A data structure that stores a measure pre-summed across several dimensions, such as month, region and product, so any slice is quick to read.
An OLAP cube (online analytical processing) holds one or more measures, such as revenue or units, organized by dimensions, such as time, region, customer and product. Each dimension can have a hierarchy, for example day to month to quarter. Users slice to one value, dice to a subset, and drill up or down a hierarchy without writing queries.
Facts from transaction tables are aggregated to every combination of dimension members the cube stores. The cell count is the product of the member counts, before any empty cells are dropped.
A cube with 12 months, 5 regions and 40 products has 12 x 5 x 40 = 2,400 cells at its lowest level. Summing the 480 cells for one region (12 x 40) gives that region's annual revenue.
Measures that cannot simply be added, such as distinct customer counts or margins, summed anyway; and a cube refreshed overnight that disagrees with the ledger by afternoon. The full guide is OLAP.