What is Translation Memory?

Translation Memory (TM) is a database of previously translated segments — sentences, paragraphs, UI strings — that a translation tool can reuse the next time the same or similar text appears. For ecommerce teams scaling content across many locales, it is the difference between linear and sublinear translation cost.

Definition

A TM stores source-target pairs together with metadata such as locale, domain, glossary tags, and approval status. When a new piece of content enters the workflow, the tool segments it and queries the TM for exact and fuzzy matches; fuzzy matches are scored by similarity, often 75% to 99%. Translators only review or edit deltas, while exact matches can be applied automatically. TMs are typically paired with a termbase, which enforces brand-specific glossary terms across all locales, and they integrate with CMS systems through connectors or APIs.

Why it matters

Ecommerce content is highly repetitive: product attributes, size guides, return policies, and shipping disclaimers recur across thousands of items. Without TM, each new locale and each catalog update triggers full re-translation. With TM, costs drop dramatically and consistency improves — brand voice stays uniform across de-DE, de-AT, and de-CH, and legal disclaimers stay synchronized across markets. For Composable-Commerce teams, TM also reduces time-to-market: a new product can ship into ten locales within hours rather than weeks because most of the boilerplate is already translated.

Use cases

A retailer connects its Headless CMS to a translation management system that pushes new product copy through TM, returns pre-translated drafts, and routes only the uncertain segments to human reviewers. A marketplace uses TM to keep checkout strings perfectly aligned across forty locales, ensuring that "Add to cart" or "Proceed to checkout" never drift. A D2C brand combines TM with AI Generated Content to draft first versions in target locales, then refines them via human-in-the-loop review backed by the existing termbase. Translation Memory also pairs well with Locale Fallback: if a fr-CA variant is missing, the system can serve fr-FR while flagging the gap for translation and capturing the eventual delta in the TM.

Related

Explore Multi-Brand and Multi-Market.