In Sajan's Open Challenge on TM Benchmarking, "We'll Let the Client Decide"

Article Details

Global Watchtower

Common Sense Advisory Blogs

Posted by Benjamin B. Sargent on March 14, 2008 in the following blogs: Translation and Localization, Technology, Market Data
Pages \| \|

Sajan was founded in 1997 as a software company, but quickly decided to get into the business of translating words, too. This turns out to have been a wise decision. One after another, virtually every technology company in the language area has bitten the dust, with each carcass being picked up by a language services provider: eTranslate, Global Sight, Logoport, Uniscape, Trados, Idiom. Cashflow from translation services provides the R&D budgets to maintain investment in software products for this industry. To date, there just doesn't seem to be enough of a market to support independent globalization software vendors beyond the startup phase. The fact is, initial investors in each of these companies grew impatient before they achieved positive cashflow or sustainable business models. Trados may be an exception -- in that it did achieve positive cashflow -- but even it decided there was no future in desktop software and the company was sold to SDL after repeated failures to gain a foothold in the enterprise market.

SDL succeeds, by financial measures. As much as 30% of SDL's revenue may now be attributable to software sales -- including some non-globalized CMS -- following last year's acquisition of Tridion. SDL's gross margins now exceed 50%. The company credits its technology platform with helping to shift work into regional production centers -- in other words, low cost countries -- where a 27% increase of "revenue throughput” only entailed a 10% growth in headcount. That difference alone supports a healthy R&D spend. SDL showed 8% organic growth overall, according to its public statements.

Lionbridge also sees its investment in technology as a positive. It reported 2007 revenue of US$452.0 million, an increase of 33.1 million or 7.9% from 2006. More importantly, the company increased R&D spending almost 17% on the year, from US$2.8 million to 3.4 million and announced it would further increase that number in 2008.

Sajan claims 140% growth in 2007, with gross margins over 60% -- though the company is still small enough to escape notice in our "Ranking of Top 20 Translation Companies." Sajan does not report its numbers publicly. Licensing technology accounts for less than 15% of revenues because Sajan lets clients use the main component, Global Communication Management System or GCMS, free of charge. Another difference between Sajan and SDL is that it started building its software for enterprise clients from day one; translators were never the main audience. In 2003 the company substantially rewrote its software from the ground up to take advantage of the on-demand model. GCMS is now a multi-tenant application in which all clients use a single instance of the software.

The company recently announced its new TMate Search technology, a "search and match" engine that serves as a stand-in for translation memory. Based on a single repository rather than a concatenation of individual TMs, TMate's contextual indexing returns results quickly, even on very large data sets. The company says the technology includes a "self-learning agent" developed by a third party but adapted to the translation environment. The primary benefit comes from improved relevance in fuzzy matching. In announcing the new capability, Sajan offers clients a challenge to benchmark TMate Search against whatever translation memory tool they are currently using. The company is setting client expectations for "10-60% net new matches above any existing TM technology on the market today."

Various parties have proposed TM benchmarking in the past, but the idea has not caught on, in part due to the non-trivial issue of how to create a fair and balanced playing field. The novelty here is that Sajan has side-stepped that problem. By laying the challenge at the doors of translation buyers, rather than to other service or technology providers, there is no question of fairness or levelness. The enterprise translation buyer only cares about their unique content, in the end, and they will set the standard and assess the results, with no need for an impartial judge. So, for instance, when we asked how will "fuzzy match" be defined? The answer was: "We'll let the client decide."

Categories