Finally, we will expand the Time dimension to make the workings of our Stock Level measure readily apparent, with regard to how it accumulates value over time.

The Year level expands, revealing the underlying Quarter level.

The Data Grid now resembles that shown in Illustration 18.

Let's take a look at a zoom of a single year to get a feel for what is happening in our Data Grid. As we can see from the two annotated examples below, the Year 1997 Total (the "sum of the 2's," in Illustration 19) is a simple total of the four Quarters that make it up, and the Q1 Total (the "sum of the 1's," in Illustration 19), represents a simple total of its member Months 1, 2, and 3. This behavior is consistent throughout the dataset, although we use Year 1997 only in our examples in Illustration 19.

While this simple summing would be quite appropriate for many measures (such as sales, expense, and other common income statement accounts, among others), semi-additive measures such as Stock Level do not, as we have discussed, appropriately add up in this way. For this reason, we need to make an adjustment in our calculated member that takes into consideration the fact that we wish to present the balance at the end of each time period as that of the last member of the respective period (say Month 3 of Q1, or Q4 of Year 1997, in our two illustrated examples above),

Let's make our adjustment now, and ascertain the correct results with the modified Stock Level calculated member.

(ClosingPeriod([Month]),[Measures].[Units Ordered]) - 

    (ClosingPeriod([Month]),[Measures].[Units Shipped])

Analysis Manager generates a message box, indicating that "Syntax is OK", once more. (If this is not the result, check your syntax input).

The Calculated Member Builder appears, with our modifications, as depicted in Illustration 20.

The Calculated Member Builder closes, and we are returned to the Data view of the Cube Editor. The Data Grid should reappear, with new values to reflect our changes, as shown in Illustration 21.

A quick review of the values reveals that we have, indeed, accomplished our objective, modifying the behavior of Stock Level measure to reflect its intended, semi-additive nature. As we see in the subset of the data, wherein we highlight parallel examples to those we noted above, the balance found in the "last" member (for both Quarter and Year) is the value retained as the "final" balance of the respective parent.

NOTE: For more information on the ClosingPeriod() function, see my Database Journal article MDX Time Series Functions, Part II: The OpeningPeriod () and ClosingPeriod() Functions.

We are returned to the Analysis Manager console.

Conclusion

In this article, we explored the nature and management of semi-additive measures, creating a calculated measure that is not fully additive, to meet the business requirements of a hypothetical group of information consumers. After discussing the nature of semi-additive measures, and describing uses for which they are appropriately employed, we created and processed a copy of the Warehouse sample cube in preparation for our practice example. We next established an illustrative set of business requirements as a specification for creating a semi-additive measure (a calculated member) in our practice cube.

We explored an initial approach to creating a simple inventory balance calculated measure, and explained why it was ineffective for our needs, due to its fully additive nature. We then modified the calculated measure to exhibit the desired semi-additive behavior of an inventory balance measure. Throughout the practice example, we discussed the results datasets obtained, within each of the steps we undertook to meet the expressed business requirements, verifying the effectiveness of our final solution.