Skip to content
Docs

Best Practices

Best practices for getting the most out of Sellestial while maintaining data quality and controlling costs.

General Principles

Section titled “General Principles”

Start Small

Section titled “Start Small”
Why this matters
  • Catch configuration errors early
  • Verify output quality
  • Understand costs
  • Identify issues quickly
How to test safely
  1. Create small test list in HubSpot (10-50 records)
  2. Configure pipeline with test list
  3. Deploy and process
  4. Review every result manually
  5. Adjust configuration
  6. Scale to larger lists

Enable Human Review Initially

Section titled “Enable Human Review Initially”

For high-impact operations:

When to require review

Always require review for:

  • Data deletions
  • Record merges
  • Bulk updates to critical fields
  • New pipeline deployments

Can auto-confirm after:

  • 50-100 successful reviews
  • Confidence in results
  • Pattern validation complete

Monitor Continuously

Section titled “Monitor Continuously”

Daily checks:

  • Processing status
  • Error rates
  • Success rates

Weekly reviews:

  • Cost analysis
  • Quality spot checks
  • Performance trends

Monthly audits:

  • Configuration review
  • Results validation
  • ROI assessment

Data Hygiene Best Practices

Section titled “Data Hygiene Best Practices”

Workflow Order

Section titled “Workflow Order”
Recommended sequence
  1. Validate — Code pipelines for basic validation
  2. Classify — Classifier pipelines for quality assessment
  3. Clean — Agent pipelines for normalization
  4. Deduplicate — Deduplication + merge pipelines
  5. Maintain — Agent pipelines for ongoing monitoring

Deduplication Strategy

Section titled “Deduplication Strategy”

Set up rules:

  1. Start with exact match on unique identifiers (domain, LinkedIn URL)
  2. Add fuzzy match cautiously
  3. Test with small dataset
  4. Publish when confident

Review pairs:

  • Check all pairs before merging
  • Document merge decisions
  • Keep audit trail

Large merges:

  • Extra caution for companies with more than 30 associations
  • Verify data quality first
  • Consider manual review

Data Cleaning Frequency

Section titled “Data Cleaning Frequency”

Continuous:

  • New record validation (Code and Classifier pipelines)

Weekly:

  • Recent record cleaning (Agent pipelines on last 7 days)

Monthly:

  • Full database cleanup pass
  • Deduplication review
  • Employment verification

Quarterly:

  • Comprehensive data audit
  • Rule review and adjustment
  • Archive obsolete records

Data Enrichment Best Practices

Section titled “Data Enrichment Best Practices”

Cost Optimization

Section titled “Cost Optimization”
Filter before enriching
Step 1: Validation (Free Code pipeline)
Step 2: Quality Assessment (Low-cost Classifier)
Step 3: Filter for high-quality records
Step 4: Data Cleaning (Moderate-cost Agent)
Step 5: Enrichment (Higher-cost service)
Savings: Avoid enriching invalid/low-quality records

Target Strategically

Section titled “Target Strategically”
What to enrich

Don’t enrich:

  • All contacts (expensive)
  • Test data
  • Invalid emails
  • Archived records

Do enrich:

  • Active opportunities
  • Recent leads (last 30 days)
  • High-value segments
  • Missing critical data

Write Mode Strategy

Section titled “Write Mode Strategy”
Choosing write modes

Default to “Write if empty”:

  • Safest option
  • Preserves existing data
  • Fills gaps only

Use “Always overwrite” for:

  • Standardization fields (Country, Industry)
  • Known poor-quality data
  • Enrichment refresh

Use “Write if not modified by user” for:

  • Job titles (change frequently)
  • Company data (needs updates)
  • Phone numbers (may update)

Allowed Values

Section titled “Allowed Values”

Configure for:

  • Country names
  • Industry classifications
  • State abbreviations
  • Standard enums

Benefits:

  • Consistent segmentation
  • Clean reporting
  • Reliable automation

Pipeline Configuration Best Practices

Section titled “Pipeline Configuration Best Practices”

Processing Limits

Section titled “Processing Limits”

Set appropriate limits:

High-volume (1000s/day):

Enrollment: 1000/day
Generation: 1000/day
Weekend blocking: OFF

Medium-volume (100s/day):

Enrollment: 100/day
Generation: 100/day
Weekend blocking: Optional

Low-volume (10s/day):

Enrollment: 50/day
Generation: 50/day
Weekend blocking: ON

Review Policies

Section titled “Review Policies”

Auto-confirm:

  • Well-tested pipelines
  • Low-risk operations (reads, logging)
  • High-volume, low-impact

Human review:

  • New pipelines (first 50-100 records)
  • High-impact operations (merges, deletes)
  • Critical data updates
  • Uncertain classifications

Timed auto-confirm (10 mins):

  • Sequences (review content quality)
  • Medium-impact updates
  • Allows quick review without blocking

Data Source Selection

Section titled “Data Source Selection”

Always enable:

  • HubSpot contact/company (core data)
  • Previous communication (avoid duplicates)

Selectively enable:

  • LinkedIn profiles (if personalization needed)
  • Company news (if timing important)

Rarely enable:

  • LinkedIn posts (expensive, often low value)
  • All external sources simultaneously

Cost Management

Section titled “Cost Management”

Budget Planning

Section titled “Budget Planning”

Calculate costs:

Pipeline cost per record × Expected volume = Daily cost


Example:
Agent pipeline: 1 credit/contact
500 contacts/day
= 500 credits/day
= ~15,000 credits/month

Set daily limits: Match budget constraints with processing limits in pipeline settings.

Cost Monitoring

Section titled “Cost Monitoring”

Weekly:

  • Check Usage → Credits tab
  • Review category distribution
  • Identify unexpected spikes

Monthly:

  • Analyze trends
  • Compare to budget
  • Adjust limits or strategy

Cost Optimization Tactics

Section titled “Cost Optimization Tactics”

1. Filter Aggressively:

Before enrichment:
- Remove test data
- Validate with Code pipelines (low/free)
- Assess quality with Classifiers
- Filter invalid
Result: Only enrich quality records

2. Target High-Value:

Instead of:
50,000 contacts × enrichment cost = High total


Target:
2,500 active opportunities × enrichment cost = Lower total
Savings: Significant cost reduction

3. Use Cheaper Pipelines First:

Code (Free/Low) → Filter
Classifier (Low-Moderate) → Filter
Then: Agent and enrichment on validated data

4. Adjust Processing Frequency:

Daily: New records only
Weekly: Recent activity
Monthly: Full database
Quarterly: Comprehensive audit

Performance Optimization

Section titled “Performance Optimization”

Speed Improvements

Section titled “Speed Improvements”

Parallel processing:

  • Use multiple pipelines simultaneously
  • Process different segments
  • Separate hygiene from enrichment

Efficient data sources:

  • Disable unnecessary sources
  • Use only required fields
  • Minimize external API calls

Batch sizing:

  • Find optimal batch size
  • Too small = overhead
  • Too large = timeouts
  • Test 50-200 record batches

Success Rate Optimization

Section titled “Success Rate Optimization”

For enrichment:

  1. Clean data first
  2. Ensure required fields present
  3. Validate company associations
  4. Use quality source lists

For classification:

  1. Provide sufficient context
  2. Use appropriate model
  3. Add validation if needed
  4. Review borderline cases

HubSpot Integration Best Practices

Section titled “HubSpot Integration Best Practices”

List Management

Section titled “List Management”

Create specific lists:

"Contacts - Missing Email"
"Contacts - Invalid Domain"
"Companies - Need LinkedIn URL"
"Contacts - Created Last 7 Days"

Use active lists:

  • Auto-updates membership
  • Continuous processing
  • No manual maintenance

Document criteria:

  • Clear naming
  • Note purpose
  • Track usage

Property Management

Section titled “Property Management”

Naming conventions:

  • Use sellestial_ prefix
  • Descriptive names
  • Consistent format

Documentation:

  • Add help text
  • Explain values
  • Link to pipelines

Organization:

  • Group related properties
  • Standard property groups
  • Easy to find

Field Mapping

Section titled “Field Mapping”

Review mappings:

  • Correct target fields
  • Appropriate write modes
  • Proper data types

Test writes:

  • Verify updates work
  • Check data format
  • Validate calculations

Security & Compliance

Section titled “Security & Compliance”

Data Privacy

Section titled “Data Privacy”

GDPR/CCPA:

  • Document data processing
  • Provide opt-out mechanisms
  • Respect preferences
  • Maintain records

Internal policies:

  • Follow company guidelines
  • Legal team review
  • Compliance checks

Access Control

Section titled “Access Control”

User permissions:

  • Admin: Full access
  • User: Pipeline management
  • Viewer: Read-only

API keys:

  • Secure storage
  • Regular rotation
  • Scope limiting
  • Audit usage

Documentation Best Practices

Section titled “Documentation Best Practices”

Pipeline Documentation

Section titled “Pipeline Documentation”

For each pipeline:

  • Purpose and goals
  • Configuration details
  • Input sources
  • Expected outputs
  • Review process
  • Owner/contacts

Change Management

Section titled “Change Management”

Document:

  • Configuration changes
  • Deployment dates
  • Version numbers
  • Reasons for changes
  • Impact analysis

Communicate:

  • Notify affected teams
  • Explain changes
  • Training if needed
  • Support available

Team Collaboration

Section titled “Team Collaboration”

Roles & Responsibilities

Section titled “Roles & Responsibilities”

Admin:

  • Platform configuration
  • Pipeline deployment
  • Settings management
  • User access control

Pipeline Owner:

  • Pipeline configuration
  • Quality monitoring
  • Issue resolution
  • Performance optimization

Reviewer:

  • Manual review tasks
  • Quality spot checks
  • Feedback reporting

Communication

Section titled “Communication”

Regular sync:

  • Weekly pipeline review
  • Monthly performance review
  • Quarterly strategy planning

Documentation:

  • Share best practices
  • Document decisions
  • Maintain runbooks
  • Update guides

Troubleshooting Workflow

Section titled “Troubleshooting Workflow”

When issues arise:

  1. Check basics:

    • Pipeline active?
    • Sources configured?
    • Limits reasonable?
    • Recent deployments?

  2. Review logs:

    • Execution logs
    • Error messages
    • Status codes
    • Timing information

  3. Test in isolation:

    • Single record test
    • Verify each component
    • Check data sources
    • Review output

  4. Adjust & retry:

    • Fix identified issues
    • Deploy changes
    • Retest
    • Monitor results

  5. Escalate if needed:

    • Document issue
    • Gather logs
    • Contact support
    • Provide context

Next Steps

Section titled “Next Steps”