Skip to content
Artificial intelligence. Clearly. RSS feed
AI NEWS AI

News. Evidence. Perspective.

BusinessAnalysis

eBay’s 2014 translation upgrade helped sell more goods across borders

A 2014 rollout offers an early example of AI changing purchases across borders. The economic estimate concerns item titles on one marketplace.

A listing can be available overseas without being easy to buy from. Someone still has to understand what the seller is offering. Translating those few words can change whether an existing product finds another customer.

In 2014, eBay upgraded its English-to-Spanish item-title translation. A 2019 Management Science study by Erik Brynjolfsson, Xiang Hui and Meng Liu estimated a 10.9% increase in the quantity of U.S. exports sold through the platform to 18 Spanish-speaking Latin American countries.[1]

eBay’s launch account helps reconstruct the audience. More than two million Latin American buyers were already shopping on its English-language, U.S.-dollar site, the company said. Its localized experience added Spanish and Portuguese, local promotions and prices displayed in local currencies. These were people already navigating a service built around another language.[2]

The language inside a listing

Product titles have their own vocabulary. In its technical account, eBay described statistical translation models trained for the noisy language people use when listing goods. Brand names and alphanumeric product specifications need to survive the translation. Abbreviations need their commercial meaning: NIB means new in box. The task requires understanding what a seller is saying about an item while preserving the identifiers a buyer needs.[3]

A fluent sentence that changes a model number can describe the wrong product. A clumsy phrase that accurately communicates condition and compatibility may still answer the buyer’s question. That is why the translation of a catalogue deserves attention as a distinct commercial task.

Following the purchases

The researchers studied 14 months around the July 2014 title upgrade, comparing export changes for longer and shorter titles within the same countries. Query translation and other localization changes had arrived in May. Their design assumes other concurrent changes did not affect sales differently by title length. The 10.9% figure is the authors’ overall estimate from that relationship.[1]

The upgrade replaced an existing translation service, so the finding concerns better translation. The estimated effects were larger for cheaper and less standardized goods, and for less experienced buyers.[1]

That makes the historical result useful when thinking about AI-led growth. The seller’s next customer may already be searching the marketplace. Making the existing offer intelligible can help that buyer decide to purchase.[1]

Follow the evidence

Sources & further reading

  1. Does Machine Translation Affect International Trade? Evidence from a Large Digital Platform Erik Brynjolfsson, Xiang Hui and Meng Liu / Management Science · 3 Sep 2019 · Accessed 19 Oct 2026