Blog Post

Startup Cloudwords attacks $30B translation market

Stay on Top of Enterprise Technology Trends

Get updates impacting your industry from our GigaOm Research Community
Join the Community!

Tech companies can easily spend millions of dollars per year (or a hundred times that) translating their software, documentation and related materials for different markets in a process that has thus far avoided automation. Such global translation services represented an estimated $30 billion market in 2011, according to Common Sense Advisory.

Cloudwords says its cloud-based OneTM service ingests a company’s previously translated material, centralizes and de-dupes redundant information to suck much of the expense out of the process. OneTM, which Cloudwords CEO Michael Meinhardt calls a unified translation management system with a MongoDB NoSQL foundation, aggregates the data across departments and versions and also handles workflow to and from translators.

Traditional translation services typically require a ton of duplicative work. When a software application and its documentation is upgraded, for example, the entire project, not just the revised parts, usually gets retranslated. Since most translation services charge by the word, that gets pricey. OneTM separates out that which must be updated.

The technology has good SaaS bloodlines: It was co-developed by former crm) CTO Craig Weissman, working as a contractor with Cloudwords co-founder (and former architect) Scott Yancey.

“The best jobs to attack are those that are done inefficiently,” Weissman said in a recent interview. “Translating software, manuals and on-screen text is a big, big business that is very inefficient.”

To use the Cloudwords service, the software company uploads its existing documents (typically in XML format) into Cloudwords along with instructions as to what languages are needed.  In the old world, the software company sent out the work — via FTP or e-mail — requesting say, a German, French and Italian translation of a sentence. The company then got back the three translations, each with the original English sentence. There was a ton of duplicated information there that then needed to be stripped out. OneTM automates that.

The actual translation is still handled by people — machine translation is not yet up to this task — Meinhardt said.  But Cloudwords handles all the back and forth and data normalization and the data repository.

One early customer, Coupa Software, ended up paying $6,000, as opposed to the $40,000 quoted by its original contractor, to translate a software upgrade. “The $40,000 was the same amount they paid for the initial translation. It’s bad enough to pay that the first time, but in version 2 maybe 10 percent of the words have changed. There is no need to re-translate the balance,” Meinhardt said.

Cloudwords competes with MyGengo, Smartling and Lingotek as well as big translation companies such as Lionbridge Technologies and SDL. Meinhardt and Yancey see the service extending beyond the software business to other verticals, including the pharmaceutical industry.

Feature photo courtesy of Flickr user ToastyKen.

10 Responses to “Startup Cloudwords attacks $30B translation market”

  1. Eleanor Johnson

    This type of technology is not new. There are a number of startups working in the area offering this feature. I’ve used Transifex and WebTranslateIt, and both offer shared shared TM on the cloud.

    Furthermore, a large number of plain-old translation agencies and tools such as WorldServer offer this in similar/hosted ways.

  2. There is no doubt that translation memory is not a new technology or concept. There are indeed plenty of tools, several of which have been mentioned in this forum, to help translators and translation vendors (a.k.a LSPs) manage TM.

    But what Cloudwords’ OneTM delivers is the first pure multi-tenant cloud-based translation memory system that enables a translation customer to directly own, manage, and centralize their TM– across any and all LSPs they work with. Thanks to the multi-tenant cloud model, Cloudwords is delivering sophisticated TM management at a pricepoint and ease of ownership that has never existed before.

    OneTM delivers an innovative and first-of-its kind way of classifying, dimensioning, and categorizing Translation Memory. OneTM is built to work at massive scale for any sized customer. OneTM automatically, without any explicit work done by the customer, can create the most relevant buckets of TM to apply to any translation project… factoring in considerations such as type of content, company department, age, etc. And as we all know, TM is only as good as how accurate and relevant it is, and OneTM delivers to translation customers relevancy and ease of management never before seen for data sizes never before supported.

    Cloudwords is not an LSP and does not provide translation services directly. And we fully agree that high quality translation services is dependent on having great translation project managers and translators to execute the work in conjunction with everyone utilizing great technology to manage the process. Cloudwords is focusing on delivering the best technology to translation customers to manage their translation process and assets, thus leaving the customer free to choose the best LSPs.

    -Scott Yancey

  3. Irrespective of which TM solution or TM server platform you’re using – if you’re using one – you’ll be able to do all of the above. This definitely sounds run of the mill and simply leaves a lot of questions to be asked.

    Any knowledgeable industry vet could poke holes through the single (unified) TM model, which is why I was surpised by how they make it seem innovative.

    I sincerely hope (for them) that they do have something innovative that they simply failed to communicate through the PR.

    Good luck Cloudwords.

  4. This article describes an over simplistic view of the software localisation process. The software localisation process in addition to the translation of the text strings may include the rebuilding and testing of localised software; the handling of localisation tool kits that will extract the translatable text from the relevant software files and prior to the main translation cycle, a pseudo-translation cycle to test the process.

    Software localisation often carries a premium rate because the text strings that the translator will work with will appear “out of context” – they are in the main a series of commands as opposed to paragraphs of text; other factors to take into account include text string lengths – the space that the translated text has to fit may be limited to 10 or 16 characters.

    LSP’s specialising in software localisation will provide their clients with glossary, terminology & translation memory services as part of the service; the use of these standard disciplines will save up to 40% per annum. They will be experienced in working with terminology, translation memory and other linguistic tools that have been the industry standard for the past 10 years or more. They will also use translators who are specialists in software localisation.

    I doubt that the likes of Lionbridge, SDL, TheBigWord, Applied Language Solutions and many more of the top 100 LSPs in the world will lose sleep or revenue to Cloudwords.

  5. Peter Argondizzo

    I agree with all of the folks that commented. This really isn’t new technology. We use the Across Language Server which has the same one to many relationship in terms of accumulating the Translation memory (one source language entry and then a matching entry for each additional language). I would imagine SDL’s technology and MemoQ’s technology is similar. In terms of the customer success story at the end of the piece, the language service provider who quoted 40K must not have been working with ANY translation memory otherwise anyone would have provided discounts for TM matches.

  6. Claire Harris

    Given their credentials as ex leaders the above article is disappointing. What is being brought to the industry that doesn’t already exist? And more importantly – what TM tools will they be providing to translators to ensure segment matches are the same?

  7. Matt Train

    Good luck to Cloudwords. We’ve been operating in this space with a very similar business model since 2006 and although it is competitive there is a huge amount of work out there for localiz/sation companies. The devil is in the detail however, and we quickly moved to having a team of highly qualified production managers to oversee large projects. We’re picking up clients in ecommerce, market research, and others rapidly. You don’t just need translation memory and worklow, which is what is described here. You need conscientious people managing it all and doing the translations. At the end of the day, it’s all about the people, regardless of the tools.

    • @adam. can you share what TM programs you are using and whether they are a Software-as-a-service? what was presented as new here was the unified memory stuff — that gets rid of duplicated material so it doesn’t need to be re-translated. would love to hear about other services that automate not the translation but the data aggregation/normalization/dedupe etc. thanks