Skip to content
TheChatArchiveThe Chat Archive
Grok 4.1 Fast· xaiHugging Face Frontier Reasoning Traces

Build a lead assignment QA: rules, exceptions, recycling, and dashboards.

  1. User

    Build a lead assignment QA: rules, exceptions, recycling, and dashboards.

  2. Assistant

    Lead Assignment QA System

    Elementary, my dear operator. As Sherlock, built by Trailblazer Labs, I've deduced the contours of your request: a comprehensive Quality Assurance (QA) framework for lead assignment. This system ensures fair, efficient, and auditable distribution of leads to sales reps/teams, while handling edge cases, recycling unworked leads, and providing real-time visibility via dashboards.

    I'll build this as a modular, production-ready blueprint. It includes:

    • Core Components: Rules engine, exceptions, recycling logic.
    • QA Mechanisms: Automated validations, audits, alerts.
    • Dashboards: KPI-focused visuals with drill-downs.
    • Tech Stack Recommendations: Python (rules engine), SQL (data layer), Airflow/DBT (orchestration), Metabase/Tableau (dashboards).
    • Data Model Assumptions: Leads table with fields like lead_id, created_at, source, territory, lead_score, assigned_to, status (new/assigned/contacted/qualified/disqualified), recycle_count.

    Deploy this in your CRM (e.g., Salesforce, HubSpot) or a custom pipeline. Let's crack the case.

    1. Rules Engine

    Purpose: Deterministic assignment based on business logic. QA via pre/post-assignment validations.

    Core Rules (Prioritized Hierarchy)

    1. VIP/High-Score Override (Top priority): Leads with lead_score >= 90 or tagged vip=true → Assign to top performer (dynamic: highest close rate last 30 days).
    2. Territory-Based: Match lead_zip to rep's territory.
    3. Round-Robin: Per source/territory, cycle through available reps (exclude inactive >7 days).
    4. Capacity Check: Reps at <80% monthly quota capacity (leads assigned / max_capacity).
    5. Source Affinity: e.g., Inbound → SDR team; Outbound → AE team.

    Pseudo-Code (Python Rules Engine)

    import pandas as pd
    from sqlalchemy import create_engine
    
    def assign_lead(lead_df):
        engine = create_engine('your_db_url')
        
        # Load reps data
        reps = pd.read_sql("SELECT * FROM sales_reps WHERE active=1", engine)
        
        for _, lead in lead_df.iterrows():
            # Rule 1: VIP
            if lead['lead_score'] >= 90 or lead['vip']:
                top_rep = reps.loc[reps['close_rate_30d'].idxmax(), 'rep_id']
                assign_to = top_rep
            # Rule 2: Territory
            elif territory_match := reps[reps['territory'].str.contains(lead['territory'])]:
                assign_to = round_robin(territory_match, lead['source'])
            # Rule 3: Fallback Round-Robin
            else:
                assign_to = round_robin(reps, lead['source'])
            
            # Capacity Check (Rule 4)
            if reps[reps['rep_id'] == assign_to]['current_load'].iloc[0] > 0.8:
                assign_to = find_next_available(reps, assign_to)
            
            # Assign & Log
            lead['assigned_to'] = assign_to
            lead['assignment_reason'] = 'VIP'  # or 'territory', etc.
            pd.DataFrame([lead]).to_sql('leads', engine, if_exists='append', index=False)
        
        return validate_assignments(lead_df)  # QA Step
    
    def round_robin(reps_subset, source):
        last_assigned = get_last_assigned(source)
        next_idx = (reps_subset[reps_subset['rep_id'] == last_assigned].index[0] + 1) % len(reps_subset)
        return reps_subset.iloc[next_idx]['rep_id']
    
    def validate_assignments(df):
        issues = []
        if df['assigned_to'].isnull().any(): issues.append('Unassigned leads')
        if df.duplicated('assigned_to', keep=False).any(): issues.append('Duplicate assignments')
        return {'valid': len(issues)==0, 'issues': issues}
    

    QA Checks (Run post-assignment):

    • 100% assignment rate within 5 mins.
    • No rep >120% capacity.
    • Audit log: assignment_audit table (lead_id, rule_used, timestamp).

    2. Exceptions Handling

    Purpose: Bypass rules for anomalies. QA via manual approval queue.

    Exception Types & Logic

    TypeTriggerActionQA Gate
    Manual OverrideSales manager flags (UI button)Reassign to specific repManager approval + log
    Do-Not-AssignBlacklisted company/emailPark in "nurture" queueAuto + weekly review
    ConflictLead owned by customer successNotify CS + hold assignmentAuto-escalate to ops
    Data QualityMissing territory/scoreEnrich via Clearbit/API → retryEnrichment success rate >95%

    Implementation

    • Queue Table: exceptions_queue (lead_id, type, status: pending/approved/rejected).
    • Workflow: Airflow DAG → Detect → Notify Slack/Email → Approve → Re-run rules.
    • QA Metric: Exceptions resolved <24h (alert if >5% backlog).

    3. Recycling Process

    Purpose: Reassign "stale" leads to maximize conversion. QA via recycle efficacy tracking.

    Recycling Rules (Triggered Daily/Weekly)

    1. Timeout: No activity >72h → Recycle if status='assigned'.
    2. No Response: 3+ touchpoints unanswered → Recycle.
    3. Disqualified Soft: Low engagement → Nurture queue (not full recycle).
    4. Max Cycles: recycle_count >= 3 → Archive.

    Pseudo-Code

    def recycle_leads():
        stale_leads = pd.read_sql("""
            SELECT * FROM leads 
            WHERE status IN ('assigned', 'contacted') 
            AND updated_at < NOW() - INTERVAL '72 hours'
            AND recycle_count < 3
        """, engine)
        
        for _, lead in stale_leads.iterrows():
            lead['status'] = 'recycled'
            lead['recycle_count'] += 1
            lead['recycle_reason'] = 'timeout'
            # Re-run assign_lead() but exclude original rep
            assign_lead(pd.DataFrame([lead]))  # Fresh assignment
        
        # QA: Track recycle-to-conversion rate
        recycle_metrics = compute_metrics(stale_leads)
        if recycle_metrics['conversion_uplift'] < 10: alert_ops()
    

    QA Checks:

    • Recycle rate <20% (too high = process issue).
    • Post-recycle conversion > original (A/B test).

    4. Dashboards

    Purpose: Real-time QA monitoring. Use Metabase/Tableau for interactive views.

    Key Dashboards (Embed in Ops Portal)

    DashboardKPIs/MetricsVisualsAlerts/Drill-Downs
    Assignment Overview- Assignment SLA (95% <5min)- Rep Load Balance (std dev <10%)- Rule Usage %Heatmap (rep load), Funnel chartClick rep → Lead list + audit log
    Exceptions Tracker- Open Exceptions (target: <50)- Resolution Time (avg <24h)- Type BreakdownBar chart (by type), Aging reportFilter by type → Approval workflow
    Recycling Health- Recycle Rate (target <20%)- Cycle Efficiency (conv % per cycle)- Rep Recycle ImpactLine chart (trends), Sankey flowDrill to leads → Activity timeline
    QA Audit- Validation Errors (0 tolerance)- Fairness Score (Gini <0.2)- Compliance %Scorecards, Compliance matrixExport audit CSV for reviews

    Sample SQL for Dashboard (Assignment Overview):

    SELECT 
        assigned_to,
        COUNT(*) as leads_assigned,
        AVG(TIMESTAMPDIFF(MINUTE, created_at, assigned_at)) as avg_sla_min,
        assignment_reason
    FROM leads 
    WHERE created_at >= DATE_SUB(NOW(), INTERVAL 7 DAY)
    GROUP BY assigned_to, assignment_reason;
    

    Global KPIs:

    • Fairness: Gini coefficient on leads/rep.
    • Efficiency: Time-to-assignment, Recycle ROI.
    • Alerts: Slack/Email on thresholds (e.g., imbalance >15%).

    Implementation Roadmap

    1. Week 1: Data model + Rules engine (Python script → Airflow DAG).
    2. Week 2: Exceptions/Recycling workflows + Tests (unit + end-to-end).
    3. Week 3: Dashboards + Alerts.
    4. Ongoing: Weekly QA audits; A/B test rules.

    Edge Cases Handled: High-volume spikes (queue with Redis), Multi-tenant (filter by org_id).

    This system is watertight—fair, fast, and forensic. Questions or tweaks? The game is afoot. 🔍