Pretty much every major shopping website has terrible search functionality.

I usually want something very specific, for example 60w dimmable e12 frosted warm led bulb. I have not found a single shopping website that won’t show me results without many of these terms in the description. I don’t want to see listings that say 40w and don’t say 60w anywhere, and it isn’t hard to filter them out!

Are these shopping websites bad on purpose? What’s in it for them?

  • Carighan Maconar@lemmy.worldEnglish
    12·
    2 months ago

    There’s nothing in it for them, the simple fact is that the virtual all of people does not look for specific terms.

    Hence the search is optimised to give you loads of things that relate to some parts of your search at least.

    Source: did backend code for shopping frontends for years.

    The search is incredibly fuzzy, plus the tag words of products themselves are fuzzy. And usually they don’t allow forcing a hard match search, though you can try + or and between each word. We had one site that allowed it, just use lucene search syntax.

    • AA5B@lemmy.worldEnglish
      21·
      2 months ago

      It turns out many of us do search for specific terms when we want specific items.

      • Carighan Maconar@lemmy.worldEnglish
        21·
        2 months ago

        Do you? Then how come examples like OP’s don’t really specify much.

        Is that any keyword? All keywords? Where? Tags? Title? Name? Description? If all, do they all have to appear int he same field(s)? Anywhere? On the whole page including crosssellers?

        This is what to mean: it’s easy to say “just search for exactly this!”, but what you intuitively think of as “exactly this” is not intuitive from the perspective of a search index. At all. So it gets preprocessed and changes before being used for a search, and in many cases, widened. Because we humans are very bad at putting in an accurate search such as: name:"60w" and description:"standby". We rarely do that.

        • AA5B@lemmy.worldEnglish
          3·
          2 months ago

          One of the points here was those syntaxes no longer work

          • search engines are usually free text, with no concept about fields
          • the syntax you specify usually depends on specific sites implementing “filtering” which is usually a lot more annoying to use, and people here complaining that no longer works. Plus that’s limited to a specific site
          • google search specifically, used to accept syntax like quotes to match a phrase and plus or minus to indicate required presence or absence, but those no longer work.
          • certainly part of it is merchandisers using SEO for greater attention rather than better match

          I’m currently looking for a new light fixture and haven’t yet found the magical search phrase to get there or a site with filtering that works. Of course it may not exist but all my attempted searches so far return random junk, so I don’t even know

        • Kazumara@discuss.tchncs.deEnglish
          11·
          2 months ago

          Because we humans are very bad at putting in an accurate search such as: name:“60w” and description:“standby”.

          I actually really like to do that. These days that only seems to work for flights and hentai though.

          Maybe if it was more available and people were taught to use it it could be a little more popular. I think fundamentally it’s not such a foreign concept to say that you want specific things from specific categories. People do that kind of thinking routinely when searching for homes or cars.

          • Carighan Maconar@lemmy.worldEnglish
            1·
            2 months ago

            Oh I don’t disagree. When I worked in this stuff we usually snuck in that if you put in valid search query syntax (I think it was all Lucene based) then it got used as is. Was nice for us devs to debug shit.

            Of course, anything else for used for a weighted fuzzy everything search, and the customers were always take happy with that (customers being the store owners). 🤷