Translation memory leverage sheet

The translation memory leverage sheet costs a release from your own analysis rather than a published rate card: the source word count of the strings that changed, the locales the product ships in, the share of those words the memory matches exactly and the share it matches fuzzily, and the rate the vendor charges for a new word, a fuzzy match and an exact match. It returns the new, fuzzy and exact words per locale, the cost of the release per locale and across every locale, what the release would cost with no memory at all, the saving the memory delivered and that saving as a share of the full cost, which is the leverage, and the effective rate per source word. It is the sheet a localisation engineer builds in a spreadsheet every release, with the arithmetic written down so the budget review can argue with the inputs instead of the answer.

Why the sheet asks for match shares, not a total

A vendor's quote is a word count times a rate, and it hides the one thing the team controls: how much of the release the memory already holds. A release that is a third exact matches and a quarter fuzzy matches costs a fraction of one that is all new words, at the same headline rate, and the analysis report every translation tool produces gives those shares before anybody agrees a price. The sheet takes each share and its rate, so the effect of a rate negotiated on fuzzy matches, or of a memory grown by one more release, can be tried on the page before the purchase order is raised.

Leverage is a percentage until it is a budget

The saving the memory delivered is the difference between the release at the full rate and the release as the analysis splits it, and the share of the full cost that saving represents is the memory's leverage. On the worked example a release that is a third exact and a quarter fuzzy across six locales costs well under three quarters of what the same words would cost with no memory, and that gap is the argument for keeping the memory as a record rather than as a file the vendor holds. Xlifflane Pro keeps every translated segment against its key so the next release's analysis reads the team's own memory, not the vendor's.

The effective rate is the figure to compare vendors on

Dividing the release's cost per locale by its word count gives the effective rate per source word, which is the honest figure when two vendors quote different rates for new words and different discounts on matches. A vendor with a higher new-word rate and a better fuzzy discount can be cheaper on a release the memory holds well and dearer on a release that is mostly new strings, and the sheet shows which one this release is. The cost with no memory beside it is the ceiling the whole conversation sits under.