A New Model for Testing Global Requirements
Alym Rayani, Director of Client Solutions, The Symbio Group, USA
Intended Audience: Software Engineers, Managers, Testers
Session Level: Intermediate
In this session, we will provide an in-depth description of a new, distributed model for localization testing that separates localization functionality testing from Linguitic Verification Testing (LVT), thereby significantly reducing overall localization testing costs.
The current prevailing model for performing L10N testing relies heavily on the use of software test engineers that are native speakers of the target language. This requires the company performing the testing to employ skilled test engineers that have strong language skills. And because the language requirements typically require native language skill levels, this has generally meant that the testing had to be done in higher cost regions such as in the US or Europe where there is a large pool of native language speaking resources available.
However, since pure linguistic bugs account for such a small percentage of the actual globalization problems found in software products, logic would dictate that test planners isolate LVT from the other techniques for L10N testing. Yet until now, the models typically proposed for separating the LVT from functionality testing, including having a language specialist and a tester work side-by-side or completing functionality testing prior to LVT, have failed to reliably produce quality, cost-effective results.
There is now a new approach that closes the loop between functionality and linguistic testing and ensures that all aspects of product testing are covered. The key to making this model work is a process, supported by technology, which enables non-technical language specialists to review all UI elements within the product. More importantly, the process is extensible so that language specialists can perform LVT for multiple localized builds as they become available without having to work with the build directly.
To ensure that the language specialists are able to uncover all of the UI elements within the product, we have developed a model built around sending screenshots to the language specialists for review. The LVT team does not have to work with the product directly and is responsible only for reviewing and editing screenshots. They are also given a UI Roadmap to help them navigate the screenshots in the order that a user would typically follow within the system.
While this new model may increase project risk, the risk can be mitigated with increased communication and the payback comes in the form of lower testing costs. Because the Localization Functionality Testing team and the Language-Independent Localization Testing team are no longer tied to any particular language, they can be located in lower-cost offshore testing centers. Also, by not requiring the language specialists performing the LVT to have specialized technical skills, the overall project cost is reduced.
Conclusions
This model does not replace the existing model for localization testing,
but it offers a very cost-effective alternative.
Removing language dependencies from the globalization test lifecycle is a natural evolution in the L10N industry. By taking out the limiting step of language from the globalization test lifecycle, L10N testing can be done anywhere, and we predict will increasingly move to an offshore testing model. China will likely be the key beneficiary of this trend since knowledge of double-byte issues will continue to be a big focus of L10N testing and there is still a cost advantage of testing in China over most other offshore locations.