When unrelated customers merge into one profile: identity graph collapse

Why this matters

This matters because when unrelated customers are merged into a single profile, personalisation breaks down. The identity graph incorrectly connects people through shared values like a shop’s phone number or a placeholder like test@example.com, resulting in one profile that contains data from multiple individuals. This distorts audience analytics—audience sizes become inflated or deflated—and leads to misinformed decisions in targeting and messaging.

Without proper identity hygiene, campaigns lose relevance. For example, a customer segment that should include only loyal buyers might instead include random users linked only by a shared placeholder. Checking how many identities a single profile links to quickly reveals such issues. The risk is not just data inaccuracy—it’s a loss of trust in analytics, which undermines campaign performance and customer insight.

The key ideas

The identity graph connects identities that appear together in the same record or event. When a shared value—like a family email or a shop’s phone number—is used, it links multiple people into one profile. This creates a collapsed identity graph where different individuals are merged into a single profile. As a result, personalisation fails because the data doesn’t reflect real user behaviour. Audience counts also become inaccurate—either inflated or deflated—because the profile includes data from unrelated people.

A common cause is the use of placeholder values, such as test@example.com. These values are not real identities and should be removed or left empty before data is sent to the system. If sent as identities, they can trigger the identity graph to merge unrelated users. Shared devices can cause similar issues with browser identities like ECID, where multiple users on the same device are incorrectly linked.

To catch this early, check how many identities a single profile links to. A high number of linked identities may signal a collapsed graph. Removing or eliminating placeholder values during ingestion stops the identity graph from incorrectly connecting people. This ensures profiles remain accurate and personalisation remains effective.

How to apply it

Before ingesting data, remove or leave empty any placeholder or default identity values like test@example.com or a generic shop phone number. These values act as weak links in the identity graph and can unintentionally merge unrelated customers into a single profile. When such values appear in data, the identity graph connects all individuals who share them—regardless of whether they are actually related—leading to a collapsed graph where multiple people’s data is mixed together.

This mix breaks personalisation and distorts audience sizes, making analytics and targeting unreliable. To catch such issues early, check how many identities a single profile is linked to. If a profile has many identities from the same source—especially common ones like a shared email or device ID—it may indicate a collapse.

For shared devices, treat browser identities like ECID with caution. If the same device appears across users, it can mimic a shared identity. Always validate identity values at ingestion and remove defaults. By enforcing clean, unique identities from the start, you preserve accurate customer profiles and prevent the identity graph from incorrectly linking unrelated people.

Mistakes to avoid

• Sending placeholder values like test@example.com or default phone numbers as identities causes unrelated people to merge into one profile. These values appear in the same record, triggering the identity graph to link all individuals sharing them. This results in a collapsed graph where multiple people’s data is mixed, breaking personalisation and distorting audience sizes.

• Shared devices can produce similar issues, especially when browser identities like ECID are used across multiple users. To avoid this, remove or leave empty any placeholder or default values before data ingestion. Regularly check how many identities a single profile links to—this reveals graph collapses early. Always validate identity values to ensure they uniquely represent individuals, not shared or generic data points.

Quick checklist

• Remove or leave empty any placeholder or default identity values (e.g., test@example.com) before ingestion
• Verify that shared values like family email addresses or shop phone numbers are not used as identities
• Check if browser identities such as ECID are coming from shared devices that could link unrelated users
• Review profile-level identity counts to detect if a single profile links to too many distinct identities
• Ensure no common values (e.g., public or generic identifiers) are being used as identity graph links
• Validate that identity graph connections are based on unique, personally identifiable attributes
• Confirm that identity values are not duplicated across events or records in the same session
• Run a test to identify profiles with more than one linked identity to detect potential collapse