A type 1 dimension keeps the current attribute value by updating the existing row. A type 2 dimension preserves changes as separate records with effective dates and a current-row indicator. The business question decides which history you need. Correcting a typo and tracking a customer's move may require different treatments.
A MERGE is useful for matching dimension keys, but it doesn't invent historical semantics. For type 2, a changed current row must close its validity interval and a new row must open. An unchanged row should not create another version. Use one stated timestamp boundary and avoid overlapping active intervals. These exercises build the comparison and current-value update first. The project then performs the close-and-open sequence and checks the resulting history, including one current row per customer. A production pipeline also needs an atomic or recoverable strategy across its writes.