Data Cloud Consultant Identity Resolution

Identity Resolution Detailed Explanation

Identity resolution is a critical process in Salesforce Data Cloud that ensures the data about a single customer from multiple sources is consolidated into one accurate, unified profile.

1. Functionality of Identity Resolution

1.1 Identity Matching

What is it?
Identity matching involves comparing data across different records to determine if they belong to the same customer. This process uses predefined rules to match records accurately.

Key Details:

• Matching Rules:
  Rules that specify which fields to compare when identifying matches. Common fields include:

  • Name
  • Email
  • Phone Number
  • Address

• Weighted Matching:
  Each field is given a weight based on its importance. For example:

  • Email might have a higher weight (e.g., 70%) compared to a phone number (e.g., 30%).
  • A total weight threshold (e.g., 90%) determines if the records are a match.

Example:
Two records:

• Record 1: Name = "John Doe," Email = "[email protected]"
• Record 2: Name = "John D.," Email = "[email protected]"
  Matching rules prioritize email over name, identifying these as the same person.

1.2 Deduplication

What is it?
Deduplication eliminates redundant records by merging duplicates into a single, clean record.

Key Details:

• Merge Records:
  Combines information from duplicates into one profile.

  • Example: Merge “John Doe” from CRM and “J. Doe” from an e-commerce platform into one record.

• Priority Rules:

  • Set rules to determine which data to keep in case of conflicts.
  • For example:
    • Keep the most recent data for fields like address.
    • Prioritize the CRM source over social media for contact details.

Example:

• Duplicate 1: Email = "[email protected]," Phone = "123-456-7890"
• Duplicate 2: Email = "[email protected]," Phone = "987-654-3210"
  Rule: Keep the most recent phone number → Result: Phone = "987-654-3210"

1.3 Unified Profiles

What is it?
A unified profile is the result of merging data from various sources to create a single, comprehensive view of the customer.

Key Features:

• Integration from Multiple Sources:
  Combine data from CRM, e-commerce, social media, and other platforms.
• Dynamic Updates:
  Real-time ingestion ensures that profiles are updated as new data comes in.

Example:
For "John Doe," a unified profile might include:

• Name: John Doe
• Email: [email protected]
• Phone: 987-654-3210
• Purchase History: 5 orders from the e-commerce platform
• Social Media Engagement: Likes and shares on Facebook

2. Technical Details

2.1 Matching Rules

What are Matching Rules?
Rules that determine how records are compared during the identity resolution process.

Types of Matching:

• Exact Matching:
  Requires fields to match exactly.

  • Example: Two records with the same email ("[email protected]") are identified as the same.

• Fuzzy Matching:
  Allows slight variations in data to match.

  • Example: "John Doe" and "J. Doe" are identified as the same person.

Threshold Configuration:
Set a score threshold to determine a match.

• Example: If the matching score is above 90%, consider the records as duplicates.

2.2 Reconciliation Rules

What are Reconciliation Rules?
Rules for resolving conflicts when two records have different values for the same field.

How it works:

• Assign a priority to the source:
  • Example: CRM > E-commerce > Social Media.
• Keep the value from the most trusted source or the most recent data.

Example of Conflict Resolution:

• Record 1: Phone = "123-456-7890" (CRM)
• Record 2: Phone = "987-654-3210" (E-commerce)
  Rule: Trust CRM data → Result: Phone = "123-456-7890"

3. Exam Focus

3.1 Differentiate Between Matching Rules and Reconciliation Rules

• Matching Rules: Identify if two records represent the same customer.
• Reconciliation Rules: Decide how to handle conflicting information between duplicate records.

3.2 Optimize Identity Resolution for Accuracy

• Use weighted matching for better precision.
• Adjust thresholds to balance accuracy and false positives/negatives.
• Prioritize trusted data sources in reconciliation.

Summary for Beginners

1. Identity Matching: Finds records that belong to the same customer using rules for comparison.
2. Deduplication: Eliminates duplicate records by merging them into one unified profile.
3. Unified Profiles: Combines data from multiple sources to provide a complete, up-to-date view of each customer.
4. Technical Details: Matching rules define how to identify duplicates, while reconciliation rules determine how to handle conflicts.

Mastering identity resolution ensures your data is accurate, reliable, and ready for actionable insights.

Identity Resolution (Additional Content)

1. Identity Graph

1.1 Why Is Identity Graph Important?

An Identity Graph is a core component of Identity Resolution in Salesforce Data Cloud. It dynamically establishes relationships between different identity attributes (such as email, phone numbers, and social media IDs) to create a unified customer profile.

1.2 Key Features of Identity Graph

• Multi-Source Identity Mapping

  • Links customer identities across multiple data sources.
  • Captures identifiers such as email, phone number, social media handles, CRM records, and transaction history.

• Graph-Based Relationship Building

  • Uses a network structure to connect identity attributes and form a 360-degree customer profile.

• Continuous Updates

  • Ensures new customer data is dynamically integrated and reconciled with existing identity records.

1.3 Example: Identity Graph in Action

A customer named John Doe might have multiple identity records across different platforms:

Data Source	Identity Attribute
CRM System	Customer ID: 12345
Email System	Email: [email protected]
Social Media	Twitter Handle: @john_doe
E-Commerce	Purchase History under Name: J. Doe

Without an Identity Graph, these records would be stored separately. With Identity Graph, Salesforce Data Cloud automatically detects and consolidates them, ensuring John Doe is recognized as a single customer.

2. Deterministic Matching vs. Probabilistic Matching

2.1 Why Is Identity Matching Important?

When merging customer records, businesses use two primary matching techniques to determine whether multiple records belong to the same person.

2.2 Deterministic Matching

Definition:
Deterministic matching relies on exact matches of unique identifiers to link records.

Key Features:

• Uses precise, one-to-one identity attributes (e.g., email, government ID, phone number).
• Low false positive rate (accurate matches).
• Higher false negative rate (may fail to match records due to slight variations).

Example of Deterministic Matching:

Customer Record 1	Customer Record 2	Match?
Email: [email protected]	Email: [email protected]	Match
Phone: +1-555-1234	Phone: +1-555-5678	No Match

2.3 Probabilistic Matching

Definition:
Probabilistic matching calculates a similarity score between multiple fields and determines matches based on probability thresholds.

Key Features:

• Uses fuzzy matching for names, addresses, or phone numbers.
• More flexible, allowing records with minor variations to be matched.
• Higher false positive rate, as it assumes partial matches could still represent the same individual.

Example of Probabilistic Matching:

Customer Record 1	Customer Record 2	Similarity Score	Match?
Name: John Doe	Name: J. Doe	85%	Match
Email: [email protected]	Email: [email protected]	90%	Match
Address: 123 Main St	Address: 123 Main Str.	95%	Match

2.4 When to Use Each Matching Method

Matching Type	Use Case	Pros	Cons
Deterministic Matching	When unique identifiers (email, customer ID) are available	Highly accurate, reduces false positives	Fails when minor discrepancies exist
Probabilistic Matching	When no unique identifier is available, but data has similarities	Flexible, can match variations	Risk of false positives

3. False Positives vs. False Negatives

3.1 Why Is Matching Accuracy Important?

Incorrect identity resolution can lead to major business risks, such as misaligned customer data, ineffective marketing, and compliance issues.

3.2 Understanding False Positives and False Negatives

Matching Issue	Definition	Business Impact
False Positive (Incorrect Match)	Different customers are mistakenly merged into one profile	Causes data confusion, leading to irrelevant marketing or security risks
False Negative (Missed Match)	The same customer is mistakenly treated as separate individuals	Causes incomplete customer profiles, impacting personalization and customer service

3.3 Solutions to Improve Matching Accuracy

1. Adjust Matching Weights:

• Increase weight for highly unique fields (e.g., email, government ID).
• Decrease weight for less reliable fields (e.g., name, address).

2. Set an Optimal Matching Threshold:

• If the similarity score is too high (e.g., 98%), it may cause false negatives.
• If the similarity score is too low (e.g., 70%), it may cause false positives.

3. Combine Automated Matching with Manual Review:

• High-risk matches (e.g., customers with similar emails but different addresses) should require manual approval.

4. Identity Resolution Performance Optimization

4.1 Why Is Performance Optimization Important?

Identity Resolution must handle millions of records efficiently, ensuring high-speed processing and accurate identity linking.

4.2 Key Optimization Strategies

1. Indexing Matching Fields

• Optimize database queries by indexing high-frequency fields (such as email and phone number).
• Example: Instead of scanning 100 million records, an indexed query narrows results to thousands of potential matches.

2. Leveraging AI & Machine Learning

• AI-driven matching models can improve accuracy over time by learning from past matching errors.
• Machine learning refines probabilistic matching thresholds dynamically based on historical match success rates.

3. Balancing Batch vs. Real-Time Processing

Processing Type	Use Case	Pros	Cons
Batch Matching	Overnight processing of large historical datasets	Efficient for large-scale identity resolution	Not useful for real-time actions
Real-Time Matching	Dynamic updates to customer profiles as new data arrives	Enables instant personalization	More computationally expensive

4.3 Example: Optimizing Identity Resolution for Speed

Optimization Step	Before Optimization	After Optimization
Unindexed Matching	12 hours for 10M records	N/A
Indexed Matching Fields	N/A	4 hours
AI-Based Matching with Machine Learning	N/A	3 hours

Conclusion

Key Takeaways

1. Identity Graph:

• Dynamically links multiple identity sources into a unified customer profile.

2. Deterministic vs. Probabilistic Matching:

• Deterministic Matching is precise but strict (ideal for structured data).
• Probabilistic Matching is flexible but has a risk of false positives (ideal for unstructured data).

3. Managing False Positives & False Negatives:

• Adjust matching weights to reduce errors.
• Use a hybrid AI/manual review system for high-risk matches.

4. Performance Optimization:

• Index high-frequency fields to speed up matching.
• Leverage AI & machine learning to improve accuracy.
• Balance batch and real-time matching based on business needs.

By mastering these Identity Resolution techniques, businesses can build accurate, scalable, and real-time customer identity systems in Salesforce Data Cloud.