Setting Up Identity Resolution

This guide walks you through configuring Zeotap’s identity resolution engine to unify customer records across multiple data sources into golden records — single, authoritative customer profiles. By the end, you’ll have merge rules, limit rules, and a resolution pipeline producing unified profiles you can explore and activate.

When You Need Identity Resolution

Identity resolution becomes essential when:

• Customers interact across channels — A customer signs up with their work email, makes a purchase with their personal email, and browses on their phone. Without resolution, this looks like three separate people.
• Data lives in multiple systems — Your CRM has one view, your product database has another, and your marketing platform has a third. Each system uses different identifiers.
• You’re building a 360-degree customer view — Computed attributes, audiences, and orchestrations are only as good as the underlying profiles. Fragmented identities lead to duplicate audiences and inconsistent experiences.
• You’re activating across destinations — Sending the same customer to a CRM as three separate leads wastes sales time and damages trust.

How Zeotap Resolves Identities

Zeotap uses a connected-components graph algorithm to merge records:

1. Collect identifiers — Zeotap gathers all identifiers (emails, phones, device IDs, account IDs) from your mapped schema attributes.
2. Build the graph — Each identifier becomes a node. Two records sharing an identifier value are connected by an edge. Identifier families define which identifiers can form connections.
3. Find connected components — The algorithm finds all groups of records that are transitively connected through shared identifiers. Each connected component becomes a unified profile.
4. Apply merge rules — When multiple records contribute conflicting values for the same attribute, merge rules determine which value wins.
5. Enforce limit rules — Limit rules prevent unreasonably large clusters (super-clusters) that can form from shared device IDs or generic email addresses.

Prerequisites

• A configured warehouse with customer data
• Relationships configured with at least one entity type that has identifier attributes mapped
• Multiple data sources or tables contributing to the same entity type (otherwise there’s nothing to resolve)

Step 1: Audit Your Identifier Landscape

Before configuring identity resolution, understand what identifiers exist across your data.

Survey Your Data Sources

For each table or data source that contributes to your customer entity type, catalog the available identifiers:

Identify Linking Opportunities

Look for identifiers that bridge across sources:

• Email bridges CRM, Product, Website (logged-in), Mobile, and Support
• Phone bridges CRM and Support
• User ID is internal to Product but can link to CRM via an integration table

Flag Problematic Identifiers

Some identifiers create false matches:

• Shared device IDs — Family members or shared computers
• Generic email addresses — info@company.com, noreply@example.com
• Recycled phone numbers — Carriers reassign phone numbers
• Test accounts — Internal testing with dummy data

You’ll address these with limit rules in Step 4.

Step 2: Define Identifier Families

Identifier families group identifiers by type and define how they participate in resolution.

1. Navigate to ID Graphs under Identity in the left sidebar
2. Open the Identifier Families tab
3. Click Create Family for each identifier type:

Recommended Identifier Families

Configuring a Family

For each family:

1. Name — Descriptive name (e.g., “Email”)
2. Columns — Select the schema attributes that belong to this family. All columns in a family are treated as the same type of identifier.
3. Priority — Determines preference when resolving conflicts. Higher-priority families produce more trusted links.
4. Normalization — Zeotap normalizes identifiers before matching:
   • Emails: lowercase, trim whitespace
   • Phones: strip formatting, apply E.164 when possible
   • All: remove leading/trailing spaces

Step 3: Configure Merge Rules

Merge rules determine which attribute value wins when multiple source records contribute conflicting values for the same attribute in a golden record.

1. Navigate to ID Graphs under Identity, then open the Merge Rules tab
2. Click Create Merge Rule

Merge Strategies

Example Merge Rules Configuration

Setting Up a Merge Rule

For each attribute you want in your golden record:

1. Select the attribute name
2. Choose the merge strategy
3. If using Source Priority, drag sources into preference order
4. If using Custom SQL, write the aggregation expression
5. Click Save

Attributes without explicit merge rules default to Most Recent.

Step 4: Set Limit Rules to Prevent Super-Clusters

A single shared identifier — a device ID from a public computer, a generic inbox — can merge thousands of unrelated records into one profile. Limit rules cap how many distinct records may share one identifier value.

This step is optional. Every enabled merge rule is already capped at 100 by default; configure a rule only where that number is wrong for your data.

1. Navigate to ID Graphs under Identity, open your graph, and find Limit Rules on the Configuration tab
2. Click Edit and set Max unique values for the families you want to override

The table shows every enabled family with the cap actually in force and whether it is Configured or Default.

Recommended Settings

A value that exceeds its cap contributes no edges. The records carrying it still resolve — as their own profiles, or merged through another identifier they share.

If you find yourself raising a cap to accommodate one bad value, exclude that value in the model’s SQL instead — a placeholder identifier is a data quality problem, and filtering it at the source is more precise than loosening the cap for everything.

Step 5: Run Initial Full Resolution

With identifier families, merge rules, and limit rules configured, run the first full resolution pass.

1. Navigate to ID Graphs under Identity, then open the Resolution Runs tab
2. Click Run Full Resolution
3. Zeotap will:
   • Collect all identifier values from your mapped schema attributes
   • Build the identity graph
   • Find connected components
   • Apply merge rules to produce golden records
   • Enforce limit rules to break super-clusters
4. Monitor the run progress — the first full run may take 10–60 minutes depending on data volume

Understanding the Resolution Summary

After the run completes, review the summary:

A merge rate above 50% often indicates overly aggressive linking — review your identifier families and limit rules.

A merge rate above 50% often indicates overly aggressive linking — review your identifier families and limit rules.

Step 6: Verify Golden Records in Profile Explorer

Inspect the resolved profiles to confirm the resolution quality.

1. Navigate to ID Graphs under Identity, then open the Profile Explorer tab
2. Search for a known customer by email, phone, or ID
3. Review the golden record:
   • Merged attributes — Check that the right values were selected by merge rules
   • Contributing records — See which source records were merged together
   • Identity graph — Visualize the identifier connections that caused the merge
4. Spot-check several profiles:
   • A customer you know has multiple accounts
   • A customer who should NOT be merged with anyone
   • A high-value customer where accuracy is critical

What to Look For

Good resolution:

• Known duplicate accounts are correctly merged
• Attribute values come from the expected authoritative source
• The identity graph shows logical connections (same email across systems)

Problematic resolution:

• Unrelated people merged into one profile (over-merging) → Tighten limit rules or remove problematic identifier families
• Known duplicates not merged (under-merging) → Check that the linking identifiers are mapped and in the right family
• Wrong attribute values in the golden record → Adjust merge rule priorities

Step 7: Schedule Incremental Resolution

Once you’re satisfied with the initial resolution, set up incremental runs to keep golden records current as new data arrives.

1. Navigate to ID Graphs under Identity, then open the Settings tab
2. Under Schedule, select the resolution frequency:

3. Configure the run window — choose off-peak hours to minimize warehouse compute impact
4. Enable incremental mode — processes only new and changed records since the last run, significantly faster than full resolution
5. Click Save

Full vs. Incremental Resolution

Schedule a full resolution run weekly or monthly to catch edge cases that incremental processing might miss (e.g., retroactive data corrections).

Common Pitfalls and Troubleshooting

Over-Merging (Clusters Too Large)

Symptom: Golden records contain records from clearly different people.

Diagnosis:

1. Open the problematic golden record in Profile Explorer
2. Examine the identity graph to find the linking identifier
3. Check if the identifier is a shared device, generic email, or recycled phone number

Solutions:

• Add the problematic identifier value to a blocklist
• Lower the link limit for the relevant identifier family
• Reduce the maximum cluster size
• Remove low-trust identifier families (e.g., cookie IDs) from resolution

Under-Merging (Too Many Duplicates)

Symptom: Known duplicate customers remain as separate golden records.

Diagnosis:

1. Search for the duplicates in Profile Explorer
2. Check what identifiers each record has
3. Verify that the linking identifier is mapped in the schema and assigned to an identifier family

Solutions:

• Ensure all relevant identifier columns are mapped in your schema
• Check that identifier normalization is working (e.g., John@example.com and john@example.com should match after lowercasing)
• Add additional identifier families if new linking opportunities exist

Resolution Runs Timing Out

Symptom: Full resolution runs fail or take excessively long.

Solutions:

• Check warehouse compute resources — resolution is CPU/memory-intensive
• For Snowflake, use a larger warehouse size (at least MEDIUM for datasets > 10M records)
• For BigQuery, ensure you have sufficient slot capacity
• Break the run into phases: resolve high-priority identifier families first, then add lower-priority ones

Merge Rule Producing Wrong Values

Symptom: Golden record attributes have unexpected values.

Diagnosis:

1. Open the golden record in Profile Explorer
2. Check the “Contributing Records” panel to see all source values
3. Verify which merge rule is applied to the attribute

Solutions:

• Adjust the source priority order in your merge rule
• Switch to a different merge strategy (e.g., Most Recent instead of Source Priority)
• Add a Custom SQL merge rule for complex resolution logic

Next Steps

• Identity Resolution concepts — Detailed documentation on the resolution engine
• Profile Explorer — Navigate and inspect golden records
• Building a Warehouse-First CDP — End-to-end CDP setup guide
• Audiences — Build audiences on top of resolved profiles