Terminology: Rollup, Consolidate, Aggregate

[John Hammond]

Folks

Here's how we use them:

Rollup - Aggregation when a dimension is removed (shrunk really).
Consolidation - Aggregation when a level of hierarchy is removed (hierarchy shrunk a level).

Are we going against the grain or is there a TM1 approved terminology everyone uses - which we will willingly follow.

[Martin Ryan]

I generally use rollup as a verb and consolidation as a noun. As in "you can rollup these elements into this consolidation".

However there is the "Rollup" button in the subset editor (provided it's opened from the cubeviewer) which creates an ad-hoc consolidation.

Cheers,
Martin

[Alan Kirk]

Folks

Here's how we use them:

Rollup - Aggregation when a dimension is removed (shrunk really).
Consolidation - Aggregation when a level of hierarchy is removed (hierarchy shrunk a level).

Are we going against the grain or is there a TM1 approved terminology everyone uses - which we will willingly follow.

I generally agree with Martin. Personally I only use the term "roll-up" when referring to ad hoc user-defined consolidations as opposed to those defined in the dimension metadata, but it does have a wider meaning than that. For example an account "roll-up" in the PeopleSoft ledger system is essentially the same as a "consolidation" in TM1.

However consolidation and aggregation really aren't the same thing, and are at best tenuously connected. A consolidation doesn't really aggregate data (at least in the sense that TI, for example, uses the term) it simply adds up the individual elements. And even that assumes that the elements of the consolidation all have a standard weighting of 1, rather than a negative or a fractional weighting. Some consolidations (though they may be rare) therefore don't really "aggregate" values at all.

Aggregation is something that I'd regard as any action which requires:
- Reading the value that's already stored at the intersection of a collection of elements;
- Adding the value of the current data row to that; then
- Writing the total of the two back to the cube.

This may indeed be the case where you're writing from a more detailed cube (which would typically have more dimensions) to a less detailed one (which would typically have fewer), however it could also apply to any data source - relational database connections, text files, whatever - which obviously don't have dimensions to remove. The process of aggregation (certainly in the TI sense) applies to any situation where you may need to write more than one value from a data source to the same combination of elements in a cube.