7. Assessment and Calculation
How the two percentages and their grades are calculated.
7.1 Converting prices into comparable ratios
Prices quoted per hour, per gigabyte-month, and per million requests cannot be combined, but ratios formed from them can. For each product category in the comparison basket, the median price across the provider's regions is taken as the reference, and every region's price for that product category is divided by it to give a price ratio. A ratio of 1.10 describes a price ten per cent above the median for that product category, and the statement means the same thing whichever unit the underlying price was quoted in.
7.2 The relative cost percentage
A region's relative cost percentage is the median of its price ratios across the basket, restated as a percentage difference from the median. A region ten per cent above the median on compute, five per cent below on storage, and twelve per cent above on databases resolves to a value between those positions. A published value of plus nine per cent therefore describes a region that is typically around nine per cent more expensive than the median region of that provider across the basket, and minus eight per cent describes a region typically around eight per cent less expensive.
Every product category in the basket carries equal weight in that median. The score is a comparison of catalogue prices and not a projection of any customer's bill, and Section 11 sets out what follows for a reader.
7.3 Evidence sufficiency and suppression
A region that prices only a small share of the comparison basket would produce a median resting on too few ratios to defend. Where a region falls below the sufficiency level the methodology sets for that purpose, no relative cost percentage and no cost grade are published for it, and the fields are left empty. The region keeps its row on the scorecard and keeps its coverage score, which does not depend on the basket.
The basket coverage percentage is published for every region, including those above the sufficiency level, so that a reader can see how much of the basket any published cost score rests on. A region scoring on most of the basket and a region scoring just above the sufficiency level both carry a cost score, and the basket coverage percentage is what separates them.
7.4 The service coverage percentage
The coverage score counts product categories instead of comparing prices. The eligible product categories a region prices are counted, and the total is divided by the count for the provider's most complete region, which by construction scores one hundred per cent. A region pricing four thousand of the leading region's five thousand eligible product categories scores eighty per cent.
The score therefore describes completeness relative to the best-stocked region of the same provider, and not relative to the provider's full theoretical catalogue or to any other provider's.
7.5 Banding the percentages into grades
Each percentage is placed into one of five bands using Jenks natural breaks, so that the band boundaries fall at the gaps that exist in the distribution. The band is published as a numeric grade from 1 to 5, with 1 the least expensive cost band and 1 the most complete coverage band.
The grade is stored with the rest of the row and is not computed when a reader loads the scorecard, because the boundaries are fitted afresh for each provider in each data version and cannot be reconstructed from a percentage alone. How a grade is presented, whether as a symbol, a word, or a colour, is a display decision taken outside this methodology, so a change of presentation does not require rescoring. The band boundaries themselves are recorded for each run under Section 9 but are not published on the scorecard.