Monday, June 20, 2016

The trouble with Memsource

A major customer of ours recently switched from SDL Trados Studio to Memsource.

I don't know why our customer (a translation agency) chose Memsource. Probably a mix of several reasons, including that no software needs to be installed, since Memsource works in the cloud. Also, they can peddle it to translators as a "free" tool.

Memsource may be suitable for simple projects, but for more complicated ones and for advanced users it suffers from several serious drawbacks.

The most serious is that in the free version of Memsource you can't use your own memories and termbases—the program lets you work only on projects prepared by translation agencies, and only using the resources selected by them. Possible solutions to this are:
  • Upgrade to a paid version of Memsource, in which case you can add your own memories and termbases to a project. 
  • Load your translation memories and termbases in Xbench (or a similar tool), and use Xbench to search them. If you do this, however, you can use your resources only for reference—Xbench doesn't offer any automatic way to add new segments to a memory. Also, this workaround needs extra steps, so it slows you down.
  • Use a different program (such as memoQ) to translate your Memsource projects. Yet, if you do so, while you can use your own memories, you lose access to Memsource's ones. 
Another problem is that sometimes Memsource is painfully slow, even on a fast Internet connection. In certain segments of a recent project, 9-10 seconds passed between the moment I hit a key and the time the corresponding character appeared—this meant typing blind. When I complained about this to Memsource support, they told me the segments in question contained joins, tags, and were long, and this slowed online processing. I believe this speaks volumes about the limitations of the tool, although, to be fair, a Memsource representative told me they know of this bug and are working to correct it.

Memsource looks and acts like a stripped-down version of more powerful tools. This might be good for those who feel overwhelmed by too many choices, but experienced translators miss the advanced features they expect from professional translation tools. The first flaws that come to mind are:
  • Far fewer find and replace options than memoQ or Studio; 
  • Find and replace in memoQ, Studio and Memsource
    Find and replace in memoQ, Studio and Memsource
  • No such thing as memoQ's LiveDocs and Muses, or the wealth of added features SDL offers through the Open Exchange;
  • Limited segment filtering when compared with memoQ or Studio—for example, no regular expressions in the filters;
  • No auto-complete in the desktop editor;
  • Fixed screen layout. You cannot increase the size of the lower panes (CAT, Search, Changes). The little you can change, such as moving the panes from the right of the screen to the left, you can't save: next time you reopen the program, the panes are back where they started.
  • No way to show tracked changes inline in the Memsource editor—you can only see the differences in the Change pane, and that is not enough.
The lack of advanced features, the occasional slowness, and the fact you either cannot use your own memories and termbases or have to rely an external tool to search them, means that Memsource makes you less productive. According to our estimate, confirmed by what other colleagues say, we suffer a 30% drop in productivity when we work in Memsource instead of memoQ or Studio.

The supposed advantages of using Memsource

If Memsource was all bad, nobody would use it. So, what advantages does Memsource offer?

According to a Memsource representative, the main advantages are that it's cloud based, that it allows simultaneous access by several people to the same translation memory, and that for translators it's free.

Let's look more closely at these claims.

For translation companies, using a cloud-base tool may offer an enticing benefit: no need to install anything locally, and no need to migrate data when changing from one version of a program to a newer one, thus avoiding the risks associated with such migrations—though I wonder if these risks, far from being cut out, are instead passed to Memsource and to their staff in charge of the cloud servers.

Also, a cloud based tool means that project managers can work on the program remotely, even without personal computers: they can create and manage projects using a tablet or even a smartphone. I doubt, though, whether trying to create and manage projects using smartphones would be wise.

For translator teams and for translation companies another claimed advantage is that Memsource allows simultaneous read and write to the same memory in real time. I normally work on projects where I am the sole translator or editor, so this isn't much of a benefit for me. Even for projects with multiple translators, though, updating the same memory in real time can lead to fast spreading of undetected errors. Better to wait until a job has been edited and proofread before sharing the memory with others. Project managers may set up memories in Memsource so that only translations that have been reviewed and corrected are shared with others, but in my experience many project managers either are unaware of this feature, or choose not to use it.

Memsource is "free" for translators.
  • This is irrelevant for translators who have already paid for tools they have chosen themselves, and, if you wish to use you own translation memories and termbases in Memsource, you have to pay. A way to sweeten the deal for more experienced translators would be for Memsource to allow using a translator's own memories and termbases even in the free version.
  • A "free" tool is a real advantage only for beginners or occasional translators, who might be reluctant to pay at the start of their career for a more powerful program. Still, even if you are a beginner or occasional translator, relying on a tool that prevents you from building up and using your own translation memories is, in the long run, self-defeating. In fact, I think that beginners would be better served by other free tools—from OmegaT to Wordfast Anywhere—since they allow you to keep control of your own resources.
Overall, working in Memsource feels as I imagine a carpenter would feel if, after assembling during his career a rich set of specialized and high-end tools, his customers demanded that he use instead the simplest tools available at a dollar store and suitable only for the occasional user.

The developers at Memsource know that they still have much work to do: they started development only a few years ago, so more mature tools such as Déjà Vu, Studio or memoQ have a huge head start on them. But until Memsource catches up with its competitors, there isn't much in the program to appeal to professional translators.


  1. We share many of the same concerns, Riccardo. Tools like these are popular among agencies unable to manage their own technical infrastructures and who do not particularly care about the efficiency and resource access of their digital sharecroppers. I hear a number of technopundits declare this sort of thing to be the "future", but most of these people are really quite divorced from the actual wordface and do not give a damn about things that might slow down and restrict the income and output quality of their pieceworking crowd.

  2. Nice article. While MemoQ is used as an example here, I think these problems you mention relate to all cloud-based tools. For example:

    Basically, the cloud technology is not ready yet for translators, high performance is a must for productivity.
    However, this gap will narrow in the future as technology advances, but I estimate it will take another 10 or more years.

    1. Sorry I said "MemoQ" when I meant MemSource...

    2. I'm less sanguine: yes, technology will improve, but the bulk of the improvements will be for things that enterprise customers (and, to a lesser extent, LSPs care about). Whatever usability improvements we'll see that specifically benefit translators will be byproducts of improvements required by enterprise clients and by LSP.
      For many of them the fact that translators cannot feed their own translation memories will be seen as a feature, not as a bug.

  3. Inspired by your article, Riccardo, I decided to take another look at memSource for a little project. I last played with it about 4 years ago and gave up in disgust at the time after I managed to create a few hundred "ghost" entries in a termbase and a few other things that I have mercifully forgotten about. Another thing I had forgotten about was that the desktop client is not properly integrated with the online system, so I had to download their goofy XLIFF variant manually and then open it in the editor. Working in that feels like getting a lobotomy. If I were in a mood to play I might put this client through Jim Wardell's test criteria for compatibility with dictation software, but not today. I need to get some work done, so to heck with it - off to memoQ with that "mxliff". And in the unlikely event I need any of the server resources, I can just leave the memSource editor open in the background while I work with my own, better resources for this subject area. This tool really is NOT ready for professional translation prime time!

  4. Kevin, I agree with you. This article is very inspired. I have a lot of related things with translations in my life so this article helped me to solve some problems.

  5. Strangely enough, I started playing around with Memsource a few days ago, on real projects, and must say that I am very pleasantly surprised. I haven't used memoQ for a years now (as it is just too damned slow), but I think Memsource is way better, both in terms of UI responsiveness, and of just getting shit done. I know I am supposed to hate it, and all online CAT tools (which, in all honesty, I actually did until about a week ago), but I am actually really enjoying it and might even pay for it once my 30-day trial runs out. My main tool is CafeTran these days, but there are a few things I prefer about Memsource at the moment.

    1. Hi Michael,
      What kind of things do you prefer about Memsource?

    2. MemoQ slow? Not in my experience. But as a Memsource newbie who uses memoQ and Studio regularly, I find Memsource easy to learn but very limited, and the Web Edition is IMPOSSIBLY slow. I say that after using Across (which I am going to dump, it's too exasperating to work with), which is always linked to the client's server, or MemoQ, which I have used 90% of the time on the client's server. response is always quick, even with my medium-speed Internet connection.

  6. Hi Riccardo,

    I'll try to update this when I have more time (which is never), but just off the top of my head (WARNING THE FOLLOWING WILL BE DICTATED):

    • The preview system is great in Memsource and just works. It also allows you to click anywhere in the preview window and it will take you to the relevant segment in the grid . memoQ's is also OK, but Studio's is a mess, and CafeTran doesn't have a built in one yet.

    • I can create MASSIVE TMs (and termbases) in Memsource, which don't seem to slow down working in the grid! Furthermore, I can also run superfast concordance searches on these massive TMs. Speaking of TMs...

    • I LOVE the fact that you can search both glossaries and TMs simultaneously with a single shortcut.

    • A very important point is UI responsiveness. As I mentioned already, over the years, memoQ you has just gotten to darned slow. I also noticed that they have revamped the termbase entry display system recently, so that it is no longer possible to quickly scroll through the termbase entries in the Translation results window using Ctrl + UP/DOWN arrow. :O This is a massive step backwards in my opinion. Clicking on pretty much anything in memoQ these days takes at least five seconds before anything at all happens. This is really annoying when doing actual work, and unacceptable. And what's crazy is that any of these things that take very long in memoQ are virtually instantaneous in Memsource. The Memsource web editor is slightly slower than the desktop editor, but both are MUCH faster than anything offered in memoQ. Thankfully, CafeTran doesn't suffer from this problem at all, as everything you do in it is lightning fast.

    • I also like the minimalist design and layout of Memsource better (which, as you probably already will have noticed, it's basically a copy of the memo to layout), better than memoQ, and better than CafeTran, whose UI is a bit messy and confusing still. It doesn't get in the way and allows me to work faster.

    OK, that's all for now. Will post more if and when I can!

  7. C'mon guys, if for some strange reasons, such as using Romanian Programmers keyboard layout, someone needs to use Alt Grey + Letter to generate his language special characters, such as ă, ș, î, ț or â .... he/she is out with MemSource, just because it is not possible to configure keyboard shortcuts, such as in Studio or MemoQ.

    Any way, someone may say ... there is an alternate keyboard available. Really? With all 5 high frequently used chars on the little finger of the right hand ... and placed in a totally different position ... it is not an option for a pro translator focusing on translation not on finding chars on the keyboard.

    For my target language is a total mess :(

  8. IYAM, the best way to easily add special characters is to use: (or just create special shortcuts for them in AutoHotkey).

    However, the Memsource desktop editor (which is what I use) DOES allow you to customise shortcuts (unlike the Memsource web editor), and is available on Windows/Linux/Mac.

    I recently also wrote a pretty long review of Memsource @ and @


  9. Well from the project manager side this tool is terrible. It lacks so basic features like correction of the TMs. There is no built in feature to search for all affected terms and correct them. There is process consisting of export of the whole TM (which is so big that it cannot be edited by anything) out of MS, correction of the affected strings and then reimporting back to the MS. Import is painfully slow just to find out it does not update strings properly. There is no log to show what had happened during import.


Thank you for your comment!

Unfortunately, comment spam has grown to the point that all comments need to be moderated. All legitimate comments will be published as soon as possible.