crm-hygiene
关于
This skill audits CRM data quality by identifying issues like missing fields, duplicates, and stale records. It's designed for use during quarterly reviews, before reporting periods, or when troubleshooting automation failures. The tool provides structured analysis based on completeness, consistency, accuracy, and timeliness frameworks.
快速安装
Claude Code
推荐npx skills add guia-matthieu/clawfu-skills -a claude-code/plugin add https://github.com/guia-matthieu/clawfu-skillsgit clone https://github.com/guia-matthieu/clawfu-skills.git ~/.claude/skills/crm-hygiene在 Claude Code 中复制并粘贴此命令以安装该技能
技能文档
CRM Hygiene
Systematically identify and resolve data quality issues in your CRM to improve forecast accuracy, reporting, and automation reliability.
When to Use This Skill
- Quarterly CRM audits
- Before major reporting periods
- When automation workflows fail
- After team changes or migrations
- Preparing for CRM integrations
Methodology Foundation
Based on Salesforce Data Quality Best Practices and RevOps Co-op Data Governance frameworks, focusing on:
- Completeness (required fields populated)
- Consistency (standardized values)
- Accuracy (correct and current data)
- Timeliness (recent updates)
What Claude Does vs What You Decide
| Claude Does | You Decide |
|---|---|
| Defines audit criteria | Which fields are required |
| Identifies data issues | Priority of fixes |
| Categorizes problems by type | Who owns cleanup tasks |
| Suggests remediation steps | Automation vs manual fix |
| Creates cleanup reports | Enforcement policies |
What This Skill Does
- Audit design - Define required fields and validation rules
- Issue identification - Find missing, inconsistent, duplicate data
- Prioritization - Rank issues by business impact
- Remediation planning - Action steps to fix issues
- Governance recommendations - Prevent future issues
How to Use
For Audit Design:
Help me create a CRM hygiene audit for [CRM System].
Key objects to audit:
- [Leads, Contacts, Accounts, Opportunities]
Critical fields for each:
- Leads: [list required fields]
- Opportunities: [list required fields]
Our main issues:
- [Describe known problems]
For Data Audit:
Audit this opportunity data for hygiene issues:
[Paste export or describe data]
Required fields:
- [List fields that must be populated]
Validation rules:
- [List rules like "Stage must match Amount > $0"]
Instructions
Step 1: Define Audit Scope
Object Priority (typical):
- Opportunities - Revenue impact
- Accounts - Relationship foundation
- Contacts - Communication accuracy
- Leads - Pipeline source
Field Categories:
- Critical - Blocks processes if missing
- Important - Affects reporting/segmentation
- Nice-to-Have - Enriches but not required
Step 2: Completeness Audit
Check required fields by object:
Opportunities:
| Field | Criticality | Why Required |
|---|---|---|
| Amount | Critical | Forecast, pipeline value |
| Close Date | Critical | Forecast timing |
| Stage | Critical | Pipeline reporting |
| Owner | Critical | Accountability |
| Account | Critical | Company linkage |
| Primary Contact | Important | Communication |
| Next Step | Important | Deal momentum |
| Competitor | Important | Win/loss analysis |
| Loss Reason | Critical (if lost) | Improvement insights |
Accounts:
| Field | Criticality | Why Required |
|---|---|---|
| Industry | Important | Segmentation |
| Employee Count | Important | ICP fit |
| Website | Important | Research, enrichment |
| Billing Address | Critical | Invoicing |
| Account Owner | Critical | Accountability |
Step 3: Consistency Audit
Check for standardization issues:
| Issue Type | Example | Impact |
|---|---|---|
| Picklist variations | "US" vs "USA" vs "United States" | Reporting breaks |
| Case inconsistency | "ACME" vs "Acme" vs "acme" | Duplicates |
| Format variations | "(555) 123-4567" vs "5551234567" | Automation fails |
| Null vs blank | NULL vs "" vs "N/A" | Filters don't work |
Step 4: Accuracy Audit
Identify likely incorrect data:
| Check | Criteria | Action |
|---|---|---|
| Stale close dates | Past date, still open | Update or close |
| Amount mismatches | $0 in late stage | Verify pricing |
| Stage regression | Stage 4 → Stage 2 | Investigate |
| Impossible values | 200% probability | Fix validation |
| Bounced emails | Marked invalid | Remove or update |
Step 5: Duplicate Detection
Find potential duplicates:
| Match Type | Criteria | Confidence |
|---|---|---|
| Exact email | Same email address | 99% duplicate |
| Domain match | Same company domain | 80% duplicate |
| Fuzzy name | Similar company name | Investigate |
| Phone match | Same phone number | 90% duplicate |
Step 6: Timeliness Audit
Flag stale records:
| Object | Staleness Threshold | Action |
|---|---|---|
| Open Opportunity | No activity 30 days | Review with rep |
| Lead | No touch 14 days | Reassign or nurture |
| Account | No contact 90 days | Re-engagement |
| Contact | Bounced + 30 days | Archive |
Examples
Example 1: Opportunity Audit
Input:
Audit these opportunities:
1. Acme Corp - Stage: Proposal, Amount: $0, Close: Jan 15 (past), Owner: Empty
2. Beta Inc - Stage: Closed Won, Amount: $50K, Close: Feb 1, Loss Reason: "Price"
3. Gamma LLC - Stage: Discovery, Amount: $100K, Close: Mar 30, No activity 45 days
4. Delta Co - Stage: Negotiation, Amount: $75K, Close: Feb 28, Next Step: Empty
5. Echo Systems - Stage: Closed Lost, Amount: $200K, Loss Reason: Empty
Output:
## CRM Hygiene Audit: Opportunities
### Summary
- **Total Records**: 5
- **Critical Issues**: 4
- **Warnings**: 2
- **Clean Records**: 0
### Critical Issues (Fix Immediately)
#### 1. Acme Corp - 3 Critical Issues
| Issue | Field | Current | Required |
|-------|-------|---------|----------|
| 🔴 Missing | Owner | Empty | Assigned rep |
| 🔴 Missing | Amount | $0 | Deal value |
| 🔴 Stale | Close Date | Jan 15 (past) | Future date |
**Action**: Assign owner, verify if deal is active, update amount and close date.
#### 2. Beta Inc - Logic Error
| Issue | Field | Current | Problem |
|-------|-------|---------|---------|
| 🔴 Inconsistent | Loss Reason | "Price" | Stage is "Closed Won" |
**Action**: Either change stage to Closed Lost, or clear Loss Reason.
#### 3. Echo Systems - Missing Required
| Issue | Field | Current | Required |
|-------|-------|---------|----------|
| 🔴 Missing | Loss Reason | Empty | Required for Closed Lost |
**Action**: Rep must document why deal was lost.
#### 4. Delta Co - Missing Next Step
| Issue | Field | Current | Impact |
|-------|-------|---------|--------|
| 🟠 Missing | Next Step | Empty | No clear action |
**Action**: Rep should define next step for Negotiation stage.
### Warnings (Review This Week)
#### 5. Gamma LLC - Stale Deal
| Issue | Indicator | Value |
|-------|-----------|-------|
| 🟡 Stale | Last Activity | 45 days ago |
**Action**: Review with rep - is this deal still active? If yes, update. If no, close lost.
### Hygiene Score: 20/100 (Poor)
### Recommended Actions by Priority
1. **Today**: Fix Acme Corp (no owner = orphaned deal)
2. **Today**: Fix Beta Inc logic error (affects win rate)
3. **This Week**: Get Echo Systems loss reason
4. **This Week**: Update Gamma LLC status
5. **This Week**: Add Delta Co next step
### Process Recommendations
- Add validation rule: Amount required when Stage > Discovery
- Add validation rule: Owner required on all opportunities
- Add validation rule: Loss Reason required when Closed Lost
- Create stale deal report (no activity 30+ days)
Example 2: Duplicate Detection
Input:
Check these accounts for duplicates:
1. Acme Corporation, acme.com, San Francisco
2. ACME Corp, acme.com, SF
3. Acme Inc, acme-inc.com, San Francisco
4. Beta Technologies, beta.io, New York
5. Beta Tech, betatech.com, NYC
6. Beta Technologies Inc, beta.io, New York
Output:
## Duplicate Detection Report
### High Confidence Duplicates (Merge)
#### Duplicate Set 1: Acme Corporation
| Record | Name | Domain | Location | Confidence |
|--------|------|--------|----------|------------|
| #1 | Acme Corporation | acme.com | San Francisco | Master |
| #2 | ACME Corp | acme.com | SF | 99% - Same domain |
**Recommendation**: Merge #2 into #1
- Same domain (acme.com)
- Same location (SF = San Francisco)
- Name variation only (Corp vs Corporation)
#### Duplicate Set 2: Beta Technologies
| Record | Name | Domain | Location | Confidence |
|--------|------|--------|----------|------------|
| #4 | Beta Technologies | beta.io | New York | Master |
| #6 | Beta Technologies Inc | beta.io | New York | 99% - Same domain |
**Recommendation**: Merge #6 into #4
- Same domain (beta.io)
- Same location (New York)
- Name variation only (Inc suffix)
### Investigate (Possible Duplicates)
#### Possible Set 1: Acme variations
| Record | Name | Domain | Location | Confidence |
|--------|------|--------|----------|------------|
| #1 | Acme Corporation | acme.com | San Francisco | - |
| #3 | Acme Inc | acme-inc.com | San Francisco | 40% |
**Recommendation**: Investigate manually
- Same location, similar name
- BUT different domains (acme.com vs acme-inc.com)
- Could be: subsidiary, different company, or typo
#### Possible Set 2: Beta variations
| Record | Name | Domain | Location | Confidence |
|--------|------|--------|----------|------------|
| #4 | Beta Technologies | beta.io | New York | - |
| #5 | Beta Tech | betatech.com | NYC | 50% |
**Recommendation**: Investigate manually
- Same location (NYC = New York)
- Similar names (Beta Tech vs Beta Technologies)
- Different domains - likely different companies
### Summary
- **Definite Duplicates**: 2 sets (4 records → 2)
- **Possible Duplicates**: 2 sets (need research)
- **Unique Records**: Pending investigation
### Merge Checklist
Before merging, verify:
- [ ] All opportunities moved to master record
- [ ] All contacts linked to master
- [ ] Activity history preserved
- [ ] Custom fields compared
- [ ] Integration IDs noted (for external systems)
Skill Boundaries
What This Skill Does Well
- Defining audit criteria systematically
- Identifying common data quality issues
- Prioritizing fixes by business impact
- Creating actionable remediation plans
What This Skill Cannot Do
- Access your actual CRM data
- Execute cleanup automatically
- Know your specific business rules
- Detect all semantic errors
When to Escalate to Human
- Merge decisions (data loss risk)
- Business rule definitions
- Enforcement policy decisions
- Cross-object relationship issues
Iteration Guide
Follow-up Prompts
- "Create a monthly hygiene report template."
- "What validation rules would prevent these issues?"
- "Prioritize the top 10 issues by revenue impact."
- "Design a data entry form that enforces these rules."
Continuous Improvement Cycle
- Audit → Identify issues
- Fix → Clean data
- Prevent → Add validation rules
- Monitor → Track hygiene score weekly
- Iterate → Refine rules quarterly
Checklists & Templates
Weekly Hygiene Report Template
## CRM Hygiene Report - Week of [Date]
### Hygiene Score: X/100
### Critical Issues (Fix Today)
| Object | Record | Issue | Owner |
|--------|--------|-------|-------|
### Warnings (Fix This Week)
| Object | Count | Issue Type |
|--------|-------|------------|
### Trends
- Issues closed this week: X
- New issues this week: X
- Net change: +/- X
### Action Items
1.
2.
3.
Data Governance Checklist
- Required fields defined per object
- Validation rules implemented
- Duplicate rules active
- Stale record reports scheduled
- Ownership assignments current
- Integration mappings documented
References
- Salesforce Data Quality Best Practices
- RevOps Co-op Data Governance Framework
- Marketo Database Health Guide
- Gartner Data Quality Market Guide
Related Skills
pipeline-forecasting- Clean data improves forecastslead-scoring- Requires complete lead datadeal-risk-scoring- Depends on activity data
Skill Metadata
- Domain: RevOps
- Complexity: Intermediate
- Mode: centaur
- Time to Value: 30-60 min for audit, varies for cleanup
- Prerequisites: CRM data export, field requirements list
GitHub 仓库
相关推荐技能
llamaguard
其他LlamaGuard是Meta推出的7-8B参数内容审核模型,专门用于过滤LLM的输入和输出内容。它能检测六大安全风险类别(暴力/仇恨、性内容、武器、违禁品、自残、犯罪计划),准确率达94-95%。开发者可通过HuggingFace、vLLM或Sagemaker快速部署,并能与NeMo Guardrails集成实现自动化安全防护。
cost-optimization
其他这个Claude Skill帮助开发者优化云成本,通过资源调整、标记策略和预留实例来降低AWS、Azure和GCP的开支。它适用于减少云支出、分析基础设施成本或实施成本治理策略的场景。关键功能包括提供成本可视化、资源规模调整指导和定价模型优化建议。
quantizing-models-bitsandbytes
其他这个Skill使用bitsandbytes库量化大语言模型,能在GPU内存有限时通过8位或4位量化减少50-75%内存占用,同时保持精度损失最小。它支持INT8、NF4、FP4等多种量化格式,可与HuggingFace Transformers无缝集成,适用于需要部署更大模型或加速推理的场景。还提供QLoRA训练和8位优化器支持,让开发者能轻松实现高效模型压缩。
dispatching-parallel-agents
其他该Skill用于并行处理3个以上无依赖关系的独立故障,可为每个问题域分派专属Claude代理同时执行调查修复。它通过并发处理多个独立问题显著提升故障排查效率,特别适用于测试文件、子系统等无共享状态的场景。
