Dimension columns for multiple values

Hi _luke_ ,

Yes, this is absolutely a valid case for snowflaking. In fact, it’s one of the few scenarios where snowflaking makes your model more efficient, not less. You have demographics that are single-response (Demog1 to Demog5), which fit nicely into your main dimCustomer table. But for Demog6, Demog7, and Demog8 — where a customer can have multiple values — modeling them as separate dimension tables linked to dimCustomer is the cleanest and most scalable approach.

Trying to cram multi-response values into delimited text columns in the customer dimension would create headaches for filtering and aggregation. Power BI doesn’t handle delimited fields natively — so filtering on "contains" or splitting them out dynamically for proper visual use becomes a nightmare. A better solution is to create bridge tables for these multi-response demographics.

For example, for dimDemog6, you can have a table structured like this:

Then, your dimDemog6 dimension might look like:

This allows customers to be associated with multiple Demog6 values without repeating rows in the customer table or using delimited strings.

Once you've linked it up:

• BridgeDemog6[CustomerId] → many-to-one → dimCustomer[CustomerId]
• BridgeDemog6[Demog6_Value] → many-to-one → dimDemog6[Demog6_Value]

Now in Power BI, if you want to count how many customers selected each Demog6 value, you can use:

Customer Count (Demog6) =
CALCULATE(
    DISTINCTCOUNT(dimCustomer[CustomerId]),
    TREATAS(VALUES(dimDemog6[Demog6_Value]), BridgeDemog6[Demog6_Value])
)

Or, if you're using bi-directional filtering between the bridge table and dimCustomer, you can simplify it to:

Customer Count (Demog6) = DISTINCTCOUNT(dimCustomer[CustomerId])

when filtered by a slicer from dimDemog6.

So yes, even if snowflaking has a reputation problem, in this context it's doing exactly what it should: allowing for multi-response attributes to be modeled cleanly and queried effectively without delimited strings, wide tables, or DAX acrobatics. You’re modeling reality more faithfully, and Power BI will thank you for it.

Best regards,