翻译：颜玉祚（北京大学计算机辅助翻译专业）

唐纳德·A·德帕尔玛 Common Sense Advisory 股份公司

在那些看似毫不起眼的细微区别背后，是大量的人力资产、大规模的流程，以及更多地， LSPs 用来管理内部工作流、与客户和供应商交互的技术以及作生产用的翻译工具。有些 LPSs 开发自己的“翻译管理系统”，其它的则购买现成的商用软件。

上个月，我们在 GALA （全球化与本地化协会）的成员公司中做了一次调查，主题是关于这些 LSPs 用以经营业务的技术。 我们调查的内容包括他们的开支计划、使用何种软件、为什么他们使用（或不使用）现成的商业软件，以及文件交换标准在他们计划中的重要性。 共有36家 LSPs 完成了该调查。

2005年末，我们也做过类似的调查，当时的样本有266家语言服务提供商。 2007年的调查对象与2005年并不完全相同，但有部分的重叠。 为给本年的数据提供一些背景，我们将会援引2005年的一些结果作为比较的对象。 若要了解2005年调查的完整论述，请阅读《语言服务2006：供应方展望》（2006年1月）。 该报告回顾了市场状态，对未来行业的、地区的以及商业的发展活动进行了预测，有些预测已经在过去两年中发生了，有的还未发生。 请保持关注。

2007调查：把钱拿出来！

Common Sense Advisory 的研究方法论是从资金开始，之后才涉及到内容、服务和技术。 因此，我们询问 LSPs 的第一个问题是“下一年贵公司的开支计划如何？”。 样本中大约三分之二（67.6%）的被试说他们计划在2008年投入更多，而大约有五分之一（21.6%）的被试说他们的开支预算会与上年持平。 对Idiom，LTC，Project Open，SDL 等全球化软件供应商来说，这无疑是个好消息。

相比2005年的调查，当时我们的被试中只有48%计划在下一年投入更多，有43%与上年持平。 尽管样本不同，我们仍然认为2007年的结果验证了过去一年中我们听到的轶事、做过的访谈以及提供的咨询数据。这些数据显示：LSPs对翻译技术和流程管理技术兴趣上升。

这样上升的关注度应该归因何处呢？ 我们可以想到一些原因：

• 规模： 全球化工作的总量在增加——更多字数、更多语种、更多产品以及更多想要走向全球化的公司。 所有这些都转变为对自动化和高效率的需求。
• 竞争： 因特网使得刚成立不久的小公司和自由译者可以在国际范围内参与竞争，其中不乏通过降价策略来拓展新业务的小公司和自由译者。 而自动化只是制度更完善的公司有效对抗“出低价者”的一种方式。
• 选择： 较之2005年，全球化软件产业中选择更多了，生存下来的软件供应商更加强健，语言服务销售量上升了，曾在2005年风头出尽的 Trados 现在也有了更多清晰而强劲的竞争对手。 还记得2005年吗？ SDL 收购了 Trados，Idiom 这样的公司根本几乎站不住脚，翻译记忆或翻译管理公司只有两种选择：要么是硬挺，要么是出售。 两年的时间，供应商这边已经是另一番景象。

LSPs 更愿购买而非开发翻译自动软件

考虑到2005年年中以来市场中有大量解决方案涌入，我们很好奇从那以后用户的购买方式是否发生了变化。 我们问道：“一般说来，您会购买还是构建（即，内部开发或付钱给为你开发的人）语言技术呢？ 在所有适用选项前打钩。”

• 购买者人数超过构建者： 今年，样本中44.7%的 LSPs 说他们愿意购买现成的商用产品，两年前这一数据是46.7%。 愿意购买但是认为自己仍需构建一些组件的被试占到42.1%，他们认为“大部分时候我们购买，但是有些软件还需要自己构建”，而2005年这一数据为30.5%。
• DIY 加快步伐 今年样本中愿意构建而非购买的 LSPs 较之2005年有增长（18.4% vs. 10.5%），但是仍然只占样本中的小部分。 持“构建+购买”战略的 LSPs —— “大部分时候我们购买，但是有些软件还需要自己构建”——从2005年的12%小幅上涨到今年的13.2%。
• 开源软件成为一个影响因素： 两年前我们没有询问有关开源软件的问题，但是这次我们做了，为的是响应早期调查中的一些填写的回答。 样本中六分之一（15.8%）的 LSPs 正在使用开源软件，这说明随着更多的开发商和学术机构发现了这一产业，有越来越多的解决方案在不断涌现。

我们还询问那些选择构建翻译系统的 LSPs 现成商用系统有什么不好： “如果你选择构建或大部分构建，为什么不购买一个商用产品呢？” 在所有适用的原因前打钩。” 排行前三位的原因是：“LSPs 认为商用产品不能满足他们所需的全部功能”，“他们的过程很独特”以及“整合是一个问题”（见图1）。 这三个不刺激购买的动机与2005年的前三位原因完全一致。

有一个答案颇为让我们注目——选择“担心从竞争对手处购买”的被试从2005年的14.3%显著增长到今年的27.8%。 这个问题源自流传在部分翻译机构中的一种感觉，它们认为 SDL 在收购了股市表现为中性的 Trados 后，会更偏好自己的服务业务而非那些使用 SDL 或 Trados 软件的 LSPs 竞争对手提供的服务。 在过去一年中，我们从 Crossgap 和宝阁文（thebigword）这样的 LSPs 为了在开发市场上出售而将内部系统产品化。 一般来说，我们建议 LSPs 避免这种商业模式，但是我们也确实意识到为什么公司会选择这样做——我们将会在后续的文章中谈到这个问题。

图1： 为什么 LSPs 不使用商用软件

数据来源：Common Sense Advisory 股份公司

LSPs 畅谈为什么不愿意购买商用软件

有些 LSPs 和我们分享了他们的一些关于商用软件为什么在他们公司过不了关的想法。 他们的解释基本可归为三类，都与构建综合性更强的解决方案以适应翻译管理有关：

1. 保有期： 有些 LSPs 多年前已经开始自己开发商用系统，主要是由于当时市场上没有好的解决方案。 现在，他们已经在构建系统上投入良多。
   
   “1999年我们开始开发的时候，市场上根本找不到能够满足我们和主要客户规格的商用产品。” Bengt Sjögren，Interverbum 公司
2. 客户需求和成本： 他们的客户有使用多种技术的习惯，这非常不方便，对软件的灵活性要求很高，而这恰恰是很多翻译管理系统不能提供的。
   
   “我们的主要客户在本地化工作流中使用不同的环境和工具。 这要求我们不断调整以满足他们的要求。 我们发现几乎没有哪个单独的商用软件足够灵活或可以开放地让用户定义（至少普通用户）和修改功能。 另外，我们还必须应付昂贵的许可证以及后续的升级费用。” Krzysztof Leporowski，Studio Gambit 公司
3. 内部的复杂性： LSPs 的内部本身就很复杂。 很多公司的主要问题是没有哪一个软件可以解决所有的问题。
   
   “没有‘单个软件’可以解决所有的客户问题。 他们的需求各不相同。 我们发现 Idiom 公司的产品为本地化工作流和内容管理系统之间起到了很好的连接作用。 对于不那么高级的用户（或较小的团体）， Author-IT 公司提供了一套很有趣的内容管理和本地化工作流方案。 对于那些需要集成内容管理和本地化工作流的用户，DocZone 公司提供了质优价廉的解决方案。 但是，总的说来，我们一直在为客户寻找‘恰当的解决方案’。” John Watkins，ENLASO 公司

有些 LSPs 不认为极可能成功的解决方案会让他们更容易做决定：

“当市场上没有清晰的产品可以最好地满足你的需求时，信任一种产品是让人担忧的。” Teddy Bengtsson，IFL 公司

标准——从来很重要，但又总是缺位

最后，每次 GALA 会议上都会讨论标准问题。 我们的专题小组也不例外。 86.4%的被试认为 TBX，TMX，和 XLIFF 这样的标准“重要”或“非常重要”。 只有13.5% 的被试认为它们“不重要”，没有人认为他们“毫无用处”。 2005年的调查中，67.3%的被试认为这些标准“重要”或“非常重要”，28.6%认为它们“不重要”，4.1%的被试认为它们“毫无用处”。 同样，虽然样本不同，但是时代特征却很明显。

有一些反对的声音，其实是在附和我们对文件交换标准在实现协作能力上不足的分析，但是这种能力正是市场要求的。

 “有些标准不够好，他们本该对我们很重要，但现在却不是那样。 例如，TMX 并不真的在不同的翻译记忆格式之间提供兼容性。” Bertrand Gillert，LocaSoft 公司

软件开发商的行军令

这些回应应指导那些指望增加 LSPs销售额的供应商做两件事情： 1) 和潜在客户交谈，找出他们的需求以及现有软件的缺点；2）在营销和技术支持上扩大投入。 在本电子通讯的《TMS 开发商如何定位产品到 LSPs》一文中，我们讨论了第二个问题。 即便是世界一流的软件，如果缺乏集中的营销、销售和支持，也无法扩大市场或劝说 LSPs 停止构建自己的解决方案。

What Do LSPs Look for in Software They Buy?

Donald A. DePalma, Common Sense Advisory, Inc.

Underlying those anodyne differentiators are hordes of human assets, extensive processes, and, increasingly, the technology they use to manage their internal workflows, interchanges with customers, their vendors, and the translation tools that do the work. Some build their own "translation management systems" while others buy commercial off-the-shelf (COTS) software.

Last month we surveyed GALA members who are language service providers about the technology they use to run their businesses. We asked about their spending plans, which software they use, why they do or do not use COTS software, and the importance of standards in their plans. Thirty-six LSPs completed the survey.

We have asked these same questions of LSPs before, most recently with a panel of 266 language service providers in late 2005. We did not survey exactly the same people in 2007 as we did in 2005, but there was some overlap in the two groups. To provide some context for this year's data, we will reference some of the 2005 results for comparison. For a complete discussion of the 2005 survey, read "Language Services 2006: Supply Side Outlook" (January 2006). That report reviewed the state of the market and predicted many of the industry, geographic, and business development activities that played out over the last two years and some that have yet to happen. Stay tuned.

The 2007 Survey: Show Me the Money!

Our research methodology at Common Sense Advisory begins with the money and only then proceeds to content, services, and technology. Thus, our first question queries LSPs about their spending plans for the coming year. Roughly two-thirds of the sample said they planned to spend more in 2008 (67.6%) versus about a fifth who said their spending would remain about the same (21.6%). That's great news for globalization software vendors such as Idiom, LTC, Project Open, SDL, and others.

By comparison, 48 percent of our 2005 respondents told us they planned to spend more money in 2006; 43 percent were planning to spend the same amount as in 2005. Although the samples are different, we would say that our 2007 results validate our anecdotal, interview, and consulting data over the last year that shows increased interest among LSPs in translation and process technology.

To what can we attribute this heightened attention? We can think of a few reasons:

• Scale. There's the increasing volume of globalization work — more words, more languages, more products, and more companies wanting to go global. All of that translates into the need to automate and work more efficiently.
  
• Competition. The internet has allowed tiny new suppliers and freelancers to compete internationally, many underpricing their way into new business. Automation is just one way that more established companies can compete more effectively against lowballers.
  
• Choice. Compared to 2005, there is more choice in some globalization software sectors, the software vendors who survived are healthier, language services sales have grown, and there is some clear, strong competition for the 2005 incumbent Trados. Remember 2005? That's when SDL bought Trados, firms like Idiom were barely holding their own, and the only other choices in translation memory or translation management were suffering or being sold off. Two years have made a lot of difference in the vendor landscape.

LSPs Prefer Buying Translation Automation Software over Building It

Given the flood of solutions that have entered the market since mid-2005, we wondered how the buying patterns have changed since then. We asked, "Generally speaking, do you buy or build (that is, internally develop or pay to have someone develop for you) language technology? Check all that apply."

• Buyers outnumber builders. This year, 44.7 percent of our sample said they buy commercial off-the-shelf products versus 46.7 percent two years ago. An almost equal number prefer buying but find that they need to build some components — 42.1 percent said that "we mostly buy but build some software" versus 30.5% in 2005.
  
• Do-it-yourselfers have picked up the pace. This year's smaller sample showed a greater tendency to build than buy (18.4% vs. 10.5%). LSPs with mixed build-buy strategies — "we mostly build but buy some software" — increased slightly from 12 percent in 2005 to 13.2 this year.
  
  
• Open source has become a factor. We didn't ask about open source solutions two years ago but did this time in response to some write-in answers to the earlier survey. One-sixth of the sample (15.8%) is working with open source, which is testament to solutions that are popping up as more developers and academics discover the sector.

We also asked those who build what's wrong with COTS: "If you build or mostly build, why don't you use commercial products? Check all the reasons that apply." The top three reasons were that LSPs determined that commercial products don't have all the features they need, their processes are unique, and integration is a problem (see Figure 1). This trio of disincentives paralleled the top three reasons from 2005.

One answer did catch our eye — "concerns about buying from a competitor" increased substantially since 2005 (27.8% versus 14.3%). This issue derives from the feeling among some translation agencies that SDL, following its acquisition of neutral Trados, would favor its services business over that of rival LSPs using SDL or Trados software. Over the last year, we have heard from LSPs like Crossgap and thebigword who have productized their internal systems in order to sell them on the open market. We generally recommend that they avoid this business model, but we do appreciate why a company might want to do that — and we'll address this in a future article.

Figure 1 : Why LSPs Do Not Use Commercial Software
Source: Common Sense Advisory, Inc.

LSPs Speak Their Mind on Why They Don't Buy

Some providers shared their thoughts on why commercial translation tools didn't make the cut at their companies. Their comments fell into three buckets, all tied to the more comprehensive solutions built for translation management:

1. Tenure. Some LSPs have been developing a commercial system for years, largely due to the absence of a good solution when they started. Now, they have a lot invested in the solution they built.
   
   "When we started the development in 1999 there were no commercial products available with the specification we and our lead customers were looking for." Bengt Sjögren from Interverbum

2. Customer needs and costs. Their clients have the inconvenient habit of using a variety of technologies, thus requiring more flexibility than some translation management systems can deliver.
   
   "Our main customers use different environments and sets of tools for localization workflow. This requires us to continuously adjust to their requirements. We've found that no commercial solution is flexible enough or opened for user-defined (at least common user) functional modifications. Besides that, we have to deal with the high cost of licenses and subsequent upgrades." Krzysztof Leporowski from Studio Gambit

3. Internal complexity. Internal LSP needs are complex themselves. The bottom line for many is that no single solution does it all.
   
   "There is no 'single solution' out there for the customers. They vary in their needs. We find that Idiom provides a good link for the localization workflow to content management systems. For less advanced users (or smaller groups), Author-IT has an interesting approach for content management and localization work flow. For those needing integrated content management and localization workflow, there is a good & affordable solution with DocZone. In general, though, we are always working to find the "right solution" for our customers ." John Watkins from ENLASO

Some LSPs don't see that slam-dunk solution that would make their decision easy:

"When there is no clear choice in terms of which tool is best suited for your needs, committing to one product is a concern." Teddy Bengtsson from IFL

Standards — Always Important, But Not Where They Should Be

Finally, at every GALA meeting there is discussion of standards. Our panel was not different. Fully 86.4 percent said that standards like TBX, TMX, and XLIFF were important or very important. Just 13.5% said they weren't important, and no one said they were useless. In 2005, 67.3% said they were important or very important, 28.6% said they weren't, and 4.1% said they were useless. Again, different samples, but the zeitgeist seems clear. There were some naysayers, echoing our own analysis of the shortcomings of standards in delivering the interoperability that the market should demand.

"As some standards are not good enough, they are not as important to us as they should be. For example, TMX doesn't really provide compatibility between the different translation memory formats." Bertrand Gillert from LocaSoft

Marching Orders for Software Suppliers

These responses should guide software vendors looking to increase their sales to language service providers do two things: 1) talk to their prospects to figure out what they need and figure out how current software falls short; and 2) spend some money marketing and supporting their products in this market sector. We discuss this second issue in "How TMS Developers Pitch Their Wares to LSPs", also in this newsletter. World-class software without focused marketing, sales, and support will not grow the market or convince LSPs to stop building their own solutions.

Don DePalma is the founder and chief research officer of the research and consulting firm Common Sense Advisory, and author of the premier book on business globalization "Business Without Borders: A Strategic Guide to Global Marketing."