THE MOST CRITICAL FEATURE FOR HODGES & FOOSHEE REALTY. Ultimate distressed property intelligence system tracking 22 distress signals across financial, life event, property, legal, and market categories. Self-improving AI that monitors performance, spawns specialized skills to fix bottlenecks, runs A/B tests, and continuously optimizes lead conversion. This is the game-changer that finds homeowners in crisis BEFORE anyone else does. Use for ALL lead generation tasks, pipeline optimization, multi-source data ingestion, and automated lead intelligence.
This is not just a feature. This is THE revenue engine for Hodges & Fooshee Realty.
Projected Impact:
Core Capability: Track 22 types of distressed property signals, find homeowners in crisis BEFORE banks/agents/competitors, validate contacts automatically, score urgency based on multiple distress signals, and deliver qualified leads to agents with phone numbers and emails ready to call.
ALWAYS trigger this skill when the user mentions:
This is THE priority feature. Treat every Lead Hunter question with maximum urgency.
Lead Hunter Prime tracks ALL types of property distress, not just foreclosures:
The Power of Multiple Signals:
Example: "Tax delinquent + Probate + Divorce + Code violation" = Tier S lead (URGENT, call immediately)
1. MONITOR → Track every lead through pipeline (22 sources)
2. ANALYZE → Identify bottlenecks automatically (which sources convert best)
3. SPAWN → Create specialized skill to fix problem (e.g., improve skip tracing for divorces)
4. TEST → Run A/B test with 50% of leads
5. DECIDE → Keep if it works, kill if it doesn't
6. REPEAT → Forever (continuous optimization)
If building from scratch:
# Check for existing lead tracking infrastructure
view /mnt/user-data/uploads/supabase/migrations
view /mnt/user-data/uploads/app/api
# Look for existing lead forms, CRM integrations, or tracking systems
view /mnt/user-data/uploads/components
Extract these details:
If improving existing system: Ask user:
Database Schema Required:
Create tables to track:
leads - Every incoming lead with source attributionlead_events - Every action on a lead (viewed, contacted, qualified, etc.)lead_experiments - A/B tests running on subsets of leadslead_metrics - Aggregated performance by source, agent, time periodlead_skills - Spawned skills and their performance impactKey Metrics to Monitor:
Read references/monitoring-framework.md for complete schema and tracking implementation.
Common Bottleneck Patterns:
Low Response Rate → Spawn "instant-responder" skill
Poor Qualification → Spawn "lead-scorer" skill
Slow Follow-Up → Spawn "follow-up-sequencer" skill
Weak Lead Sources → Spawn "source-optimizer" skill
Analysis Query Template:
-- Find conversion rates by pipeline stage
SELECT
stage,
COUNT(*) as lead_count,
COUNT(*) FILTER (WHERE converted = true) as converted_count,
(COUNT(*) FILTER (WHERE converted = true)::float / COUNT(*)) * 100 as conversion_rate
FROM lead_events
WHERE created_at > NOW() - INTERVAL '30 days'
GROUP BY stage
ORDER BY stage;
-- Identify slowest stage transitions
SELECT
from_stage,
to_stage,
AVG(time_in_stage) as avg_duration,
PERCENTILE_CONT(0.5) WITHIN GROUP (ORDER BY time_in_stage) as median_duration
FROM lead_stage_transitions
GROUP BY from_stage, to_stage
ORDER BY avg_duration DESC;
When a bottleneck is identified, create a targeted skill to fix it.
Skill Spawning Protocol:
Define the problem clearly
Design the intervention
Create A/B test plan
Implement tracking
Example Spawned Skill: instant-responder
// Auto-respond to new leads within 60 seconds
export async function handleNewLead(lead: Lead) {
// Check if this lead is in the treatment group for instant-responder experiment
const experiment = await getActiveExperiment('instant-responder');
const isControl = lead.id % 2 === 0; // Simple 50/50 split
if (isControl) {
// Control group: existing process (manual agent follow-up)
await assignToAgent(lead);
return;
}
// Treatment group: instant AI response
const personalizedMessage = await generateResponse(lead);
await sendSMS(lead.phone, personalizedMessage);
await logEvent(lead.id, 'instant_response_sent', { experiment_id: experiment.id });
// Still assign to agent for human follow-up
await assignToAgent(lead);
}
Testing Framework:
Read references/ab-testing-protocol.md for detailed implementation, but core principles:
Decision Criteria:
interface ExperimentResults {
control: {
sample_size: number;
conversion_rate: number;
avg_time_to_convert: number;
};
treatment: {
sample_size: number;
conversion_rate: number;
avg_time_to_convert: number;
};
lift: number; // % improvement
p_value: number; // statistical significance
recommendation: 'KEEP' | 'KILL' | 'CONTINUE_TESTING';
}
Keep the skill if:
Kill the skill if:
Continue testing if:
The Continuous Improvement Cycle:
Priority Ranking System:
Always optimize in this order:
If no lead system exists:
Don't optimize what you haven't measured first.
If conversion rates vary wildly week-to-week:
Prioritize by potential impact:
Only test ONE skill at a time. Running multiple experiments simultaneously creates confounding variables.
Example: Instant-responder works great for Zillow leads but hurts conversion on referral leads.
Solution: Segment-specific deployment
lead_experiments tableComplete database schema for tracking:
Read this when: Setting up performance monitoring or designing tracking infrastructure.
Detailed guide to running statistically valid A/B tests:
Read this when: Launching a new experiment or analyzing test results.
Library of pre-built skills for common bottlenecks:
Read this when: Identifying a bottleneck and need a proven solution pattern.
User: "Only 40% of our leads are getting contacted within 24 hours. How do we fix this?"
Response:
[Analyzes current lead routing and agent workflow]
The bottleneck is agent response time. Let me spawn an instant-responder skill.
**The Skill:**
- Triggers within 60 seconds of lead capture
- Uses LLM to generate personalized SMS based on property address and lead source
- Still assigns to agent, but lead gets immediate acknowledgment
**A/B Test Design:**
- Control: existing process (agent contacts within 24h)
- Treatment: instant AI response + agent follow-up
- Metric: % of leads that respond to initial contact
- Sample size: 200 leads (100 per group)
- Duration: 2 weeks
**Hypothesis:** Instant response increases engagement rate from 40% to 65%+
[Implements skill, sets up tracking, launches experiment]
I'll check results in 2 weeks and decide: keep, kill, or continue testing.
User: "We're getting tons of leads but only 10% are qualified. We're wasting time on tire-kickers."
Response:
[Analyzes lead sources and qualification criteria]
The bottleneck is lead scoring. Most leads aren't pre-qualified before agent contact.
**Spawning lead-scorer skill:**
Analyzes each incoming lead for:
- Property characteristics (price, condition, equity position)
- Lead behavior (time on site, pages viewed, form urgency signals)
- Demographics (location, device type, referral source)
Assigns A/B/C/D score:
- A leads: high equity, motivated seller signals → immediate agent contact
- B leads: potential, needs nurturing → follow-up sequence
- C leads: low quality, may convert later → drip campaign
- D leads: spam or unqualified → archive
**A/B Test Design:**
- Control: all leads routed to agents equally
- Treatment: leads prioritized by score (A/B get fast response, C/D get automated nurture)
- Metric: % of agent time spent on qualified leads that convert
- Duration: 3 weeks
[Implements lead scoring model, sets up routing rules, launches test]
Use this skill when:
Don't use this skill for:
Built for Hodges & Fooshee Realty by AICA 🔥