MealWords

Localization is the rare agency service where the underlying platform does most of the labor, which means the margin lives entirely in the workflow wrapper: terminology governance, QA, and version control. Agencies that own that wrapper sell a repeatable global-ready retainer; agencies that resell a translation subscription compete on price against the client's own CMS plugin. The window to claim the wrapper is open now because AI translation quality has crossed the threshold where clients notice the difference between governed and ungoverned output.

Locale debt is the accumulated backlog of untranslated, hardcoded, or context-poor strings that grows every sprint a product ships without a localization pipeline. Like technical debt, it charges interest: each new market multiplies the surface area that must be retrofitted, and each release widens the gap between source and localized versions. Agencies inherit this debt when they take over a client's multilingual estate, and it is usually invisible in the pitch deck. The framework asks one question before quoting a global-ready retainer: how many strings exist, how many are already localized, and who owns the glossary? Platforms such as Crowdin and Transifex reduce future accrual through continuous sync with repositories, while Phrase adds quality scoring so drift is measurable rather than assumed. A client launching into eight languages with no terminology base will pay for the same review work repeatedly. Pricing the cleanup as a separate discovery phase protects margin on the ongoing retainer.

Terminology Lock-In is the degree to which a client's brand vocabulary, product names, and approved translations live inside one localization platform's glossary and translation memory rather than in a portable asset the agency controls. Every approved string compounds the lock: a glossary of 4,000 terms with reviewer sign-off across eight locales is expensive to rebuild and cheap to keep. Agencies that treat the glossary as client-owned IP (exported, versioned, documented) can move a retainer between platforms or renegotiate pricing from strength; agencies that let it accumulate inside a single vendor's memory own nothing and are effectively reselling. Lokalise routes copy through custom AI profiles that enforce brand terminology and tone, and DeepL's glossary system does similar work at document level, which is exactly why the glossary, not the subscription, is the asset worth auditing at renewal.

Localization Margin Ladder ranks where an agency sits on the value chain: reselling a subscription, running the translation pipeline, or owning the terminology and QA layer that governs every language the client ships. Each rung up the ladder raises gross margin and switching cost, because the client's brand glossary, review rules, and release cadence become agency assets rather than vendor settings. The ladder matters because AI translation has collapsed the price of raw word conversion, so the only defensible retainer is one tied to governance the client cannot replicate in-house. A concrete example: Lokalise routes copy through custom AI profiles that enforce brand terminology and tone, which means an agency that authors those profiles owns the standard every future translation is judged against. Phrase adds quality scoring on top of workflow automation, giving agencies a measurable QA artifact to attach to monthly reporting. Without that layer, a client can swap vendors in a week and the retainer disappears with them.

Treat the translation platform as the engine and sell the terminology, review, and QA layer you own, because a pipeline anyone can subscribe to is a price, not a service.

Build the review, terminology, and release workflow as agency-owned IP before quoting multilingual delivery, and price the retainer on that workflow rather than on platform seats or word counts.

IF a client's multilingual surface is a single marketing site refreshed a few times a year, THEN resell a website translation layer and keep the margin thin but the delivery cost near zero. IF the client ships product strings, help docs, and campaign copy across three or more locales on a continuous cadence, THEN own the workflow: terminology, QA gates, and release process built on a translation management platform, because that is the only part a client cannot buy directly from a vendor.

A fixed-scope engagement that takes one client property (site, app, or product surface) from single-language to launch-ready in two to four target locales, with version-controlled strings, a terminology glossary, and a QA gate that a client can audit.