Sales Data Hygiene

# CRM Data Hygiene & Quality Learnings

Accumulated tips, gotchas, and corrections discovered during use. Claude reads this at the start of each invocation and appends new learnings as they're discovered. Once significant learnings have accumulated, use `/sales-request-skill` to share them back to the community. Shared and declined entries are marked so they won't be re-prompted.

<!-- Add entries below in format: **YYYY-MM-DD**: Learning description -->


# CRM Data Hygiene Platform Guide

## Data Quality Audit Framework

Before fixing data, measure what's broken:

1. **Completeness** — what % of records have all critical fields filled?
   - Email: target 95%+ for contacts
   - Phone: target 70%+ for key personas
   - Company: target 100% for accounts
   - Title/department: target 90%+ for contacts

2. **Accuracy** — what % of filled fields are actually correct?
   - Email deliverability: verify a sample with an email verification tool
   - Phone connectivity: check a sample of direct dials
   - Job title currency: compare against LinkedIn for a sample

3. **Duplication rate** — what % of records are duplicates?
   - Contact-level: same person, multiple records
   - Account-level: same company, different spellings
   - Cross-object: leads that are also contacts

4. **Decay rate** — how fast does your data go stale?
   - Industry average: 30% of B2B data decays annually
   - Sales contacts: ~20% change jobs each year
   - Direct dials: ~15% become invalid annually
   - Emails: ~22% bounce rate after 12 months without refresh

5. **Consistency** — are the same things called the same thing?
   - Job titles: "VP Sales" vs "Vice President of Sales" vs "VP, Sales"
   - Industries: "SaaS" vs "Software" vs "Technology"
   - Company names: "IBM" vs "International Business Machines" vs "IBM Corp"

## Deduplication Strategy

| Approach | When to use | Risk level |
|----------|------------|------------|
| **Exact match** | Email, phone, domain — safest | Low |
| **Fuzzy match** | Names, company names, addresses | Medium — review matches before merging |
| **Rule-based** | Combine multiple fields (name + company + title) | Medium |
| **ML-based** | Large datasets with complex patterns | Low (if trained well) — but expensive |

**Merge rules** (which record wins):
- Most recently updated record keeps modifiable fields
- Most complete record keeps enrichment data
- Oldest record keeps the original owner/source
- Always preserve: original lead source, first touch date, opt-in status

## Data Normalization

| Field | Common problems | Solution |
|-------|----------------|----------|
| Job title | Abbreviations, variations, custom titles | Map to standard taxonomy (C-Level, VP, Director, Manager, IC) |
| Industry | Free-text, overlapping categories | Map to SIC/NAICS or your internal taxonomy |
| Company name | Abbreviations, legal suffixes, DBA names | Normalize to official name, store variants as aliases |
| Phone | Mixed formats, extensions, country codes | E.164 format (+1XXXXXXXXXX) |
| Address | Inconsistent formatting, abbreviations | USPS standardization or Google Maps API |
| Country | Mix of codes and names | ISO 3166-1 alpha-2 codes |

## Enrichment Automation

Set up ongoing enrichment to prevent decay:

1. **Trigger-based** — enrich when a record is created or updated
2. **Scheduled** — monthly/quarterly batch enrichment of all records
3. **Decay-based** — re-enrich records older than X days
4. **Event-based** — re-enrich when a contact's company has a news event (funding, acquisition)

## Platform-Specific Guidance

### In ZoomInfo (OperationsOS)

ZoomInfo OperationsOS is purpose-built for CRM data management at scale.

**Deduplication**:
- OperationsOS identifies duplicates across contacts, leads, and accounts using fuzzy matching
- Configurable match rules: email, name+company, phone, domain
- Bulk merge with configurable "winning record" rules
- Cross-object dedup: find leads that already exist as contacts

**Data orchestratio