In February 2012, XTRF Management Systems released its eponymous translation management system, XTRF 2.5. The system will be showcased at the TMS Inspiration Days conference in Krakow, April 19-20. The conference is co-sponsored by XTRF and its sister company, Lido-Lang Technical Translations. As might be expected from its roots in the language services industry, XTRF focuses on business and project management capabilities from the perspective of language service providers (LSPs).
That said, the company has brought in senior software engineering and software product management expertise that is readily evident in the maturation of the product itself. Having previously set the tone by pioneering deep integration with various translation environment tools, the company’s latest release takes another large step in the direction of openness and interoperability.  
I recently had the opportunity to discuss the product vision with technical director Dominik Radziszowski. I want to focus on the current product version, but I think it is important to see this version as movement along an arc with a definite goal in mind. Radziszowski recognizes the amazing diversity among LSPs, and rather than trying to be “all things to all people,” he has set a course that allows XTRF to focus on core capabilities while providing technical “hooks” on which tech-savvy LSPs can hang their individual customizations.
There is a well-documented web services application programming interface (API) that allows customers to implement their own preprocessing mini-applications and then create new projects in XTRF automatically. There is also a flexible workflow definition capability that allows customers to create their own automatic actions — essentially the ability to embed automated processing steps such as custom quality assurance checking into the overall workflow. These steps can be executed without manual intervention and without risk of manually introduced errors in file handling. Both of these capabilities are relatively immature in the current release, but it is obvious that they have been designed in a way that can be extended in the future based on feedback from early adopters. In fact, the automatic actions capability is essentially an extension of the previously released deep integration capabilities.
XTRF uses the term deep integration to refer to the approach to seamlessly interacting with external translation environment tools (TEnTs). With this approach, localization project managers can create projects and upload source files directly into XTRF. XTRF then interacts programmatically with the designated TEnT to leverage and analyze the files based on the appropriate translation memory (TM). The resultant TM logfile is then automatically available within XTRF to generate end customer pricing — computer-assisted translation (CAT) receivables in XTRF terms. Vendor assignment can be accomplished in the traditional manual way, but selection of an appropriate vendor is supported by a rich and customizable combination of language pair, domain specialty and quality evaluation characteristics. There is also a capability to do vendor assignment automatically or semi-automatically based on specifications within the automated workflow. In all of these cases, the file analyses provided by the linked TEnT are available along with vendor pricing profiles to create an accurate purchase order and CAT payable directly from XTRF. Currently supported TEnTs include the server-based MemoQ (from Kilgray Translation Technologies) and web-based XTM Cloud (from XTM International).
XTRF is based on open technologies both for its underlying data storage and its user interface. Whether it is installed on a customer’s server or hosted remotely by XTRF, the application itself runs entirely within a browser. I have tested it with both Firefox and Internet Explorer without issues and have seen only minor anomalies with Chrome. There seems to be some compatibility problems with Safari on a Mac, but switching to the free version of Firefox solves those. The presentation is clean and functional. There is also a consistent layout regardless of where you are in the system. The left side of the page presents various context-sensitive filtering and selection criteria, while the main panel contains the list of selected items. Columns can be reordered and either included or excluded to provide customized views of the data. These views can also be shared among users, or used to generate scheduled e-mail reports for management or others who do not work with the system on a regular basis. They can also be used to export data in a variety of standard formats for further analysis or transfer to a third-party system such as an accounting package. Of course, the concept of deep integration applies here as well since, for US releases, there is already a supported interface for QuickBooks.  
Clicking on any displayed item takes you directly to the appropriate detail page. Editing can be done from the detailed page by selecting the Edit icon on the main page or by selecting multiple items and using the Actions button. The system is also very good about reminding you if you try to leave a page without saving changes.
When it comes to further analysis, there are also some strong features within XTRF that are available. Many commonly used financial metrics such as total revenue per language pair or profitability over time are available as predefined expressions. If you want to be more creative, it is very easy to define new expressions that calculate total or relative values and allow you to display or graph them over time (see Figure 1).
 A feature that many business owners will like is the ability to define this kind of financial report and have it automatically generated and distributed by e-mail on a fixed schedule. Properly used, this can be an aid in focusing management attention on production challenges, sales opportunities and general business decision making. The same approach can also be used to define and monitor key performance indicators to support ISO certification or general adherence to process. Before we get any further in the review process, however, I want to address a major stumbling block that most companies have when it comes to implementing a system like XTRF. And let me hasten to add that this is not limited to translation management systems or even narrowly to the language services industry. None of these wonderful capabilities comes without a price, and it is no different with XTRF. In fact, the “price” manifests itself in at least two distinct ways. In the first place, there is a substantial effort required to define and configure vendors, customers, price lists, specialties, assignable activities, workflows, trackable metrics, target margins and so on. Even if you already use a different or homegrown system to maintain this information, it is highly unlikely that your data will be organized and linked properly in your current system. Note that by “properly” I am assuming the perspective of the new system. And if you think about it, the very features that we find so attractive in a new system under consideration are the ones that are not supported in our current one, so this requirement should not come as the surprise that it so often engenders. That said, XTRF does provide a limited data import capability that is helpful. It provides a defined format for a CSV (comma-separated values) file, which you can then generate from your existing system. Given the data model mismatch mentioned above, though, there will still be plenty of work to do after the basic data is imported. One potential improvement along these lines would be for XTRF to support a data import API. At least for customers with technical capabilities, that might allow for a more complete migration of legacy data into XTRF without threatening data integrity.
In the second place, it should be noted that any new system will have built-in assumptions about processes that probably do not match your legacy processes. Again, one of the attractions of the new system is precisely the fact that it streamlines and automates error-prone manual processes, but that also means that you will not be able to proceed with business as usual. Changing tools without redesigning processes is likely to be frustrating for everyone and unlikely to result in a substantial return on investment. The focus should be on “How do I accomplish the business goal with the new system?” and not on “How do I replicate the file structure (or any other attribute) of the old system?” Fortunately, XTRF provides substantial help for those who genuinely want to reengineer their processes. Chief among these is the flexible approach to workflow definition (Figure 2).  
Using the graphical interface, workflows containing an arbitrary number of activities can be created and manipulated. Source and reference files can be bundled for processing, and different bundles can be routed differently in the workflow. For instance, files that govern final layout can be routed to the desktop publishing (DTP) or proofreading step while only the translatable text is routed to translation. For large volume projects, files can be bundled for purposes of workload balancing, with each bundle assigned to a different provider for a particular activity. Providers can be manually assigned or pre-assigned as part of the workflow definition. Setting this up properly for each customer (or class of customers) and each type of project does take time, but doing so generates substantial time savings during project execution when it really matters. The configuration effort is essentially planning time, which allows things to run efficiently and quickly once the customer deadline is looming.
Once the project is underway, project managers can monitor progress, assign activities and manage the schedule using a customizable dashboard (Figure 3). Both the displayed columns and the selected rows can be configured according to user preference. There is a drill down capability that allows the user to expand a project into tasks (effort associated with a specific language pair) and tasks into activities (individual steps in the task workflow). The status of any particular project, task or activity is clearly indicated with color-coding that focuses attention where it is needed. At the activity level, unassigned activities include a button that activates a context-sensitive assignment dialog. Pre-populated selection criteria such as language pair, domain specialty and so on can be augmented by other selection criteria to identify and assign qualified providers.  
This is also a good place to mention the flexible e-mail communication capability provided by XTRF. E-mail templates for various types of communications (activity assignment, requests for interest) can be predefined and are automatically sent to prospective vendors based on flexible configuration rules. There is a vendor portal that also allows vendors to accept or reject job offers, download or upload files using secure file transfer protocol, and even update their own qualifications. Similar capabilities are also included for customer e-mail communications, and a customer portal allows customers to submit files directly into the system and request a quote, or even preauthorize the project.
There is also a flexible approach to creating both quotes and invoices directly from the system. This capability is based on predefined templates that allow various levels of detail or summation based on a strict project/task/activity level. This is one area where XTRF could make some substantial improvements, however. In the first place, the template definitions are arcane, and the controlling logic is not centralized. This means that it takes programmer-level expertise to construct a consistent set of expressions and display logic to address anything beyond very basic customization needs. In the second place, the output format is an uneditable PDF, and the PDF generation does not contain any presentation intelligence such as widow or orphan control, the ability to keep related items on the same page and so on. Furthermore, while it is theoretically possible to build in the capability to add project-specific commentary text, the effort required to do that in a consistent manner is significant. Most users would probably prefer an editable format for these key customer communications, even though that means accepting responsibility for errors that would result from changing the actual numbers calculated by XTRF. Another potential shortcoming for some organizations may be the over-reliance on language pair as an organizational framework. While it is true that traditional localization processes can easily fit into that framework, it is less clear that emerging challenges such as multimedia localization can be as easily accommodated. More complex workflows involving a sequence of tasks performed against the customer source files themselves (and dependent only on the source language) followed by a “split” into target-language-dependent subtasks would be welcome. In the meantime, however, organizations with significant needs in this area can probably address those through customized preprocessing applications focused on the source file activities that then utilize the XTRF API to create the language pair specific projects.
All in all, though, this is a system deserving of consideration for most small-to-medium sized language services companies. As previously mentioned, there is more diversity among LSPs than is generally recognized, so your mileage may vary. This is why it is always wise to rely on an individual “test drive” rather than a sales demo or even a review. The time and energy required to run a full pilot in your environment are cheap insurance against making the wrong choice.