 3 · ORCHESTRATION — n8n CLOUD FL Scraper Launcher7IlBFNm8ERe5mcvR · ACTIVE Lis Pendens ScraperDkS5LlNaiP3A75Qs · ACTIVE Lead Creator692HEkm5hncMzyY6 · INACTIVE Skip Trace SubmitRjwzFTInKSiMmO9p · INACTIVE GHL OutreachGJa5c2amnxxRzKMa · INACTIVE LAYER 4 · STORAGE — SUPABASE POSTGRESQL + POSTGIS public.*properties · geo_locationbulk_upsert_properties() RPC scraping.*lis_pendens_raw · apify_runsraw_payloads · error_logs crm.*leads · skip_tracesoutreach_logs · conversations analysis.*market_stats · florida_hotnessrefresh_florida_hotness() RPC LAYER 5 · AI INTELLIGENCE Claude 3.5 Sonnet (Primary)Lead scoring · Motivation analysis · SMS generation DeepSeek V3 / GPT-4o-miniOpenRouter · Secondary classification Hotness Score FormulaAbsorption×0.4 + Δprice×0.2+ DOM×0.2 + Distress×0.2 Lead Score Thresholds≥85 → Auto skip trace60–84 → Review queue LAYER 6 · SKIP TRACE WATERFALL Tracerfy (Primary)$0.04/record · Address → Phone + Email NO HIT Skipify.ai (Fallback)$0.085–0.10/record · Tracerfy misses LLC DETECTED BatchData (LLC Pierce)$0.07/record · 76% right-party rate HIT crm.leads UPDATEDphone_numbers + emails + dnc_clean LAYER 7 · CRM & OUTREACH GoHighLevel CRMSMS · Voice · Email · Calendar Twilio SMSDelivery layer via GHL Claude AI (Conv. SMS)Context-aware reply generation RedisConv. thread memory Appt. Booked→ Under Contract LAYER 8 · VISUALIZATION & BI DealRadar UI (React 19 + Vite)Heatmap · Leads · Pipeline · Analytics Looker StudioBusiness intelligence dashboards Florida Heatmap (SVG)Zip code hotness overlay · Click-to-pull Supabase Realtime + PostGISLive feed · Geo queries · Materialized views

End-to-End Data Flow

Every property travels through a deterministic pipeline: ingestion → dedup → scoring → qualification → skip trace → outreach → deal.

🏠 Scrape 100K+ properties/mo (Max)
🔁 Dedup Unique by address_normalized
🤖 AI Score 0–100 Claude 3.5 Sonnet
📊 Qualify Score ≥85 Auto-advance queue
🔍 Skip Trace Phone + Email 3-provider waterfall
💬 SMS Outreach AI-Driven Claude conversation
📅 Appointment Booked Calendar via GHL
💰 Closed Deal $10K Assignment fee

Distress Signal Sources

Lis Pendens (Pre-Foreclosure)
TinyFish county clerk scraping · Highest priority · Active foreclosure filing
Tax Delinquency
ATTOM Data API · Unpaid property taxes signal distress
Code Violations
ATTOM + Municipal portals · Property in disrepair
Auction Notices
ATTOM Data API · Imminent forced sale
Price Reductions
Apify (Zillow/Redfin) · Active seller motivation signal
High Days on Market (>90)
Apify · Stale listing, seller fatigue likely
Deed Transfers / Quit Claims
ATTOM · Ownership changes, possible motivated seller
Probate / Divorce (Future)
TinyFish county clerk expansion · Inherited / forced sales

Stored Procedures

bulk_upsert_properties(JSONB)public
get_unscored_properties(limit)public
get_leads_needing_skip_trace(limit)crm
refresh_florida_hotness()analysis
+ Public RPC wrappers for all non-public schema functions

TinyFish Scraper Notes

Uses stealth browser profiles + US residential proxies — required because county clerk sites block datacenter IPs. 7 pilot counties active (Hillsborough, Orange, Miami-Dade, Broward, Palm Beach, Duval, Lee). Scalable to all 67 FL counties via SSE streaming API.

Scoring Architecture

Two-layer AI scoring: a market-level Hotness Score per zip code, and a Lead Score per property combining hotness with AI-generated motivation analysis.

40% Hotness Score
(zip code market)
×
60% Motivation Sentiment
(AI-analyzed, per owner)
=
0–100 AI Lead Score
≥85

Auto-Advance
Skip trace + GHL outreach immediately queued. No human review.

60–84

Review Queue
Standard queue for manual review before outreach decision.

<60

Archive
Stored in database for potential future reassessment. No outreach.

Zip Code Hotness Score (Materialized View)

40% Absorption Rate absorption_rate / 30
20% Price Change % price_change_pct / 15
20% Days on Market 1 − (median_dom / 120)
20% Distressed Leads distressed_count / 50
-- analysis.florida_hotness materialized view formula
hotness_score = (absorption_rate / 30 × 0.40) + (price_change_pct / 15 × 0.20) + ((1 − median_dom / 120) × 0.20) + (distressed_lead_count / 50 × 0.20)

-- Classifications:
strong_seller  |  moderate_seller  |  neutral  |  buyer  |  strong_buyer

3-Provider Waterfall

Maximum hit rate through automatic provider cascading. Starting with lowest cost, escalating to LLC-piercing when corporate ownership is detected.

1
Tracerfy
Primary — address → owner + phone + email
$0.04/record
▼ IF NO HIT
2
Skipify.ai
Fallback for Tracerfy misses
$0.085–0.10
▼ IF LLC DETECTED
3
BatchData (LLC Pierce)
Finds the person behind corporate ownership
$0.07/record

BatchData LLC Piercing

76% right-party contact rate when piercing corporate-owned LLCs. Finds the actual human decision-maker behind shell companies commonly used by investors and landlords.

DNC Compliance

Every traced number is checked against the Do Not Call registry before outreach. The dnc_clean and litigator_flagged fields on crm.leads gate all outreach attempts. Zero-tolerance compliance to prevent litigation exposure.

Cost Per Skip Trace (Blended)

Tracerfy hit (most common)$0.04
Skipify fallback$0.085–0.10
BatchData LLC pierce$0.07
Blended average (est.)~$0.05–0.06

crm.skip_traces Schema

lead_idFK→leads
providerTEXT
statusTEXT
phones_foundTEXT[]
emails_foundTEXT[]
cost_usdNUMERIC

Autonomous Conversational SMS

GHL pushes the initial contact. Owner replies are routed through Redis-backed Claude conversations, maintaining full context across multi-turn exchanges.

Step 1

Lead Push to GHL

Once skip-traced, leads are pushed to GoHighLevel via the GHL Outreach n8n workflow. GHL triggers the first SMS through Twilio using the DNC-verified number.

Step 2

AI-Driven Replies

Inbound SMS replies hit an n8n webhook, retrieve conversation context from Redis, and pass the full thread to Claude 3.5 Sonnet for context-aware response generation. Responses are human, never robotic.

Step 3

Appointment Booked

When an owner shows interest, Claude books an appointment directly into GHL calendar. crm.conversations.appointment_booked = true. Deal enters pipeline as "Under Contract."

Conversation State in Redis

Redis stores the full message history for each active conversation, keyed by ghl_contact_id. Claude receives the complete thread on each reply — no context lost, no repetition, no robotic "Hi [FIRSTNAME]" restarts. This is the single most important differentiator in reply quality vs. template-based SMS campaigns.

crm.conversations fields
messagesJSONB
turn_countINTEGER
sentimentTEXT
appointment_bookedBOOLEAN
appointment_atTIMESTAMPTZ
outreach pipeline stages
newInitial lead
skip_tracedOwner ID'd
in_ghlActive outreach
under_contract🎯 Closed

Supabase PostgreSQL — 4 Schemas

xjoxxxqhexlovrsfjuop · PostgreSQL + PostGIS · Dedup on address_normalized · PostGIS geo queries for heatmap rendering

public.properties

CORE TABLE
idPK · UUID
address_normalizedUNIQUE
geo_locationGEOGRAPHY(POINT)
listing_price / zestimateNUMERIC(12,2)
days_on_marketINTEGER
listing_statusFOR_SALE · PRICE_REDUCED
price_historyJSONB
zpid · redfin_id · propwire_idcross-platform IDs
source_platformTEXT

scraping.lis_pendens_raw

FORECLOSURE DATA
idPK · UUID
county · county_codeTEXT
case_numberUNIQUE per county
filing_dateDATE
plaintiffLender/bank filing
defendantOwner being foreclosed
parcel_numberTEXT
lead_statuspending|scored|sent|archived
tinyfish_run_idTEXT

crm.leads

CORE CRM
idPK · UUID
property_idFK→properties
ai_lead_scoreINTEGER (0–100)
motivation_scoreINTEGER (0–100)
motivation_summaryAI-generated TEXT
phone_numbersTEXT[]
dnc_clean · litigator_flaggedBOOLEAN
ghl_contact_idTEXT
outreach_statusnew|skip_traced|in_ghl|contract

analysis.market_stats

ZIP CODE INTEL
zip_code · countyTEXT
absorption_rateNUMERIC(5,2)
months_of_supplyNUMERIC(5,2)
median_domINTEGER
median_price_change_pctNUMERIC(5,2)
distressed_lead_countINTEGER
hotness_scoreNUMERIC(3,2)
market_classificationstrong_seller → strong_buyer
florida_hotness (view)MATERIALIZED

Automation Inventory

Five core workflows on n8n Cloud (fyxie.app.n8n.cloud) — two active, three staged for activation.

← Scroll for full workflow details →
Workflow ID Status Trigger Purpose
FL Scraper Launcher 7IlBFNm8ERe5mcvR Active Cron schedule Triggers Apify actors per zip batch, logs runs to scraping.apify_runs
TinyFish Lis Pendens Scraper DkS5LlNaiP3A75Qs Active Cron · 7 counties Scrapes county clerk lis pendens filings, upserts to scraping.lis_pendens_raw
Lis Pendens Lead Creator 692HEkm5hncMzyY6 Inactive Webhook / cron Converts raw lis pendens records → crm.leads with AI scoring
Skip Trace Submit RjwzFTInKSiMmO9p Inactive Webhook trigger Waterfall skip trace: Tracerfy → Skipify → BatchData for leads ≥85
GHL Outreach GJa5c2amnxxRzKMa Inactive Webhook trigger Pushes scored+skip-traced leads to GoHighLevel for SMS/voice campaigns

3-Tier Pricing Structure

Start lean and proven, scale on results. Each tier is profitable at just one deal per month.

Starter $340/mo
50–100 zip codes · 7 counties · 500 skip traces/mo
Supabase Pro$25
n8n Starter$24
Apify Starter$29
TinyFish$30
Tracerfy$20
Skipify$5
ATTOM Starter$95
GHL Starter$97
SMS Usage$15
Growth — Recommended $1,092/mo
300–500 zip codes · 25–30 counties · 2,500 skip traces/mo
Supabase Pro$25
n8n Starter$24
Apify Scale$199
TinyFish$30
Tracerfy + Skipify$25
BatchData$14
ATTOM Standard$200
GHL Unlimited$297
SMS + Misc$278
Max — Full Florida $3,820/mo
All 1,464 zip codes · 67 counties · 10,000 skip traces/mo
Supabase Pro$25
n8n Business$800
Apify Business$999
TinyFish (67 counties)$150
All Skip Trace Providers$250
ATTOM Premium$500
GHL Unlimited$297
PropStream Pro$199
SMS + Infra$600

Unit Economics & Deal Funnel

Growth tier projections. Each stage of the funnel narrows the universe to the highest-conviction leads only.

Growth Tier Funnel (Monthly)

Properties Scraped100%
25,000–50,000
AI Scored (≥60)~30%
~7,500
High Priority (≥85)~10%
~2,500
Skip Traced~80% hit
~2,000
Contacted via SMS~70%
~1,400
Responded / Interested~3–5%
~42–70
Appointments Booked~50%
~21–35
Deals Closed
3–5
$10K
Avg Assignment Fee
$364
Cost Per Closed Deal (Growth)

27×
Min ROI on Op Cost
46×
Max ROI on Op Cost

Revenue Projections by Tier

Starter (~1 deal/mo)$10K/mo
Growth (~3–5 deals/mo)$30–50K/mo
Max (~10–15 deals/mo)$100–150K/mo

Key Unit Economics

Starter cost per closed deal$340
Growth cost per closed deal$218–364
Max cost per closed deal$255–382
Avg margin per deal$9,618–9,782

DealRadar vs. The Market

Existing tools are data providers — they give you lists. DealRadar is an autonomous acquisition engine — it works the list for you.

← Scroll to see full comparison →
Feature DealRadar PropStream BatchLeads DealMachine InvestorLift
AI Lead Scoring ✓ Claude 3.5 Sonnet Partial Partial
Autonomous SMS Outreach ✓ Full AI conversations Basic templates
Lis Pendens Scraping ✓ 67 FL counties Some counties Limited
LLC Owner Piercing ✓ BatchData (76% rate)
Multi-Source Skip Trace ✓ 3-provider waterfall Single provider Single provider
Zip Code Hotness Scoring ✓ Custom PostGIS formula Basic market data Basic
Custom Data Pipeline ✓ Fully owned ✗ Platform lock-in ✗ Platform lock-in ✗ Platform lock-in ✗ Platform lock-in
Monthly Cost (Growth equiv.) $1,092/mo $499/mo (data only) $299/mo (data only) $599/mo Custom pricing
End-to-End Automation ✓ Zero manual prospecting ✗ Manual workflows ✗ Manual workflows ✗ Manual workflows Partial

Core Competitive Moat

DealRadar's advantage isn't any single feature — it's the vertical integration. Competitors sell data or tools. DealRadar automates the entire acquisition workflow from raw court record to booked appointment. The data is proprietary (not a licensed feed), the scoring is customized for FL distressed properties, and the AI outreach quality improves with every conversation in Redis context history.

FL → Southeast → National

A phased expansion strategy with clear milestones. Prove unit economics in Florida first, then leverage the same infrastructure for geographic expansion.

Phase 1 · Q1–Q2 2025 Florida
Activate all 5 n8n workflows
Scale to 30 FL counties via TinyFish
First 10 closed deals (proof of ROI)
DealRadar UI v1 live on Vercel
Activate BatchData LLC piercing
Build out probate court scraping
Phase 2 · Q3–Q4 2025 Southeast
Expand to Georgia, Texas, North Carolina
Replicate county clerk scraping per state
Hire 1 acquisitions manager
Launch wholesale buyer network (cash buyers list)
Build double-close deal structures
Multi-state DealRadar UI with state selector
Phase 3 · 2026+ National
Full 50-state county clerk coverage
White-label DealRadar for other wholesalers (SaaS)
AI-powered ARV (After Repair Value) modeling
Direct lender relationships for assignment fee lines
AI-generated offer letters + contract automation
PropTech acquisition target or licensing

Technical Risk Mitigation

  • Scraper blocking: Rotating US residential proxies via TinyFish stealth profiles. County sites that block — fall back to ATTOM.
  • Platform changes (Zillow/Redfin): Apify actor community maintains selectors. Multi-source redundancy means no single point of failure.
  • TCPA/DNC compliance: Pre-outreach DNC verification on every number. Litigator flag on crm.leads gates all SMS sends.
  • Supabase downtime: n8n error handling + retry logic. Apify raw_payloads table acts as durable buffer before DB write.

Team & Operations

  • Founder: Daniel Fernandez, Fyxie AI — system architecture, AI orchestration, sales
  • Acquisitions: Remote hire (Phase 2) — reviews score 60–84 leads, negotiates contracts
  • Dispositions: Remote hire (Phase 2) — manages cash buyer network, closes assignments
  • Current state: Fully founder-operated. Automation handles prospecting; founder handles negotiations.

DealRadar UI Architecture

React 19 + TypeScript + Tailwind CSS v4 + Framer Motion · Deployed on Vercel · Real-time via Supabase Realtime

Pages

DashboardMain
Stats row · Florida heatmap · Lead feed · Pipeline · Analytics row · Market table
Heatmap PageMap
Interactive FL SVG with zip code hotness overlay · Click-to-pull-leads
Leads PageCRM
Sortable/filterable lead list · Inline skip trace actions · AI score display
Pipeline PageDeals
Kanban: new → skip_traced → in_ghl → under_contract
Analytics PageBI
Market trends · Conversion funnels · Outreach performance
Skip Trace PageOps
Bulk skip trace widget with cost tracking and provider selection
SettingsConfig
API key management · Workflow configuration

Key Components

FloridaHeatmapInteractive SVG + zip overlay
LeadFeedReal-time scored leads + actions
LeadDetailPanelSlide-out: property + owner data
PipelineKanban board · deal stages
SkipTraceWidgetBulk trace + cost tracker
AIOutreachPanelSMS viewer + AI preview
MarketTableZip code rankings by hotness
StatsRowKPI cards: leads · traced · outreach · closed

Looker Studio BI Layer

Supabase connector feeds Looker Studio for board-level dashboards. Used for investor reporting, pipeline reviews, and marketing performance tracking. Separate from the operational DealRadar UI.

DEALRADAR · FYXIE AI · ORLANDO, FLORIDA · DANIEL FERNANDEZ · CONFIDENTIAL
Built on n8n · Supabase · Apify · TinyFish · ATTOM · Claude · GoHighLevel · Twilio