Loading CSVs into a Delta table with optimization is what the Load to Tables feature is designed to do. It can take the CSV files found in Subfolder1, convert them into a Delta table, and apply V-Order (Z-order) optimization during the load so related data is co-located on disk. This makes queries that filter on the most commonly used columns faster, because the data layout is arranged to minimize reads. The other options don’t provide this combined ingestion-and-optimization flow. Copying to Delta usually focuses on moving or duplicating existing Delta data, not creating a new Delta table from a set of CSV files with V-Order. Export Data is about moving data out of the lakehouse, not converting a folder of CSVs into a Delta table with optimization. Convert to Delta via CLI is a separate conversion step that generally doesn’t integrate the V-Order optimization during the ingestion from a directory like Subfolder1.

When querying a tabular model through an XMLA endpoint, you want a query that returns exactly the rows and columns you need, without pulling unnecessary data. The best approach is to start with EVALUATE, then build the shape of the result with SUMMARIZE to pick only the Store columns you want, and finally apply FILTER on the specific field (such as OpenDate) to restrict the rows. This produces a compact, filtered table as the query output, which is precisely what the endpoint should return. Using SUMMARIZE to select just the needed columns keeps the result lean, rather than returning the entire base table. Applying a FILTER inside the query ensures the data is filtered server-side, reducing data transfer and improving performance. In contrast, filtering SAMPLEdirectly on the base table without shaping the output can return more columns than necessary, and adding all columns with ADDCOLUMNS would swell the result set. CALCULATE without a filter wouldn’t produce a filtered result at all.

To build a new column whose values come from a condition, you create a derived column by adding a new column and defining its value with a conditional rule. The tab that contains this capability is Add Column, where you can choose Conditional Column (for if/then/else rules) or use Custom Column for a broader M expression. The other tabs serve different tasks: Home for general actions, Transform for altering existing columns, and View for UI options. So the Add Column tab is the place to create a derived column based on a conditional statement.

Preserving historical states of a dimension to enable point-in-time analysis is the idea here. When a dimension attribute changes, a new version of the row is created instead of updating the existing one. This is the approach of a slowly changing dimension of type 2: each change results in a new row with its own surrogate key and a validity period, often represented with start and end timestamps (or a current flag). With this setup you can look back at any date and reconstruct how the dimension looked at that moment by selecting the version whose validity contains that date. This works well for analyses that need to know attributes as they existed in the past, such as a customer’s status or product attributes at a specific time. It’s different from updating the current row (type 1), which overwrites history; it’s different from storing previous values in separate columns for only a subset of attributes (type 3); and it’s different from not tracking changes at all (type 0). That combination of versioned rows and timestamps is why this approach is chosen for point-in-time analysis.