Enthusiasm for the "Economic Theory of Index Numbers" seems to ebb due to its inherent operational problems of index compilation. The alternative concept of inflation measurement is known as "cost-of-good" or "(fixed) basket approach". Though more convenient in practice it may have a less impressive theoretical underpinning. It is not uncommon even to call its rationale in question altogether. This paper attempts to justify the basket approach and to specify "pure comparison" in operational terms. In essence this requires a focus on a time series of indices (as opposed to comparing two adjacent periods in isolation), and the interpretation of the index formula in terms of expenditures.