Microsoft MB-260 Retired: When Customer Profiles Collide

A retail company identifies the same customer in its ecommerce platform, loyalty application and support database. One system stores an abbreviated name, another uses a shared household email and a third has an obsolete address. Combining every matching field into one profile seems attractive, but it can merge two different people or preserve an outdated consent decision.

Microsoft MB-260, Dynamics 365 Customer Insights (Data) Specialist , retired on November 30, 2024. Its historical focus on ingestion, profile unification, segmentation, predictions and administration is still relevant to customer data architecture.

Identity resolution is a policy

Matching is not simply a mathematical score. A strong identifier such as a verified customer account may deserve more weight than a common surname, but even strong identifiers can be reused or mistyped. Define which fields are trustworthy, how conflicts are handled and when a proposed match needs review. Customer Insights data unification depends on a documented strategy that the business can explain and audit.

Build a miniature dataset containing a parent and teenager sharing a telephone number, two unrelated people with similar names and a customer whose address has changed. Try several matching rules. Count both false merges and missed joins, and decide which error is more damaging. The correct threshold depends on the downstream use: a recommendation system may tolerate uncertainty that a regulated account statement cannot.

Source preparation shapes the profile

Raw customer records often differ in timestamps, country codes, names and status values. Power Query and other transformation approaches can standardize fields before unification, but normalization can also erase useful distinctions. A mailbox shared by several employees should not necessarily become one individual identity; a company account may need a different data model from a consumer profile.

Inspect source freshness and lineage. A nightly CRM export and a near-real-time commerce event stream may disagree temporarily about the customer’s latest purchase or preference. Decide which source has authority over each attribute. When a support case closes, should that update the customer profile instantly or after validation? The data-pipeline decision determines what downstream teams are able to know and when they are allowed to rely on it.

Consent must survive unification

A unified record can bring together email addresses and marketing preferences from multiple systems. It must not convert uncertain or revoked consent into permission by accident. Record origin, purpose, update time and applicable channel when designing privacy-sensitive attributes. A customer may permit service notifications yet reject promotional outreach; collapsing those values into one generic opt-in field creates an operational and compliance risk.

Simulate conflicting consent timestamps from a web form and an older loyalty database. Determine how the latest valid preference becomes authoritative and how the system explains the decision. Check that exports, segments and connected destinations respect the result. A unification pipeline that gets names right but permissions wrong does not create a trustworthy customer view.

Segments should describe real populations

After unification, marketers may define segments based on recency, purchase frequency, product interest or service history. Measures and predictions can make these groups more useful, but only when their definitions are stable. A predicted churn score is not proof that a customer will leave. Teams should distinguish measured behavior, inferred likelihood and assumptions about what intervention will help.

Test a segment for customers with repeat purchases and an unresolved support issue. Validate it against sample records, including returns and duplicate activity. Ask whether a planned promotion should exclude customers awaiting a refund. Customer data preparation is not a neutral step: the definitions chosen upstream directly influence what customers experience downstream. Review how segment membership changes when late events arrive.

Governance continues after the model runs

Profile unification and refresh schedules need ownership, monitoring and change management. A revised matching rule can affect millions of identities and invalidate past metrics. Measure match quality and report the effect of changes before deployment. Apply the least privilege needed for data analysts and external integrations, and establish retention and deletion processes that include derived profiles and exported audiences.

A useful historical MB-260 exercise starts with three disagreeing sources and ends with a defensible customer view: matched identities, precedence rules, consent boundaries and testable segments. The certification itself is retired, but those judgments remain central to customer data work. The distinction matters because professionals should invest in current product documentation while learning durable design principles from legacy objectives.