Fiddling with search
I was hoping by now to be able to report upon the wonderful events at the Second Life fifth birthday celebrations, alas I haven't quite found the time yet to have a proper look. I did pop along the other day but it was a brief visit but it was no lag fest, so that has to be good.
No I've been trying to work out what the heck happens with the new search and the bottom line is that I still haven't got a clue. Something happens during the week that makes my traffic peak and trough, I'm not sure what, but there's a pattern, it would be a clear pattern if I took the time to note the days traffic peaked and troughed.
I've resisted changing keywords on my parcels for a couple of weeks now, and I'm still seeing traffic at steady levels and then at extreme levels, almost as if a couple of days a week I get raised visibility and a couple of days a week I get lowered visibility.
Whatever it is, I don't like it and this is the problem with the new search, there's not enough information about it to make it work in a reasonable manner. I want more metrics!
I want to know what search words are bringing people to my parcels. This will of course aid me, but if for some reason people are coming to my parcels for a word that isn't helpful, I can save them the trip by removing it.
I'm also not keen on the whole relevance issue, people have claimed (and it looks to be true) that a parcel name holds the most relevance. However parcel names end up leading to generic descriptions that can leave those using search frustrated, there should be more options, search by price, brand name, whatever, people need more choice to make this search work.
How about an option to find all products by a certain designer? All locations selling a certain designer's products, the ability to sort by price, surely these are options that would be helpful for the person seeking an item.
Of course not all people want to buy a product, sometimes they want to go somewhere. They may want to go to an art gallery, however finding an art gallery can be frustrating because any stores selling art will load art into the parcel name and description and could very well be higher up the search than a specific art gallery.
Then of course there is the issue of people using the search in a handy fashion, if you want "Bling Stripper heels", you search for "Bling Stripper Heels", but how many people use the speech marks? Many people will just type Bling Stripper Heels, which will return all results with bling, stripper and heels, rather than just "Bling Stripper Heels" the phrase. This leads to frustration.
The new search is supposed to be evolving, yet I've seen little sign of evolution so far since it's inception and I'm really of the mind that classifieds make little difference to overall results, although I'm constantly looking at this.
I'd just like to see a little more progress being made on the search issue and a little less secrecy about how it's working. This one size fits all solution is very frustrating.
Technorati Tags: 
Delicious
Digg
Technorati
The new search is based on
The new search is based on Google technology. A huge part of my RL job consists of gaming Google technology. There's even a name for this - Search Engine Optimization - SEO. However, you play a game, whose rules are not written down and which can be changed any day. While I see some similarities between the SL search and the Google search - for example the "link popularity" in Google (how many sites link to me?) seems to have its counterpart in the "profile picks" of (payment info) avatars - other areas are completely unclear. But just as Google, Linden Lab does not seem likely to explain what aspect bears what relevance. In Google some site try to game the algorithm to rip off others, and the same would happen in SL.
I personally think that the new Search brought no advantages. I agree that traffic - at least in the way it was implemented - was a bad way to measure the relevance. But I for my part still use the old search and dread the day the new becomes amndatory.
-- Peter Stindberg
They don't have to be PIOF
Payment info on file isn't needed for someone to count in terms of picks, but they do need to meet a certain criteria. This is I'm sure by design, but an active resident will count if they carry out certain tasks.
The problem is that Linden Lab have introduced a search appliance for places, it's not good for commerce and that's where the problem lies. Parcel names have been ruined by the new search because parcel names appear to count most in relevancy.