Get your own customer support community
 

What are the software recommendations intended to do?

I continue to be confused by your "software recommendations"; never mind that the "relevance" hovers around 50% (what does that actually mean?) I have rarely if ever seen anything on the list that is *actually* relevant to me.

Also, there's just one "block" in the newsletter, but on the website I'm forced to choose between Windows, OS X or Linux - why not "all"?

Can you explain what it is you are actually trying to do with these "recommendations"? I'm getting the impression it may be based on similarity within a software category. Wouldn't difference have a better chance of being relevant? Chances are one has already looked at "similar" applications, and picked the one being used for a reason; an application that does things *differently* might provide a fresh look.
 
indifferent I’m confused
Inappropriate?
1 person has this question

The company marked this question as answered.


  • Stitch33
    Inappropriate?
    I find that the recommendations are relevant to me. The main purpose of the software recommendations (in my opinion) is for the user to learn about new applications on the web.
     
    indifferent I’m undecided
  • Inappropriate?
    Hey Marjolein,

    The recommendations are generated using a mathematical formula that compares your usage to the overal usage of people like you. Say, for example, that you use firefox, microsoft office and winamp, and someone uses firefox, microsoft office and videolan, then you would be similar, and you will get videolan as a recommendation.

    Of course, on Wakoopa this happens on a much larger scale.

    With regards to the percentage, yes, we cheat a little here by normalising the relevance score, so thats why it's always around 50%.
     
    happy
    Sprite_screen The company says this answers the question
  • Marjolein Katsma
    Inappropriate?
    Kilian, thanks for the reply.

    It's interesting to hear how it works, but with all respect, explaining an algorithm doesn't explain the underlying assumptions or what - as I asked - it is intended to do.

    Let's have a look at my recommendations in the last batch:

    1. Google SketchUp A 3D graphics tool - relevance: 61%

    2. Chromium A browser - relevance: 59%

    3. Monopoly City Streets A strategy game - relevance: 58%

    4. Instructables A social network tool - relevance: 57%

    5. Userscripts.org A development manager - relevance: 57%



    So what does that mean for me?

    1. Google SketchUp is a powerful yet easy-to-learn 3D software tool (...) that streamlines and simplifies 3D design.

      I do not do, or need anything 3D; I even actively avoid 3D graphics boards because they are too expensive for what I do do: a good 2D graphics card costs a fraction. Conclusion: I'm not even tempted to look at it; relevance 0%.

    2. OK, once of the things I do is web design, so I'll need Chromium at some point. But I hear it's very poor on Linux and thjat's my main development platform ... I'm dragging my feet. But a relevance of around 50% would actually be right.

    3. This one is simple: I don't game, period. Relevance 0% - it can't be anything else.

    4. (...) the Biggest How To and DIY community (...)

      Well, admittedly I do a lot of social networking. But if I need any DIY information a social site would be the last place to look, and it's not something I'd spend any time on anyway (I'd be very a-social on such a site). I'd find the necessary information elsewhere, precisely because I do not have, or want, a social network for that type of information. Nice try though, so I'd give it a relevance of maybe 20-30%.

    5. Apart from the fact that I have no idea what "a development manager" means (does it somehow manage my development? I do develop software but I don't think it is capable of managing that.

      Userscripts.org has thousands of scripts, lightweight Web Browser applications, created by a community of thousands of developers, that enrich your experienc[e]...

      My "experience" online is quite "rich" enough, annoyingly so, in fact. To help me with that, a programmable proxy could be helpful ... oh wait, I'm already using The Proxomitron, and could not find an equivalent for Linux. ;) Relevance? Well, I do browse a LOT, so I'll give it better than 0 - let's say 20%.



    So what do we have? I do need Chromium but I already knew that - I'm waiting for it to get usable on Linux: while it's "right", it doesn't help me at all. And the other four don't even tempt me to look.

    In fact, for Google Sketchup or Monopoly City Streets I do not even recognize how the algorithm you described comes into play: I don't do 3D anything, I don't game, ever.

    So it's lists liek these that prompted my question: what are these recommendations intended to do? Put differently: What are the assumptions behind the algorithm?

    Stitch33 says the recommendations are relevant to him/her, so you must be doing something right. Maybe it needs fine tuning?

    Here's an idea: if someone doesn't click through on a suggestion, lower the relevance teh next time the algorithm comes up with it; vice versa: if someone does click through, increase the relevance next time. If someone actualy tries a recommendation, "reward" the algorithm. In other words: let it learn.
     
    happy I’m hoping it can learn (it's beta, after all ;))
User_default_medium