Web Portal localization testing estimation as % of estimated functional testing efforts?

Question

I'm currently estimating QA efforts for huge enterprise web portal (about 5k+ hrs DEV efforts / 1.5k hrs QA), based on Liferay / Alfresco platforms / solutions. One of the requirements - 3 interface languages user may switch from any page (place) of the solution.

So the question is that simple: is there any common practice to estimate localization testing as percentage of overall functional testing efforts?

The following assumptions / aspects are actual as well:

• 2 of 3 languages are common and do not require any special skills for checking from QA team; the 3rd language is uncommon (however, not TOTALLY different, as e.g. English and Japanese). Any criteria for estimations relating that 3rd language, perhaps x2 / x3 efforts required for localization testing of common language?
• Which approach is better (or easier to explain) - %% based on overall functional testing efforts OR based on the approximate number of separate web portal pages?

At the moment the following makes sense for me:

1. 5% of overall testing efforts (i.e. about 75 hrs) for common language localization testing;
2. Approx. 10-15%, i.e. about 200 hrs for UNcommon language localization testing.

For the record - web portal will have about 120-140 different types of pages that should have localized elements.

Any suggestions / help / ideas / approach (as best - based on real experience / cases) are highly appreciated!

P.S. no sample link available - it's requirements refining stage.

Answer 1

As well as "what Joe said, in bulldozer-sized loads", I have a few pointers to add from experience with localization/translation.

To test internationalization - which will need to happen first - you'll want something that's really obvious so that in a test environment your team can go through every interaction in the system to check that everything that should be translated is translated. This alone in a sufficiently large system is extremely time-consuming because it won't be possible to simply go to each page - you'll need to perform the actions that will trigger each error message/information message/popup/etc. That should be happening before the actual translation is tested. The testing I did used a bogus language (consisting of random numbers) so it was really obvious if anything wasn't translated. You open the page and see words anywhere that's not pulled from the database, the internationalization hasn't been done.

To test translation - this will depend on whether or not you have someone on hand who actually speaks the language. If you have a native speaker or bilingual tester, it will take about the same amount of time as internationalization, maybe a little more if the language takes a lot more space.

If you don't have a native speaker or bilingual tester on hand, figure extra time to check that the translation is actually more or less correct. The more technical the text they need to check, the more time they'll need.

Some other caveats I've run into: you'll want to allow extra time on the development end for GUI modifications: English is one of the more concise languages around. Almost every other language requires more space to display a caption than the English equivalent.

If the language is like Japanese, it requires much more memory per character. I ran into an issue where the Japanese localization of an application was overflowing buffers because the memory requirements of some 200 lines of kanji was about the same as the memory requirements of some 5000 lines of English (that particular application represented the kanji as images rather than via code pages).

Answer 2

As always, it depends.

If the system is already sufficiently internationalized, and other localizations have already been performed, then future localizations may be well-defined, documented efforts and their scope may be easily estimated based on past efforts.

On the other hand, if the system isn't internationalized at all, then you may need to incur the overhead of that effort within your first localization task. Hence, it could take much, much longer.

Additionally, the required accuracy of the translation effort and the domain over which it operates needs to be factored in. For example, it might be simple to test the translation a general-interest page from English to French (particularly when accuracy isn't vital), but it will be quite a bigger effort if the page is medical literature where the accuracy of each word in context might be of utmost importance.

I know of no magic formula which could express the "localization testing" as a percent of the "overall testing". I can't imagine that there is such a general formula that could be useful across different applications/projects. What if the rest of the project turns out to be simple to test (perhaps because you have tested the English version hundreds of times before) - then your localization percentage must be higher. But what if you are using brand new technology as part of your localization effort - then the localization effort might be an insignificant subset of the overall testing project.

You might be best served to analyze what is in front of you, take a guess at the first localization, use it as a pilot project, then estimate the remaining localization based on what you learn to that point.

If you are forced to present a guesstimate using a percentage, try 10% for the "common" language, and 20% for the "uncommon" language, and hope for the best.

Good luck!