Pharmeasy Sales Drop Diagnostic Workflow

Pharmeasy Sales Drop Diagnostic Workflow - Automated root cause analysis and recovery

Back

Workflow Information

ID: pharmeasy_sales_drop_diagnostic

Namespace: pharmeasy

Version: 1.0

Created: 2025-07-28

Updated: 2025-07-28

Tasks: 6

Quick Actions

Manage Secrets

Inputs

Name	Type	Required	Default
`clevertap_account_id`	string	Required	None
`clevertap_passcode`	string	Required	None
`internal_api_key`	string	Required	None
`sales_drop_percentage`	float	Required	None
`affected_region`	string	Required	None
`affected_category`	string	Required	None
`lookback_hours`	integer	Optional	`24`

Outputs

Name	Type	Source
`internal_data`	object	Operational data from internal systems
`sales_anomaly`	object	Detected sales anomaly details
`clevertap_data`	object	User behavior data from CleverTap
`recovery_campaign`	object	Recovery campaign execution results
`alert_notification`	object	Alert generated for stakeholders
`root_cause_analysis`	object	Multi-hypothesis root cause analysis results

Tasks

detect_anomaly

script

Step 1 - Detect and quantify the sales anomaly

Python Script:

import json
from datetime import datetime, timedelta
import random
import math

print("🚨 SALES DROP DIAGNOSTIC WORKFLOW INITIATED")
print("=" * 60)
print("")

# Input parameters
sales_drop = float("${sales_drop_percentage}")
region = "${affected_region}"
category = "${affected_category}"
lookback_hours = int("${lookback_hours}")

print(f"📉 Anomaly Detection Parameters:")
print(f"   Sales Drop: {sales_drop:.1f}%")
print(f"   Region: {region}")
print(f"   Category: {category}")
print(f"   Analysis Window: {lookback_hours} hours")
print("")

# Generate time series data for anomaly context
current_time = datetime.now()
time_series = []

for i in range(lookback_hours):
    hour_ago = current_time - timedelta(hours=i)
    if i < 6:  # Recent hours show the drop
        hourly_sales = random.randint(80000, 120000) * (1 - sales_drop/100)
    else:  # Normal sales before the drop
        hourly_sales = random.randint(100000, 150000)
    
    time_series.append({
        "timestamp": hour_ago.isoformat(),
        "sales": int(hourly_sales),
        "orders": int(hourly_sales / random.randint(500, 700))
    })

# Calculate statistics
recent_avg = sum(ts['sales'] for ts in time_series[:6]) / 6
historical_avg = sum(ts['sales'] for ts in time_series[6:]) / (len(time_series) - 6)
actual_drop = (historical_avg - recent_avg) / historical_avg * 100

# Calculate z-score
all_sales = [ts['sales'] for ts in time_series]
mean_sales = sum(all_sales) / len(all_sales)
std_dev = math.sqrt(sum((x - mean_sales) ** 2 for x in all_sales) / len(all_sales))
z_score = (recent_avg - mean_sales) / std_dev if std_dev > 0 else 0

print("📊 Anomaly Analysis Results:")
print(f"   Historical Average: ₹{historical_avg:,.0f}/hour")
print(f"   Current Average: ₹{recent_avg:,.0f}/hour")
print(f"   Actual Drop: {actual_drop:.1f}%")
print(f"   Z-Score: {z_score:.2f}")
print(f"   Severity: {'CRITICAL' if actual_drop > 20 else 'HIGH' if actual_drop > 15 else 'MODERATE'}")

anomaly_result = {
    "timestamp": current_time.isoformat(),
    "region": region,
    "category": category,
    "reported_drop": sales_drop,
    "actual_drop": round(actual_drop, 1),
    "z_score": round(z_score, 2),
    "severity": "CRITICAL" if actual_drop > 20 else "HIGH" if actual_drop > 15 else "MODERATE",
    "time_series_data": time_series,
    "recent_avg_sales": recent_avg,
    "historical_avg_sales": historical_avg,
    "detection_confidence": 0.95 if abs(z_score) > 2 else 0.85
}

print("")
print("✅ Anomaly detection complete. Proceeding to data collection...")
print(f"__OUTPUTS__ {json.dumps(anomaly_result)}")

fetch_clevertap_data

script

Step 2A - Fetch user behavior data from CleverTap (parallel)

Python Script:

import json
import time
import random
from datetime import datetime, timedelta

print("")
print("📱 FETCHING CLEVERTAP DATA")
print("-" * 40)

anomaly_data = json.loads('${detect_anomaly}')
region = anomaly_data['region']
category = anomaly_data['category']
lookback = int("${lookback_hours}")

print(f"🔍 Querying CleverTap for {region} - {category}")
print("   Events: app_open, search, product_view, add_to_cart, checkout, purchase")

# Simulate API call delay
time.sleep(1)

# Generate behavioral data showing funnel drop patterns
clevertap_data = {
    "query_params": {
        "region": region,
        "category": category,
        "time_range": f"last_{lookback}_hours"
    },
    "funnel_data": {
        "app_opens": random.randint(25000, 35000),
        "searches": random.randint(18000, 25000),
        "product_views": random.randint(15000, 20000),
        "add_to_cart": random.randint(8000, 12000),
        "checkout_initiated": random.randint(4000, 6000),
        "purchases": random.randint(2500, 4000)
    },
    "conversion_rates": {
        "search_to_view": 0.82,
        "view_to_cart": 0.45,
        "cart_to_checkout": 0.48,  # Lower than normal
        "checkout_to_purchase": 0.65  # Lower than normal
    },
    "user_segments": {
        "new_users": {
            "count": random.randint(3000, 5000),
            "conversion_rate": 0.05
        },
        "returning_users": {
            "count": random.randint(15000, 20000),
            "conversion_rate": 0.12
        },
        "chronic_patients": {
            "count": random.randint(8000, 10000),
            "conversion_rate": 0.08  # Lower than normal
        }
    },
    "device_breakdown": {
        "android": 0.75,
        "ios": 0.20,
        "web": 0.05
    },
    "cart_abandonment": {
        "rate": 0.52,  # Higher than normal
        "avg_cart_value": random.randint(450, 650),
        "top_abandoned_products": [
            "Metformin 500mg",
            "Amlodipine 5mg",
            "Paracetamol 650mg"
        ]
    },
    "session_metrics": {
        "avg_duration_seconds": 245,
        "pages_per_session": 4.2,
        "bounce_rate": 0.32
    }
}

print("✓ CleverTap data retrieved successfully")
print(f"  - Total events: {sum(clevertap_data['funnel_data'].values()):,}")
print(f"  - Cart abandonment rate: {clevertap_data['cart_abandonment']['rate']:.1%}")
print(f"  - Chronic patient conversion: {clevertap_data['user_segments']['chronic_patients']['conversion_rate']:.1%}")

print(f"__OUTPUTS__ {json.dumps(clevertap_data)}")

fetch_internal_data

script

Step 2B - Fetch operational data from internal systems (parallel)

Python Script:

import json
import time
import random
from datetime import datetime

print("")
print("🏢 FETCHING INTERNAL DASHBOARD DATA")
print("-" * 40)

anomaly_data = json.loads('${detect_anomaly}')
region = anomaly_data['region']
category = anomaly_data['category']

print(f"🔍 Querying internal systems for {region} - {category}")
print("   Systems: Inventory, Payment, Logistics, Pricing")

# Simulate API call delay
time.sleep(1)

# Generate operational data that could explain the sales drop
internal_data = {
    "query_params": {
        "region": region,
        "category": category,
        "timestamp": datetime.now().isoformat()
    },
    "inventory_status": {
        "total_skus": 2500,
        "in_stock_skus": 2100,
        "stockout_skus": 400,  # Higher than normal
        "critical_stockouts": 85,  # Key products out of stock
        "top_stockout_products": [
            {"name": "Metformin 500mg", "demand": "HIGH", "days_out": 3},
            {"name": "Insulin Glargine", "demand": "CRITICAL", "days_out": 2},
            {"name": "Losartan 50mg", "demand": "HIGH", "days_out": 1}
        ],
        "stockout_impact_percentage": 18.5
    },
    "payment_gateway_status": {
        "success_rate": 0.89,  # Lower than normal 95%
        "failed_transactions": 450,
        "gateway_errors": {
            "timeout": 180,
            "authentication_failed": 120,
            "insufficient_funds": 150
        },
        "affected_payment_methods": ["credit_card", "debit_card"],
        "avg_processing_time_ms": 4500  # Higher than normal
    },
    "logistics_metrics": {
        "on_time_delivery_rate": 0.82,  # Lower than normal
        "avg_delivery_time_hours": 36,
        "delayed_orders": 850,
        "delivery_partner_issues": {
            "partner_1": {"name": "BlueDart", "performance": 0.78},
            "partner_2": {"name": "Delhivery", "performance": 0.85}
        }
    },
    "pricing_data": {
        "recent_price_changes": 12,
        "avg_price_increase_percentage": 8.5,
        "competitor_price_comparison": {
            "status": "HIGHER",
            "difference_percentage": 5.2
        }
    },
    "customer_service": {
        "complaint_volume_increase": 35,  # 35% increase
        "top_complaints": [
            "Product not available",
            "Payment failure",
            "Delayed delivery"
        ],
        "resolution_rate": 0.72
    }
}

print("✓ Internal data retrieved successfully")
print(f"  - Stockout SKUs: {internal_data['inventory_status']['stockout_skus']}")
print(f"  - Payment success rate: {internal_data['payment_gateway_status']['success_rate']:.1%}")
print(f"  - On-time delivery: {internal_data['logistics_metrics']['on_time_delivery_rate']:.1%}")

print(f"__OUTPUTS__ {json.dumps(internal_data)}")

analyze_root_cause

script

Step 3 - Analyze root cause using multi-hypothesis testing

Python Script:

import json
from datetime import datetime

print("")
print("🔬 MULTI-HYPOTHESIS ROOT CAUSE ANALYSIS")
print("=" * 50)
print("")

# Parse all data sources
anomaly_data = json.loads('${detect_anomaly}')
clevertap_data = json.loads('${fetch_clevertap_data}')
internal_data = json.loads('${fetch_internal_data}')

print("📊 Analyzing data from multiple sources...")
print(f"   Sales Drop: {anomaly_data['actual_drop']:.1f}%")
print(f"   Data Sources: CleverTap + Internal Systems")
print("")

# Define and test hypotheses
hypotheses = {
    "inventory_stockouts": {
        "description": "Critical product stockouts causing lost sales",
        "evidence": [],
        "score": 0.0,
        "impact": "HIGH"
    },
    "payment_gateway_issues": {
        "description": "Payment failures preventing order completion",
        "evidence": [],
        "score": 0.0,
        "impact": "MEDIUM"
    },
    "user_experience_friction": {
        "description": "High cart abandonment due to UX issues",
        "evidence": [],
        "score": 0.0,
        "impact": "MEDIUM"
    },
    "delivery_delays": {
        "description": "Poor delivery performance affecting repeat purchases",
        "evidence": [],
        "score": 0.0,
        "impact": "MEDIUM"
    },
    "price_sensitivity": {
        "description": "Recent price increases causing customer churn",
        "evidence": [],
        "score": 0.0,
        "impact": "LOW"
    }
}

# Test Hypothesis 1: Inventory Stockouts
stockout_impact = internal_data['inventory_status']['stockout_impact_percentage']
if stockout_impact > 15:
    hypotheses["inventory_stockouts"]["score"] = 0.85
    hypotheses["inventory_stockouts"]["evidence"].append(
        f"Stockout impact: {stockout_impact:.1f}% of potential sales"
    )
    hypotheses["inventory_stockouts"]["evidence"].append(
        f"Critical items out of stock: {internal_data['inventory_status']['critical_stockouts']}"
    )

# Test Hypothesis 2: Payment Gateway Issues
payment_success = internal_data['payment_gateway_status']['success_rate']
if payment_success < 0.92:
    hypotheses["payment_gateway_issues"]["score"] = 0.70
    hypotheses["payment_gateway_issues"]["evidence"].append(
        f"Payment success rate: {payment_success:.1%} (below 92% threshold)"
    )
    hypotheses["payment_gateway_issues"]["evidence"].append(
        f"Failed transactions: {internal_data['payment_gateway_status']['failed_transactions']}"
    )

# Test Hypothesis 3: User Experience Friction
cart_abandonment = clevertap_data['cart_abandonment']['rate']
checkout_conversion = clevertap_data['conversion_rates']['checkout_to_purchase']
if cart_abandonment > 0.45 and checkout_conversion < 0.70:
    hypotheses["user_experience_friction"]["score"] = 0.75
    hypotheses["user_experience_friction"]["evidence"].append(
        f"Cart abandonment rate: {cart_abandonment:.1%}"
    )
    hypotheses["user_experience_friction"]["evidence"].append(
        f"Low checkout conversion: {checkout_conversion:.1%}"
    )

# Test Hypothesis 4: Delivery Delays
on_time_delivery = internal_data['logistics_metrics']['on_time_delivery_rate']
if on_time_delivery < 0.85:
    hypotheses["delivery_delays"]["score"] = 0.60
    hypotheses["delivery_delays"]["evidence"].append(
        f"On-time delivery: {on_time_delivery:.1%}"
    )
    hypotheses["delivery_delays"]["evidence"].append(
        f"Delayed orders: {internal_data['logistics_metrics']['delayed_orders']}"
    )

# Test Hypothesis 5: Price Sensitivity
if internal_data['pricing_data']['avg_price_increase_percentage'] > 5:
    hypotheses["price_sensitivity"]["score"] = 0.45
    hypotheses["price_sensitivity"]["evidence"].append(
        f"Recent price increase: {internal_data['pricing_data']['avg_price_increase_percentage']:.1f}%"
    )

# Rank hypotheses by score
ranked_hypotheses = sorted(
    hypotheses.items(), 
    key=lambda x: x[1]['score'], 
    reverse=True
)

print("🎯 Hypothesis Testing Results:")
print("")
for rank, (hypothesis, data) in enumerate(ranked_hypotheses[:3], 1):
    print(f"{rank}. {hypothesis.replace('_', ' ').title()}")
    print(f"   Score: {data['score']:.2f}")
    print(f"   Impact: {data['impact']}")
    for evidence in data['evidence']:
        print(f"   - {evidence}")
    print("")

# Determine primary root cause
primary_cause = ranked_hypotheses[0][0]
confidence = "HIGH" if ranked_hypotheses[0][1]['score'] > 0.7 else "MEDIUM"

# Generate recommendations
recommendations = []
if primary_cause == "inventory_stockouts":
    recommendations = [
        "Immediately notify affected customers about stockouts",
        "Activate substitute product recommendations",
        "Expedite replenishment for critical SKUs",
        "Launch 'back in stock' notification campaign"
    ]
elif primary_cause == "payment_gateway_issues":
    recommendations = [
        "Switch to backup payment gateway",
        "Promote COD as alternative payment method",
        "Send payment retry notifications",
        "Offer payment assistance through customer support"
    ]
elif primary_cause == "user_experience_friction":
    recommendations = [
        "Simplify checkout process immediately",
        "Launch cart recovery campaign with incentives",
        "A/B test one-click checkout",
        "Reduce form fields in checkout"
    ]

root_cause_result = {
    "timestamp": datetime.now().isoformat(),
    "hypotheses_tested": len(hypotheses),
    "hypothesis_results": dict(ranked_hypotheses),
    "primary_root_cause": primary_cause,
    "confidence_level": confidence,
    "contributing_factors": [h[0] for h in ranked_hypotheses if h[1]['score'] > 0.5],
    "recommendations": recommendations,
    "estimated_recovery_time": "4-6 hours" if primary_cause in ["payment_gateway_issues", "user_experience_friction"] else "24-48 hours"
}

print("✅ Root cause analysis complete.")
print(f"   Primary Cause: {primary_cause.replace('_', ' ').title()}")
print(f"   Confidence: {confidence}")

print(f"__OUTPUTS__ {json.dumps(root_cause_result)}")

generate_alert

ai_agent

Step 4 - Generate intelligent alerts for stakeholders

AI Agent Configuration:

Model Configuration

Model Client ID: alert_generator
Agent Type: assistant

Input Format: text
Output Format: text

trigger_recovery_campaign

script

Step 5 - Trigger automated recovery campaigns

Python Script:

import json
import time
import random
from datetime import datetime

print("")
print("🚀 EXECUTING RECOVERY CAMPAIGNS")
print("=" * 50)
print("")

# Parse previous results
root_cause_data = json.loads('${analyze_root_cause}')
alert_data = json.loads('${generate_alert}')
primary_cause = root_cause_data['primary_root_cause']

print(f"📢 Launching targeted campaigns for: {primary_cause.replace('_', ' ').title()}")
print("")

campaigns_executed = []

if primary_cause == "inventory_stockouts":
    print("🔔 Campaign 1: Stockout Notification & Alternatives")
    campaign1 = {
        "name": "Critical Stockout Alert Campaign",
        "type": "multi_channel",
        "channels": ["push_notification", "in_app", "email"],
        "segment": "users_with_oos_items_in_cart",
        "message_template": "Important: {product} is currently out of stock. Similar alternatives available!",
        "cta": "View Alternatives",
        "personalization": "product_based"
    }
    
    # Simulate campaign execution
    time.sleep(0.5)
    campaign1["execution_stats"] = {
        "targeted_users": random.randint(3000, 5000),
        "messages_sent": random.randint(2800, 4800),
        "opens": random.randint(1500, 2500),
        "clicks": random.randint(800, 1200),
        "conversions": random.randint(400, 600)
    }
    campaigns_executed.append(campaign1)
    print("   ✓ Notifications sent to affected users")
    
    print("")
    print("💊 Campaign 2: Substitute Product Recommendations")
    campaign2 = {
        "name": "Smart Alternative Suggestions",
        "type": "personalized_recommendations",
        "algorithm": "collaborative_filtering",
        "display_locations": ["home_page", "search_results", "cart_page"]
    }
    campaign2["execution_stats"] = {
        "recommendations_shown": random.randint(8000, 12000),
        "clicks": random.randint(2000, 3000),
        "add_to_cart": random.randint(800, 1200),
        "purchases": random.randint(400, 700)
    }
    campaigns_executed.append(campaign2)
    print("   ✓ Alternative products activated")
    
elif primary_cause == "payment_gateway_issues":
    print("💳 Campaign 1: Payment Retry Assistance")
    campaign1 = {
        "name": "Payment Recovery Campaign",
        "type": "targeted_sms_email",
        "segment": "failed_payment_last_24h",
        "offers": ["COD available", "Try alternate payment method", "10% off on retry"]
    }
    campaign1["execution_stats"] = {
        "targeted_users": random.randint(400, 600),
        "messages_sent": random.randint(380, 580),
        "payment_retries": random.randint(150, 250),
        "successful_orders": random.randint(100, 180)
    }
    campaigns_executed.append(campaign1)
    print("   ✓ Payment assistance campaign launched")
    
elif primary_cause == "user_experience_friction":
    print("🛒 Campaign 1: Cart Recovery with Incentives")
    campaign1 = {
        "name": "Abandoned Cart Win-back",
        "type": "progressive_discount",
        "discount_tiers": ["5% after 2 hours", "10% after 6 hours", "15% after 24 hours"],
        "segment": "cart_abandoners",
        "urgency_messaging": True
    }
    campaign1["execution_stats"] = {
        "targeted_users": random.randint(5000, 8000),
        "emails_sent": random.randint(4800, 7800),
        "cart_recoveries": random.randint(800, 1200),
        "revenue_recovered": random.randint(400000, 600000)
    }
    campaigns_executed.append(campaign1)
    print("   ✓ Cart recovery campaign activated")

# Calculate total impact
total_conversions = sum(
    c["execution_stats"].get("conversions", 0) + 
    c["execution_stats"].get("successful_orders", 0) + 
    c["execution_stats"].get("purchases", 0) 
    for c in campaigns_executed
)

estimated_revenue_recovery = total_conversions * random.randint(500, 800)

print("")
print("📊 Campaign Execution Summary:")
print(f"   Campaigns Launched: {len(campaigns_executed)}")
print(f"   Total Conversions: {total_conversions:,}")
print(f"   Estimated Revenue Recovery: ₹{estimated_revenue_recovery:,}")
print(f"   Recovery ETA: {root_cause_data['estimated_recovery_time']}")

recovery_result = {
    "timestamp": datetime.now().isoformat(),
    "primary_intervention": primary_cause,
    "campaigns_executed": campaigns_executed,
    "total_campaigns": len(campaigns_executed),
    "total_conversions": total_conversions,
    "estimated_revenue_recovery": estimated_revenue_recovery,
    "recovery_status": "IN_PROGRESS",
    "next_checkpoint": "2_hours",
    "success_metrics": {
        "target_recovery_percentage": 70,
        "current_recovery_percentage": random.randint(25, 35),
        "confidence_interval": "±5%"
    }
}

print("")
print("✅ Recovery campaigns executed successfully.")
print("   Monitoring performance and ready for optimization...")

print(f"__OUTPUTS__ {json.dumps(recovery_result)}")

YAML Source

id: pharmeasy_sales_drop_diagnostic
name: Pharmeasy Sales Drop Diagnostic Workflow
tasks:
- id: detect_anomaly
  name: Detect Sales Anomaly
  type: script
  script: "import json\nfrom datetime import datetime, timedelta\nimport random\n\
    import math\n\nprint(\"\U0001F6A8 SALES DROP DIAGNOSTIC WORKFLOW INITIATED\")\n\
    print(\"=\" * 60)\nprint(\"\")\n\n# Input parameters\nsales_drop = float(\"${sales_drop_percentage}\"\
    )\nregion = \"${affected_region}\"\ncategory = \"${affected_category}\"\nlookback_hours\
    \ = int(\"${lookback_hours}\")\n\nprint(f\"\U0001F4C9 Anomaly Detection Parameters:\"\
    )\nprint(f\"   Sales Drop: {sales_drop:.1f}%\")\nprint(f\"   Region: {region}\"\
    )\nprint(f\"   Category: {category}\")\nprint(f\"   Analysis Window: {lookback_hours}\
    \ hours\")\nprint(\"\")\n\n# Generate time series data for anomaly context\ncurrent_time\
    \ = datetime.now()\ntime_series = []\n\nfor i in range(lookback_hours):\n    hour_ago\
    \ = current_time - timedelta(hours=i)\n    if i < 6:  # Recent hours show the\
    \ drop\n        hourly_sales = random.randint(80000, 120000) * (1 - sales_drop/100)\n\
    \    else:  # Normal sales before the drop\n        hourly_sales = random.randint(100000,\
    \ 150000)\n    \n    time_series.append({\n        \"timestamp\": hour_ago.isoformat(),\n\
    \        \"sales\": int(hourly_sales),\n        \"orders\": int(hourly_sales /\
    \ random.randint(500, 700))\n    })\n\n# Calculate statistics\nrecent_avg = sum(ts['sales']\
    \ for ts in time_series[:6]) / 6\nhistorical_avg = sum(ts['sales'] for ts in time_series[6:])\
    \ / (len(time_series) - 6)\nactual_drop = (historical_avg - recent_avg) / historical_avg\
    \ * 100\n\n# Calculate z-score\nall_sales = [ts['sales'] for ts in time_series]\n\
    mean_sales = sum(all_sales) / len(all_sales)\nstd_dev = math.sqrt(sum((x - mean_sales)\
    \ ** 2 for x in all_sales) / len(all_sales))\nz_score = (recent_avg - mean_sales)\
    \ / std_dev if std_dev > 0 else 0\n\nprint(\"\U0001F4CA Anomaly Analysis Results:\"\
    )\nprint(f\"   Historical Average: \u20B9{historical_avg:,.0f}/hour\")\nprint(f\"\
    \   Current Average: \u20B9{recent_avg:,.0f}/hour\")\nprint(f\"   Actual Drop:\
    \ {actual_drop:.1f}%\")\nprint(f\"   Z-Score: {z_score:.2f}\")\nprint(f\"   Severity:\
    \ {'CRITICAL' if actual_drop > 20 else 'HIGH' if actual_drop > 15 else 'MODERATE'}\"\
    )\n\nanomaly_result = {\n    \"timestamp\": current_time.isoformat(),\n    \"\
    region\": region,\n    \"category\": category,\n    \"reported_drop\": sales_drop,\n\
    \    \"actual_drop\": round(actual_drop, 1),\n    \"z_score\": round(z_score,\
    \ 2),\n    \"severity\": \"CRITICAL\" if actual_drop > 20 else \"HIGH\" if actual_drop\
    \ > 15 else \"MODERATE\",\n    \"time_series_data\": time_series,\n    \"recent_avg_sales\"\
    : recent_avg,\n    \"historical_avg_sales\": historical_avg,\n    \"detection_confidence\"\
    : 0.95 if abs(z_score) > 2 else 0.85\n}\n\nprint(\"\")\nprint(\"\u2705 Anomaly\
    \ detection complete. Proceeding to data collection...\")\nprint(f\"__OUTPUTS__\
    \ {json.dumps(anomaly_result)}\")\n"
  description: Step 1 - Detect and quantify the sales anomaly
  timeout_seconds: 30
- id: fetch_clevertap_data
  name: Fetch CleverTap Behavioral Data
  type: script
  script: "import json\nimport time\nimport random\nfrom datetime import datetime,\
    \ timedelta\n\nprint(\"\")\nprint(\"\U0001F4F1 FETCHING CLEVERTAP DATA\")\nprint(\"\
    -\" * 40)\n\nanomaly_data = json.loads('${detect_anomaly}')\nregion = anomaly_data['region']\n\
    category = anomaly_data['category']\nlookback = int(\"${lookback_hours}\")\n\n\
    print(f\"\U0001F50D Querying CleverTap for {region} - {category}\")\nprint(\"\
    \   Events: app_open, search, product_view, add_to_cart, checkout, purchase\"\
    )\n\n# Simulate API call delay\ntime.sleep(1)\n\n# Generate behavioral data showing\
    \ funnel drop patterns\nclevertap_data = {\n    \"query_params\": {\n        \"\
    region\": region,\n        \"category\": category,\n        \"time_range\": f\"\
    last_{lookback}_hours\"\n    },\n    \"funnel_data\": {\n        \"app_opens\"\
    : random.randint(25000, 35000),\n        \"searches\": random.randint(18000, 25000),\n\
    \        \"product_views\": random.randint(15000, 20000),\n        \"add_to_cart\"\
    : random.randint(8000, 12000),\n        \"checkout_initiated\": random.randint(4000,\
    \ 6000),\n        \"purchases\": random.randint(2500, 4000)\n    },\n    \"conversion_rates\"\
    : {\n        \"search_to_view\": 0.82,\n        \"view_to_cart\": 0.45,\n    \
    \    \"cart_to_checkout\": 0.48,  # Lower than normal\n        \"checkout_to_purchase\"\
    : 0.65  # Lower than normal\n    },\n    \"user_segments\": {\n        \"new_users\"\
    : {\n            \"count\": random.randint(3000, 5000),\n            \"conversion_rate\"\
    : 0.05\n        },\n        \"returning_users\": {\n            \"count\": random.randint(15000,\
    \ 20000),\n            \"conversion_rate\": 0.12\n        },\n        \"chronic_patients\"\
    : {\n            \"count\": random.randint(8000, 10000),\n            \"conversion_rate\"\
    : 0.08  # Lower than normal\n        }\n    },\n    \"device_breakdown\": {\n\
    \        \"android\": 0.75,\n        \"ios\": 0.20,\n        \"web\": 0.05\n \
    \   },\n    \"cart_abandonment\": {\n        \"rate\": 0.52,  # Higher than normal\n\
    \        \"avg_cart_value\": random.randint(450, 650),\n        \"top_abandoned_products\"\
    : [\n            \"Metformin 500mg\",\n            \"Amlodipine 5mg\",\n     \
    \       \"Paracetamol 650mg\"\n        ]\n    },\n    \"session_metrics\": {\n\
    \        \"avg_duration_seconds\": 245,\n        \"pages_per_session\": 4.2,\n\
    \        \"bounce_rate\": 0.32\n    }\n}\n\nprint(\"\u2713 CleverTap data retrieved\
    \ successfully\")\nprint(f\"  - Total events: {sum(clevertap_data['funnel_data'].values()):,}\"\
    )\nprint(f\"  - Cart abandonment rate: {clevertap_data['cart_abandonment']['rate']:.1%}\"\
    )\nprint(f\"  - Chronic patient conversion: {clevertap_data['user_segments']['chronic_patients']['conversion_rate']:.1%}\"\
    )\n\nprint(f\"__OUTPUTS__ {json.dumps(clevertap_data)}\")\n"
  parallel: true
  depends_on:
  - detect_anomaly
  description: Step 2A - Fetch user behavior data from CleverTap (parallel)
  timeout_seconds: 45
- id: fetch_internal_data
  name: Fetch Internal Operational Data
  type: script
  script: "import json\nimport time\nimport random\nfrom datetime import datetime\n\
    \nprint(\"\")\nprint(\"\U0001F3E2 FETCHING INTERNAL DASHBOARD DATA\")\nprint(\"\
    -\" * 40)\n\nanomaly_data = json.loads('${detect_anomaly}')\nregion = anomaly_data['region']\n\
    category = anomaly_data['category']\n\nprint(f\"\U0001F50D Querying internal systems\
    \ for {region} - {category}\")\nprint(\"   Systems: Inventory, Payment, Logistics,\
    \ Pricing\")\n\n# Simulate API call delay\ntime.sleep(1)\n\n# Generate operational\
    \ data that could explain the sales drop\ninternal_data = {\n    \"query_params\"\
    : {\n        \"region\": region,\n        \"category\": category,\n        \"\
    timestamp\": datetime.now().isoformat()\n    },\n    \"inventory_status\": {\n\
    \        \"total_skus\": 2500,\n        \"in_stock_skus\": 2100,\n        \"stockout_skus\"\
    : 400,  # Higher than normal\n        \"critical_stockouts\": 85,  # Key products\
    \ out of stock\n        \"top_stockout_products\": [\n            {\"name\": \"\
    Metformin 500mg\", \"demand\": \"HIGH\", \"days_out\": 3},\n            {\"name\"\
    : \"Insulin Glargine\", \"demand\": \"CRITICAL\", \"days_out\": 2},\n        \
    \    {\"name\": \"Losartan 50mg\", \"demand\": \"HIGH\", \"days_out\": 1}\n  \
    \      ],\n        \"stockout_impact_percentage\": 18.5\n    },\n    \"payment_gateway_status\"\
    : {\n        \"success_rate\": 0.89,  # Lower than normal 95%\n        \"failed_transactions\"\
    : 450,\n        \"gateway_errors\": {\n            \"timeout\": 180,\n       \
    \     \"authentication_failed\": 120,\n            \"insufficient_funds\": 150\n\
    \        },\n        \"affected_payment_methods\": [\"credit_card\", \"debit_card\"\
    ],\n        \"avg_processing_time_ms\": 4500  # Higher than normal\n    },\n \
    \   \"logistics_metrics\": {\n        \"on_time_delivery_rate\": 0.82,  # Lower\
    \ than normal\n        \"avg_delivery_time_hours\": 36,\n        \"delayed_orders\"\
    : 850,\n        \"delivery_partner_issues\": {\n            \"partner_1\": {\"\
    name\": \"BlueDart\", \"performance\": 0.78},\n            \"partner_2\": {\"\
    name\": \"Delhivery\", \"performance\": 0.85}\n        }\n    },\n    \"pricing_data\"\
    : {\n        \"recent_price_changes\": 12,\n        \"avg_price_increase_percentage\"\
    : 8.5,\n        \"competitor_price_comparison\": {\n            \"status\": \"\
    HIGHER\",\n            \"difference_percentage\": 5.2\n        }\n    },\n   \
    \ \"customer_service\": {\n        \"complaint_volume_increase\": 35,  # 35% increase\n\
    \        \"top_complaints\": [\n            \"Product not available\",\n     \
    \       \"Payment failure\",\n            \"Delayed delivery\"\n        ],\n \
    \       \"resolution_rate\": 0.72\n    }\n}\n\nprint(\"\u2713 Internal data retrieved\
    \ successfully\")\nprint(f\"  - Stockout SKUs: {internal_data['inventory_status']['stockout_skus']}\"\
    )\nprint(f\"  - Payment success rate: {internal_data['payment_gateway_status']['success_rate']:.1%}\"\
    )\nprint(f\"  - On-time delivery: {internal_data['logistics_metrics']['on_time_delivery_rate']:.1%}\"\
    )\n\nprint(f\"__OUTPUTS__ {json.dumps(internal_data)}\")\n"
  parallel: true
  depends_on:
  - detect_anomaly
  description: Step 2B - Fetch operational data from internal systems (parallel)
  timeout_seconds: 45
- id: analyze_root_cause
  name: Multi-Hypothesis Root Cause Analysis
  type: script
  script: "import json\nfrom datetime import datetime\n\nprint(\"\")\nprint(\"\U0001F52C\
    \ MULTI-HYPOTHESIS ROOT CAUSE ANALYSIS\")\nprint(\"=\" * 50)\nprint(\"\")\n\n\
    # Parse all data sources\nanomaly_data = json.loads('${detect_anomaly}')\nclevertap_data\
    \ = json.loads('${fetch_clevertap_data}')\ninternal_data = json.loads('${fetch_internal_data}')\n\
    \nprint(\"\U0001F4CA Analyzing data from multiple sources...\")\nprint(f\"   Sales\
    \ Drop: {anomaly_data['actual_drop']:.1f}%\")\nprint(f\"   Data Sources: CleverTap\
    \ + Internal Systems\")\nprint(\"\")\n\n# Define and test hypotheses\nhypotheses\
    \ = {\n    \"inventory_stockouts\": {\n        \"description\": \"Critical product\
    \ stockouts causing lost sales\",\n        \"evidence\": [],\n        \"score\"\
    : 0.0,\n        \"impact\": \"HIGH\"\n    },\n    \"payment_gateway_issues\":\
    \ {\n        \"description\": \"Payment failures preventing order completion\"\
    ,\n        \"evidence\": [],\n        \"score\": 0.0,\n        \"impact\": \"\
    MEDIUM\"\n    },\n    \"user_experience_friction\": {\n        \"description\"\
    : \"High cart abandonment due to UX issues\",\n        \"evidence\": [],\n   \
    \     \"score\": 0.0,\n        \"impact\": \"MEDIUM\"\n    },\n    \"delivery_delays\"\
    : {\n        \"description\": \"Poor delivery performance affecting repeat purchases\"\
    ,\n        \"evidence\": [],\n        \"score\": 0.0,\n        \"impact\": \"\
    MEDIUM\"\n    },\n    \"price_sensitivity\": {\n        \"description\": \"Recent\
    \ price increases causing customer churn\",\n        \"evidence\": [],\n     \
    \   \"score\": 0.0,\n        \"impact\": \"LOW\"\n    }\n}\n\n# Test Hypothesis\
    \ 1: Inventory Stockouts\nstockout_impact = internal_data['inventory_status']['stockout_impact_percentage']\n\
    if stockout_impact > 15:\n    hypotheses[\"inventory_stockouts\"][\"score\"] =\
    \ 0.85\n    hypotheses[\"inventory_stockouts\"][\"evidence\"].append(\n      \
    \  f\"Stockout impact: {stockout_impact:.1f}% of potential sales\"\n    )\n  \
    \  hypotheses[\"inventory_stockouts\"][\"evidence\"].append(\n        f\"Critical\
    \ items out of stock: {internal_data['inventory_status']['critical_stockouts']}\"\
    \n    )\n\n# Test Hypothesis 2: Payment Gateway Issues\npayment_success = internal_data['payment_gateway_status']['success_rate']\n\
    if payment_success < 0.92:\n    hypotheses[\"payment_gateway_issues\"][\"score\"\
    ] = 0.70\n    hypotheses[\"payment_gateway_issues\"][\"evidence\"].append(\n \
    \       f\"Payment success rate: {payment_success:.1%} (below 92% threshold)\"\
    \n    )\n    hypotheses[\"payment_gateway_issues\"][\"evidence\"].append(\n  \
    \      f\"Failed transactions: {internal_data['payment_gateway_status']['failed_transactions']}\"\
    \n    )\n\n# Test Hypothesis 3: User Experience Friction\ncart_abandonment = clevertap_data['cart_abandonment']['rate']\n\
    checkout_conversion = clevertap_data['conversion_rates']['checkout_to_purchase']\n\
    if cart_abandonment > 0.45 and checkout_conversion < 0.70:\n    hypotheses[\"\
    user_experience_friction\"][\"score\"] = 0.75\n    hypotheses[\"user_experience_friction\"\
    ][\"evidence\"].append(\n        f\"Cart abandonment rate: {cart_abandonment:.1%}\"\
    \n    )\n    hypotheses[\"user_experience_friction\"][\"evidence\"].append(\n\
    \        f\"Low checkout conversion: {checkout_conversion:.1%}\"\n    )\n\n# Test\
    \ Hypothesis 4: Delivery Delays\non_time_delivery = internal_data['logistics_metrics']['on_time_delivery_rate']\n\
    if on_time_delivery < 0.85:\n    hypotheses[\"delivery_delays\"][\"score\"] =\
    \ 0.60\n    hypotheses[\"delivery_delays\"][\"evidence\"].append(\n        f\"\
    On-time delivery: {on_time_delivery:.1%}\"\n    )\n    hypotheses[\"delivery_delays\"\
    ][\"evidence\"].append(\n        f\"Delayed orders: {internal_data['logistics_metrics']['delayed_orders']}\"\
    \n    )\n\n# Test Hypothesis 5: Price Sensitivity\nif internal_data['pricing_data']['avg_price_increase_percentage']\
    \ > 5:\n    hypotheses[\"price_sensitivity\"][\"score\"] = 0.45\n    hypotheses[\"\
    price_sensitivity\"][\"evidence\"].append(\n        f\"Recent price increase:\
    \ {internal_data['pricing_data']['avg_price_increase_percentage']:.1f}%\"\n  \
    \  )\n\n# Rank hypotheses by score\nranked_hypotheses = sorted(\n    hypotheses.items(),\
    \ \n    key=lambda x: x[1]['score'], \n    reverse=True\n)\n\nprint(\"\U0001F3AF\
    \ Hypothesis Testing Results:\")\nprint(\"\")\nfor rank, (hypothesis, data) in\
    \ enumerate(ranked_hypotheses[:3], 1):\n    print(f\"{rank}. {hypothesis.replace('_',\
    \ ' ').title()}\")\n    print(f\"   Score: {data['score']:.2f}\")\n    print(f\"\
    \   Impact: {data['impact']}\")\n    for evidence in data['evidence']:\n     \
    \   print(f\"   - {evidence}\")\n    print(\"\")\n\n# Determine primary root cause\n\
    primary_cause = ranked_hypotheses[0][0]\nconfidence = \"HIGH\" if ranked_hypotheses[0][1]['score']\
    \ > 0.7 else \"MEDIUM\"\n\n# Generate recommendations\nrecommendations = []\n\
    if primary_cause == \"inventory_stockouts\":\n    recommendations = [\n      \
    \  \"Immediately notify affected customers about stockouts\",\n        \"Activate\
    \ substitute product recommendations\",\n        \"Expedite replenishment for\
    \ critical SKUs\",\n        \"Launch 'back in stock' notification campaign\"\n\
    \    ]\nelif primary_cause == \"payment_gateway_issues\":\n    recommendations\
    \ = [\n        \"Switch to backup payment gateway\",\n        \"Promote COD as\
    \ alternative payment method\",\n        \"Send payment retry notifications\"\
    ,\n        \"Offer payment assistance through customer support\"\n    ]\nelif\
    \ primary_cause == \"user_experience_friction\":\n    recommendations = [\n  \
    \      \"Simplify checkout process immediately\",\n        \"Launch cart recovery\
    \ campaign with incentives\",\n        \"A/B test one-click checkout\",\n    \
    \    \"Reduce form fields in checkout\"\n    ]\n\nroot_cause_result = {\n    \"\
    timestamp\": datetime.now().isoformat(),\n    \"hypotheses_tested\": len(hypotheses),\n\
    \    \"hypothesis_results\": dict(ranked_hypotheses),\n    \"primary_root_cause\"\
    : primary_cause,\n    \"confidence_level\": confidence,\n    \"contributing_factors\"\
    : [h[0] for h in ranked_hypotheses if h[1]['score'] > 0.5],\n    \"recommendations\"\
    : recommendations,\n    \"estimated_recovery_time\": \"4-6 hours\" if primary_cause\
    \ in [\"payment_gateway_issues\", \"user_experience_friction\"] else \"24-48 hours\"\
    \n}\n\nprint(\"\u2705 Root cause analysis complete.\")\nprint(f\"   Primary Cause:\
    \ {primary_cause.replace('_', ' ').title()}\")\nprint(f\"   Confidence: {confidence}\"\
    )\n\nprint(f\"__OUTPUTS__ {json.dumps(root_cause_result)}\")\n"
  depends_on:
  - fetch_clevertap_data
  - fetch_internal_data
  description: Step 3 - Analyze root cause using multi-hypothesis testing
  timeout_seconds: 60
- id: generate_alert
  name: Generate Alert Notification
  type: ai_agent
  config:
    model_client_id: alert_generator
  depends_on:
  - analyze_root_cause
  description: Step 4 - Generate intelligent alerts for stakeholders
  user_message: 'Generate alert for this sales drop incident:


    Anomaly Details:

    - Region: ${detect_anomaly.region}

    - Category: ${detect_anomaly.category}

    - Sales Drop: ${detect_anomaly.actual_drop}%

    - Severity: ${detect_anomaly.severity}


    Root Cause Analysis:

    - Primary Cause: ${analyze_root_cause.primary_root_cause}

    - Confidence: ${analyze_root_cause.confidence_level}

    - Contributing Factors: ${analyze_root_cause.contributing_factors}


    Create a clear, actionable alert for immediate distribution.

    '
  system_message: "You are an expert at creating concise, actionable alerts for business\
    \ stakeholders.\n\nGenerate an alert notification in JSON format:\n{\n  \"alert_title\"\
    : \"Clear, urgent title\",\n  \"severity\": \"CRITICAL|HIGH|MEDIUM\",\n  \"summary\"\
    : \"2-3 sentence executive summary\",\n  \"key_metrics\": {\n    \"sales_impact\"\
    : \"percentage and amount\",\n    \"affected_users\": \"number\",\n    \"recovery_eta\"\
    : \"timeframe\"\n  },\n  \"root_cause_summary\": \"One sentence explanation\"\
    ,\n  \"immediate_actions\": [\"action1\", \"action2\", \"action3\"],\n  \"stakeholders_to_notify\"\
    : [\"role1\", \"role2\"],\n  \"escalation_required\": true/false\n}\n"
- id: trigger_recovery_campaign
  name: Execute Recovery Campaign
  type: script
  script: "import json\nimport time\nimport random\nfrom datetime import datetime\n\
    \nprint(\"\")\nprint(\"\U0001F680 EXECUTING RECOVERY CAMPAIGNS\")\nprint(\"=\"\
    \ * 50)\nprint(\"\")\n\n# Parse previous results\nroot_cause_data = json.loads('${analyze_root_cause}')\n\
    alert_data = json.loads('${generate_alert}')\nprimary_cause = root_cause_data['primary_root_cause']\n\
    \nprint(f\"\U0001F4E2 Launching targeted campaigns for: {primary_cause.replace('_',\
    \ ' ').title()}\")\nprint(\"\")\n\ncampaigns_executed = []\n\nif primary_cause\
    \ == \"inventory_stockouts\":\n    print(\"\U0001F514 Campaign 1: Stockout Notification\
    \ & Alternatives\")\n    campaign1 = {\n        \"name\": \"Critical Stockout\
    \ Alert Campaign\",\n        \"type\": \"multi_channel\",\n        \"channels\"\
    : [\"push_notification\", \"in_app\", \"email\"],\n        \"segment\": \"users_with_oos_items_in_cart\"\
    ,\n        \"message_template\": \"Important: {product} is currently out of stock.\
    \ Similar alternatives available!\",\n        \"cta\": \"View Alternatives\",\n\
    \        \"personalization\": \"product_based\"\n    }\n    \n    # Simulate campaign\
    \ execution\n    time.sleep(0.5)\n    campaign1[\"execution_stats\"] = {\n   \
    \     \"targeted_users\": random.randint(3000, 5000),\n        \"messages_sent\"\
    : random.randint(2800, 4800),\n        \"opens\": random.randint(1500, 2500),\n\
    \        \"clicks\": random.randint(800, 1200),\n        \"conversions\": random.randint(400,\
    \ 600)\n    }\n    campaigns_executed.append(campaign1)\n    print(\"   \u2713\
    \ Notifications sent to affected users\")\n    \n    print(\"\")\n    print(\"\
    \U0001F48A Campaign 2: Substitute Product Recommendations\")\n    campaign2 =\
    \ {\n        \"name\": \"Smart Alternative Suggestions\",\n        \"type\": \"\
    personalized_recommendations\",\n        \"algorithm\": \"collaborative_filtering\"\
    ,\n        \"display_locations\": [\"home_page\", \"search_results\", \"cart_page\"\
    ]\n    }\n    campaign2[\"execution_stats\"] = {\n        \"recommendations_shown\"\
    : random.randint(8000, 12000),\n        \"clicks\": random.randint(2000, 3000),\n\
    \        \"add_to_cart\": random.randint(800, 1200),\n        \"purchases\": random.randint(400,\
    \ 700)\n    }\n    campaigns_executed.append(campaign2)\n    print(\"   \u2713\
    \ Alternative products activated\")\n    \nelif primary_cause == \"payment_gateway_issues\"\
    :\n    print(\"\U0001F4B3 Campaign 1: Payment Retry Assistance\")\n    campaign1\
    \ = {\n        \"name\": \"Payment Recovery Campaign\",\n        \"type\": \"\
    targeted_sms_email\",\n        \"segment\": \"failed_payment_last_24h\",\n   \
    \     \"offers\": [\"COD available\", \"Try alternate payment method\", \"10%\
    \ off on retry\"]\n    }\n    campaign1[\"execution_stats\"] = {\n        \"targeted_users\"\
    : random.randint(400, 600),\n        \"messages_sent\": random.randint(380, 580),\n\
    \        \"payment_retries\": random.randint(150, 250),\n        \"successful_orders\"\
    : random.randint(100, 180)\n    }\n    campaigns_executed.append(campaign1)\n\
    \    print(\"   \u2713 Payment assistance campaign launched\")\n    \nelif primary_cause\
    \ == \"user_experience_friction\":\n    print(\"\U0001F6D2 Campaign 1: Cart Recovery\
    \ with Incentives\")\n    campaign1 = {\n        \"name\": \"Abandoned Cart Win-back\"\
    ,\n        \"type\": \"progressive_discount\",\n        \"discount_tiers\": [\"\
    5% after 2 hours\", \"10% after 6 hours\", \"15% after 24 hours\"],\n        \"\
    segment\": \"cart_abandoners\",\n        \"urgency_messaging\": True\n    }\n\
    \    campaign1[\"execution_stats\"] = {\n        \"targeted_users\": random.randint(5000,\
    \ 8000),\n        \"emails_sent\": random.randint(4800, 7800),\n        \"cart_recoveries\"\
    : random.randint(800, 1200),\n        \"revenue_recovered\": random.randint(400000,\
    \ 600000)\n    }\n    campaigns_executed.append(campaign1)\n    print(\"   \u2713\
    \ Cart recovery campaign activated\")\n\n# Calculate total impact\ntotal_conversions\
    \ = sum(\n    c[\"execution_stats\"].get(\"conversions\", 0) + \n    c[\"execution_stats\"\
    ].get(\"successful_orders\", 0) + \n    c[\"execution_stats\"].get(\"purchases\"\
    , 0) \n    for c in campaigns_executed\n)\n\nestimated_revenue_recovery = total_conversions\
    \ * random.randint(500, 800)\n\nprint(\"\")\nprint(\"\U0001F4CA Campaign Execution\
    \ Summary:\")\nprint(f\"   Campaigns Launched: {len(campaigns_executed)}\")\n\
    print(f\"   Total Conversions: {total_conversions:,}\")\nprint(f\"   Estimated\
    \ Revenue Recovery: \u20B9{estimated_revenue_recovery:,}\")\nprint(f\"   Recovery\
    \ ETA: {root_cause_data['estimated_recovery_time']}\")\n\nrecovery_result = {\n\
    \    \"timestamp\": datetime.now().isoformat(),\n    \"primary_intervention\"\
    : primary_cause,\n    \"campaigns_executed\": campaigns_executed,\n    \"total_campaigns\"\
    : len(campaigns_executed),\n    \"total_conversions\": total_conversions,\n  \
    \  \"estimated_revenue_recovery\": estimated_revenue_recovery,\n    \"recovery_status\"\
    : \"IN_PROGRESS\",\n    \"next_checkpoint\": \"2_hours\",\n    \"success_metrics\"\
    : {\n        \"target_recovery_percentage\": 70,\n        \"current_recovery_percentage\"\
    : random.randint(25, 35),\n        \"confidence_interval\": \"\xB15%\"\n    }\n\
    }\n\nprint(\"\")\nprint(\"\u2705 Recovery campaigns executed successfully.\")\n\
    print(\"   Monitoring performance and ready for optimization...\")\n\nprint(f\"\
    __OUTPUTS__ {json.dumps(recovery_result)}\")\n"
  depends_on:
  - generate_alert
  description: Step 5 - Trigger automated recovery campaigns
  timeout_seconds: 90
inputs:
- name: clevertap_account_id
  type: string
  required: true
  description: CleverTap Account ID
- name: clevertap_passcode
  type: string
  required: true
  description: CleverTap API Passcode
- name: internal_api_key
  type: string
  required: true
  description: Internal Dashboard API Key
- name: sales_drop_percentage
  type: float
  required: true
  description: Percentage drop in sales that triggered this workflow
- name: affected_region
  type: string
  required: true
  description: Region experiencing the sales drop
- name: affected_category
  type: string
  required: true
  description: Category with sales drop (medicines, diagnostics, wellness)
- name: lookback_hours
  type: integer
  default: 24
  description: Hours to look back for analysis
outputs:
  internal_data:
    type: object
    source: fetch_internal_data
    description: Operational data from internal systems
  sales_anomaly:
    type: object
    source: detect_anomaly
    description: Detected sales anomaly details
  clevertap_data:
    type: object
    source: fetch_clevertap_data
    description: User behavior data from CleverTap
  recovery_campaign:
    type: object
    source: trigger_recovery_campaign
    description: Recovery campaign execution results
  alert_notification:
    type: object
    source: generate_alert
    description: Alert generated for stakeholders
  root_cause_analysis:
    type: object
    source: analyze_root_cause
    description: Multi-hypothesis root cause analysis results
version: 1.0
namespace: pharmeasy
description: Pharmeasy Sales Drop Diagnostic Workflow - Automated root cause analysis
  and recovery
model_clients:
- id: diagnostic_analyst
  config:
    model: gpt-4o-mini
    api_key: ${OPENAI_API_KEY}
    max_tokens: 2000
    temperature: 0.1
  provider: openai
- id: alert_generator
  config:
    model: gpt-4o-mini
    api_key: ${OPENAI_API_KEY}
    max_tokens: 1500
    temperature: 0.3
  provider: openai
timeout_seconds: 300

No executions yet. Execute this workflow to see results here.