Identify and Manage Duplicate and Disconnected Records
Learning Objectives
After completing this unit, you’ll be able to:
- Define intentional duplicates, unintentional duplicates, and disconnected records.
- Explain when to merge records versus when to unify profiles.
- Describe what false positives are, what causes them, and why they matter.
The Challenge of Building Customer Context from Fragmented Data
A contextual, consistent, current, and correct understanding of a customer depends on a complete view of all relevant records, transactions, and interactions across data sources. These sources can include CRM data and other systems like commerce, loyalty, and order management. It can also include interaction data, such as email, that might be captured as unstructured content.
Duplicate and disconnected records can quietly undermine any initiative that depends on trusted customer context, including analytics, automation, and agentic AI. The issue is not just having too many records. The real problem is a broken or incomplete view of the customer, which can lead to:
- Incomplete context, like missed orders, missed cases, or missed warranties
- Incorrect segmentation or routing, like misclassified customer type or value
Luna was part of Northern Trader Outfitters’ (NTO) Customer Journey Mapping exercise. She knows that customers share different details at different times: an email or phone number for marketing promotions, shipping and billing details for ecommerce orders, and the most convenient phone, email, or social handle for customer support. To choose the right path, she first maps where these signals live across CRM, orders, loyalty, and interaction data.
Luna is about to begin a process called identity resolution. This process matches records to unify customer profiles. Unified customer profiles act as a single source of truth, helping teams see a clear, complete, and up-to-date view of NTO’s customers.
To learn more about the importance of customer context for AI, take the Data Reliability Risk Factors for AI badge.
Explore Duplication and Disconnect Risks
Use this table to define, detect, and decide what to do before you merge records or configure identity resolution.
What Causes the Customer Context Gap |
How to Detect It |
What to Do |
|---|---|---|
|
Unintentional duplicates: Two or more records represent the same entity (for example, real-world person, account, or case) and context due to data entry, integrations, or process gaps. |
High match potential across key identifiers (for example, email, phone, or name plus address), repeated values, and overlapping activity or transactions suggesting the same entity. |
Merge in the system of record when appropriate. This typically reconciles field values on the surviving record and re-parents child records. Use tools or processes that support unmerge or rollback to safely revert mistakes. |
|
Intentional duplicates: Separate records are intentionally distinct for governance or operational reasons and aren’t considered true duplicates that cause customer context gaps. |
Similar identifiers might exist, but business context shows they must remain separate (for example, Experience Cloud self-service versus employee-created records or supplier versus customer accounts). |
Do not merge. Preserve as distinct, and use a unification strategy like profile linkage or the key-ring approach to enable a compliant, contextual view. Match and relate to other records with Data 360 identity resolution match and reconciliation rules and Data Spaces when contextual views are needed. Document data governance policy indicating why distinct records must be maintained, for example, channel separation. |
|
Disconnected records: Transactions relate to a customer but aren’t connected to a customer record. |
Relationship gaps across objects or systems, for example, cases not linked to a contact or guest orders without a customer ID. |
Assess if there are other identifiers in the transactional record, such as verified personal contact points or loyalty numbers. Establish relationships using deterministic rules, often using Data 360 data transforms. |
|
Invisible duplicates: Two records can’t match within a single object because each is sparsely populated (for example, one only has phone and another only has email). They become detectable only when combined with other sources. |
Match records from the source object to unified customer records. Quantify source records that match to a unified customer record, even though they were not identified as duplicates within the source object. |
Use Data 360 to create unified profiles, then identify which source records share the same unified profile ID. If these records are unintentional duplicates that you previously identified as invisible duplicates, you can clean them using merge tools. Be mindful of shared contact points across individuals that can impact match quality. |
Why NTO’s Order Header Is Also Customer Data
NTO wants to measure and use total lifetime value (TLV) to improve how cases are handled and routed. Luna quickly sees that TLV depends on having a full view of each customer’s activity. However, some customer details aren’t stored on contact or account records.
To get a complete view, she needs to look at transaction records. These records also contain important customer data. For example, in ecommerce systems, order data is often spread across multiple records. Order header has order day, buyer, payer, receiver, order data and amount information. Order line details have quantity, individual item, and unit price.
To understand a complete understanding of its customers, NTO needs to not just review ecommerce profiles but data that is part of orders, including guest orders that are not linked to a customer profile. To calculate TLV correctly, all orders can be matched and connected to existing customer profiles. Instead of merging the transactional records, NTO uses unified profiles to create a link between the guest activity and the known customer.
Differentiate Data 360 Key Ring Versus MDM Golden Record
Now that you understand the risks of having gaps in customer context, how to detect them, and how to address them, it’s time to examine two approaches organizations use when they want to unify customer profiles.
All organizations want to have a contextual understanding of their business relationships. For example:
- Who is the customer, and what are all of their interactions with marketing, sales, and service?
- Who are the job applicants, and have they applied previously?
- What are the different business relationships with an organization—are they a customer, supplier, or partner?
This requires all data in siloes to be related to a given organization and individual for the contextual views to be presented to different user or stakeholder groups based on their business need and compliance requirements.
“Most CRM match-merge efforts are well-intentioned but fail. Unmerge is expensive, and risk aversion can cause projects to halt midstream. Newer solutions, such as identity resolution, give business users what they want: a consistent understanding, without loss of context or worry about undoing incorrect merges.”
—Mehmet Orun, Datablazer and Salesforce MVP
Systems that deliver the right information in the right context, while respecting security and access controls, are systems of context. Salesforce Data 360 is designed to support this need.
Rather than forcing everything into a single record, Data 360 uses a key-ring model that links a person or account across multiple source systems while preserving the original context and traceability of those records. This approach lets organizations maintain multiple business contexts while still providing users with a unified customer view when needed.
To understand how these records stay connected without being merged, it helps to understand the role of a universally unique identifier (UUID).
Each record in this key ring is linked by a UUID, which serves as a consistent reference across systems. A UUID is a system-generated value that uniquely identifies a person or entity, even if records belonging to it exist across multiple sources. With a UUID, Data 360 can relate records without merging them or losing their original context. At NTO, this means customer service reps can access a single customer profile brought together from Service Cloud contacts and cases, Commerce Cloud customer profiles and guest orders, and point-of-sale transactions connected through a loyalty program.
A contextual, unified profile isn’t sufficient for some organizations, especially organizations in regulated industries that require a single authoritative source of truth, or golden record, that represents a customer across all systems. For example, a life sciences company must ensure that a medical professional holds a valid license, and a bank must understand a customer’s identity across multiple relationship types. In these situations, where stronger governance and stewardship are required, master data management (MDM) solutions are designed to meet those needs.
MDM systems can work alongside systems of context. For example, when paired with Data 360, Informatica MDM can maintain the governed global customer identifier and managed master profile, while Data 360 provides the contextual view that brings together customer records, transactions, and interactions across operational systems.
Examine False Customer Matches
A false match, often used interchangeably with false positive, occurs when the match process incorrectly identifies two records as a match when they should not be matched in the real world. This usually happens because the data used to match them is wrong, missing, or misleading.
In identity resolution, this means the system works as designed—but the data leads it to the wrong conclusion.
False matches are risky because they combine the wrong records. This creates a false view of the customer and can lead to:
- Wrong service decisions
- Incorrect customer groups (segmentation)
- Unreliable AI results
Common Causes
- Bad or misleading data: For example, placeholder email like na@na.com, shared phone numbers, or reused company contact info)
- Unusual values used too often: For example, the same address appearing across many different customers
- Match rules that are too broad: For example, fields that are not unique enough
False Match at NTO
At NTO, two customers—Sam Smith and Samuel Smith—both had the email na@na.com. Because of this shared value, their records were merged.
As a result, an agent could see Sam’s information when helping Samuel or the other way around.
In industries like healthcare or life sciences, this kind of mistake could lead to serious consequences for both the customer and the organization.
To prevent false matches, Luna identifies outlier values in NTO’s customer data sets and designs a process to exclude them from the identity resolution process.
What’s Next?
It’s time to move on to the next unit to explore how to handle inconsistent and unreliable values.