crm-data-cleaner

name	crm-data-cleaner
description	CRM veri temizleme. Duplicate detection, data standardization ve validation rules.
triggers	{"keywords":["CRM cleaning","veri temizleme","duplicate detection","data deduplication","data standardization","CRM hygiene"]}
auto_load_when	Kullanıcı CRM veri kalitesi, duplicate temizleme, veri standardizasyonu veya validation kuralları talep ettiğinde
agent	researcher
tools	["Read","Write","Grep","Glob"]

CRM Data Cleaner (CRM Veri Temizleme Uzmanı)

Odak Alanı: CRM sistemlerindeki veri kalitesini artırmak, duplicate'ları tespit etmek, verileri standardize etmek ve validation kuralları oluşturmak.

---

Pattern 1: Duplicate Detection (Tekrar Tespiti)

1.1 Duplicate Türleri ve Tespit Yöntemleri

Duplicate Categories:
├── Hard Duplicates (Exact Match)
│   ├── Identical email address
│   ├── Identical company name
│   └── Identical phone number
│
├── Soft Duplicates (Fuzzy Match)
│   ├── Similar company names (typos, variations)
│   ├── Similar names (first/last name variations)
│   └── Similar emails (domain variations)
│
├── Partial Duplicates
│   ├── Same company, multiple contacts
│   ├── Same person, multiple records
│   └── Same lead, different sources
│
└── Phantom Duplicates
    ├── Same real entity, different spelling
    ├── Acquired company variations
    └── International character variations

1.2 Duplicate Detection Tree

Detection Algorithm:
├── Exact Match Rules
│   ├── Email = [exact match] → Mark as duplicate
│   ├── Phone = [exact match] → Flag for review
│   └── Company + Name = [exact] → Hard merge candidate
│
├── Fuzzy Match Rules
│   ├── Company name >80% similarity → Manual review
│   ├── Contact name >85% similarity → Review required
│   ├── Email domain match + name similarity → Flag
│   └── Address >70% match → Merge candidate
│
└── Pattern-Based Rules
    ├── Multiple records from same domain (5+) → Company dedup
    ├── Same email different timestamps → Merge
    └── Multiple records with no email → Review needed

1.3 Duplicate Resolution Framework

Merge Decision Tree:
├── Auto-Merge (Safe)
│   ├── All fields match exactly
│   ├── No conflicting data
│   └── Same source attribution
│
├── Auto-Merge (Smart)
│   ├── Non-conflicting fields combine
│   ├── Most recent data wins
│   ├── More complete record wins
│   └── Activity history merges
│
├── Manual Review
│   ├── Conflicting critical fields
│   ├── Same email different names
│   └── Similar but uncertain matches
│
└── Don't Merge
    ├── Different companies (similar names)
    ├── Different people (similar names)
    └── Same person, different companies

---

Pattern 2: Data Standardization (Veri Standardizasyonu)

2.1 Company Name Standardization

Company Name Rules:
├── Standardization Steps
│   ├── Remove legal entities: Inc, LLC, Ltd, Corp, GmbH
│   ├── Standardize case: Title Case
│   ├── Remove punctuation: & → and, - → space
│   ├── Fix spacing: Double spaces → single
│   └── Handle special chars: ä→a, ö→o, etc.
│
├── Examples
│   ├── "ABC Corporation, Inc." → "ABC Corporation"
│   ├── "Smith & Jones LLC" → "Smith and Jones"
│   ├── "XYZ-Corp" → "XYZ Corp"
│   └── "Müller GmbH" → "Muller"
│
└── Special Cases
    ├── "Google Inc" → Keep "Google" as standard
    ├── "Microsoft Corporation" → Keep full
    └── "The Walt Disney Company" → Keep full

2.2 Contact Data Standardization

Name Standardization:
├── First Name
│   ├── All lowercase → Title Case
│   ├── Remove prefixes: Mr., Ms., Dr.
│   ├── Handle nicknames: "Bill" → "William"
│   └── Standardize international: "Jean-Pierre" → "Jean Pierre"
│
├── Last Name
│   ├── Handle suffixes: Jr., Sr., III
│   ├── Handle prefixes: von, van, de
│   └── Keep compound names intact
│
└── Email Address
    ├── All lowercase
    ├── Remove . in Gmail (john.smith = johnsmith)
    └── Handle + addressing (keep as-is)

2.3 Industry ve Title Standardization

Industry Classification:
├── Standard Industry List
│   ├── Technology → SaaS, Hardware, IT Services
│   ├── Finance → Banking, Insurance, Fintech
│   ├── Healthcare → Provider, Pharma, Biotech
│   ├── Manufacturing → Industrial, Automotive
│   └── Retail → E-commerce, Brick & Mortar
│
├── Job Title Standardization
│   ├── C-Level → CEO, CFO, CTO, COO, CMO
│   ├── VP-Level → VP, SVP, EVP
│   ├── Director → Director, Sr. Director
│   ├── Manager → Manager, Senior Manager
│   └── IC → Individual Contributor titles
│
└── Level Mapping
    ├── Decision Maker → C-Level, VP, Director
    ├── Influencer → Manager, Senior IC
    └── User → IC, Specialist

---

Pattern 3: Validation Rules (Doğrulama Kuralları)

3.1 Field Validation Matrix

Required Fields (Minimum):
├── Contact
│   ├── Email (valid format + deliverable)
│   ├── First name (required)
│   └── Last name (required)
│
├── Company
│   ├── Company name (required)
│   ├── Industry (required)
│   └── Company size (required if available)
│
└── Lead
    ├── Lead source (required)
    ├── Status (required)
    └── Created date (auto)

3.2 Data Quality Rules

Validation Check List:
├── Email Validation
│   ├── Format check: @domain.extension
│   ├── Domain exists: DNS lookup
│   ├── Disposable email: Block known disposables
│   ├── Role-based: catch-all, info@, support@
│   └── Role check: Ceo, admin, sales@ (flag)
│
├── Phone Validation
│   ├── Format: +[country][area][number]
│   ├── Country code presence
│   ├── Valid digit count
│   └── Phone carrier lookup (optional)
│
├── Company Validation
│   ├── Employee count reasonable
│   ├── Industry code valid
│   ├── Website resolves (optional)
│   └── LinkedIn company exists (optional)
│
└── Address Validation
    ├── Postal code format
    ├── City-State match
    └── Country format (ISO code)

3.3 Data Quality Scoring

Record Quality Score (/100):
├── Completeness (30 points)
│   ├── All required fields: +15
│   ├── Optional fields filled: +10
│   └── Custom fields filled: +5
│
├── Validity (30 points)
│   ├── Email valid: +10
│   ├── Phone valid: +5
│   ├── Company verified: +10
│   └── Address valid: +5
│
├── Freshness (20 points)
│   ├── Updated <30 days: +10
│   ├── Updated <90 days: +5
│   └── Activity <180 days: +5
│
└── Uniqueness (20 points)
    ├── No duplicates: +10
    ├── Single record: +5
    └── De-duped: +5

---

Pattern 4: Data Cleaning Workflow

4.1 Cleaning Process Steps

Standard Cleaning Workflow:
├── Step 1: Assessment
│   ├── Total records count
│   ├── Duplicate percentage
│   ├── Missing data by field
│   ├── Stale records percentage
│   └── Data source quality
│
├── Step 2: Quick Wins
│   ├── Remove obvious duplicates
│   ├── Fix formatting errors
│   ├── Fill obvious missing data
│   └── Standardize obvious variations
│
├── Step 3: Deep Clean
│   ├── Fuzzy matching and resolution
│   ├── Data enrichment
│   ├── Validation fixes
│   └── Quality scoring
│
├── Step 4: Prevention
│   ├── Set up validation rules
│   ├── Create dedup workflows
│   ├── Establish data entry standards
│   └── Set up regular cleaning schedule
│
└── Step 5: Ongoing Maintenance
    ├── Weekly duplicate review
    ├── Monthly quality check
    ├── Quarterly deep clean
    └── Annual complete audit

4.2 Tool Integration

Cleaning Tools by Platform:
├── HubSpot
│   ├── Native duplicate detection
│   ├── Property validation rules
│   ├── Workflow-based cleaning
│   └── Data quality dashboard
│
├── Salesforce
│   ├── Duplicate rules (matching rules)
│   ├── Data.com enrichment
│   ├── Validation rules
│   └── Duplicate job scheduler
│
├── Pipedrive
│   ├── Manual duplicate merge
│   ├── Custom field validation
│   └── Export/import cleaning
│
└── Custom/Other
    ├── External tools: ZoomInfo, Clearbit
    ├── Dedupe tools: Dedupely, FullEnrich
    └── Cleaning services: CloudSponge

---

Pattern 5: Quality Monitoring

5.1 Quality Dashboard Metrics

Key Metrics to Track:
├── Volume Metrics
│   ├── Total contacts
│   ├── Total companies
│   ├── Active contacts
│   └── Leads by source
│
├── Quality Metrics
│   ├── Duplicate percentage (target <5%)
│   ├── Complete records % (target >80%)
│   ├── Valid email % (target >90%)
│   └── Freshness score
│
├── Activity Metrics
│   ├── Records updated (30 days)
│   ├── New records by source
│   ├── Stale records (>180 days no activity)
│   └── Bounce rate
│
└── ROI Metrics
    ├── Clean data impact on conversion
    ├── Time saved by sales team
    └── Unqualified lead reduction

5.2 Automated Maintenance

Automation Rules:
├── Duplicate Prevention
│   ├── Block submission if email exists
│   ├── Auto-merge safe duplicates
│   ├── Alert on potential duplicates
│   └── Block import of known duplicates
│
├── Data Entry Standards
│   ├── Required field enforcement
│   ├── Dropdown for standard fields
│   ├── Auto-format on entry
│   └── Auto-capitalize where appropriate
│
├── Stale Data Handling
│   ├── Auto-archive after X months inactive
│   ├── Periodic re-validation emails
│   ├── Re-engagement campaigns
│   └── Lead scoring decay
│
└── Quality Alerts
│   ├── Alert on mass import
│   ├── Alert on quality drop
│   ├── Alert on duplicate spike
│   └── Alert on missing required data

---

Key Patterns (Özet)

Pattern	Odak	Uygulama
Duplicate Detection	Tespit	Exact + fuzzy matching
Data Standardization	Format	Names, emails, companies
Validation Rules	Doğrulama	Format + completeness
Cleaning Workflow	Süreç	Assessment → Clean → Prevent → Monitor
Quality Monitoring	Ölçüm	Dashboard + automated alerts

---

Anti-Patterns

❌ Yasaklı Yaklaşımlar

Process errors:
  - Manual-only cleaning (no automation)
  - No duplicate prevention
  - No ongoing maintenance
  - Ignoring data quality
  
Technical errors:
  - Over-aggressive merging
  - Loss of historical data
  - Deleting instead of archiving
  - Breaking related records

✅ Doğru Yaklaşımlar

Best practices:
  - Always archive before delete
  - Test rules on small batch first
  - Document all changes
  - Regular maintenance schedule
  
Prevention:
  - Validation at entry point
  - Duplicate alerts for sales
  - Source data quality requirements
  - Regular quality audits

---

Quick Reference

Metric	Target	Warning
Duplicate rate	<5%	>10%
Complete records	>80%	<60%
Valid email rate	>90%	<70%
Stale records	<20%	>40%

Validation	Rule	Action
Email	Format + domain	Reject invalid
Phone	International format	Standardize
Company	Required field	Block entry if missing
Name	Not empty	Reject if blank

Schedule	Task	Frequency
Duplicate check	Scan + merge	Weekly
Data validation	Run rules	Real-time
Quality audit	Full assessment	Monthly
Deep clean	Complete refresh	Quarterly

🌍 Universal Language Support

• Turkish Native: This skill natively supports Turkish. If the user prompt is in Turkish, all analysis, formatting, and output MUST be entirely in Turkish. You do not need explicit "write in Turkish" instructions.