You think this would be easier. After all, the name of the website is Search Engine Watch. Somehow it seems that talking about search engines would be simple. But when it comes to search and Second Life, it's anything but simple. It's even more complicated now, because of the Tinfoil Hat Theory- because there's a vast difference between what current users of Second Life need out of search and the new userbase Linden Lab is trying to attract. At this point it's a Gordian knot so big that untangling it seem nearly impossible. But what the hell. I'll give it a shot.
When is a search engine not a search engine?
When it's broken, inconsistent, and impossible to apply SEO.
First let's define what broken means, for all of those who are about to say something ridiculous like "it's not broken- if I type in a search term I get a result." However, the results you get may not be the results you actually *should* get, in terms of relevancy. It's further complicated when using a different Second Life viewer gets you wildly different search results, showing that the method by which search is calculated is not consistent from one viewer to the next. It becomes nearly impossible when every attempt at Search Engine Optimization is viewed by Linden Lab as "gaming the system". Yet all of these things are true of Second Life's current search engine. But it wasn't always like this. Search used to work just fine. What happened and why has it changed?
Search in Second Life is not as straightforward as you might think. Different viewers handle search differently, and some viewers seem to have more problems than others when it comes to the process. More complex than that, what's problematic about in-world search is not the same as the issues going on with marketplace search. So a bit of a breakdown is in order.
Marketplace is a purpose driven search. When shopping on Marketplace you're looking for something specific. You have an idea in mind. While you may browse listings idly, you don't tend to think "Oh I need something to do, I'll look at marketplace." It's more "Hey I need a red sweater, let me see what I can find" and go from there. Although the new design of Marketplace encourages browsing through the use of associated items, you need to know what you're looking for in the first place.
Unfortunately, listings at marketplace seem to be prioritized by what's been added to the site most recently, rather than what may be actually truly relevant to your search. This puts people whose listings migrated from Xstreet at a disadvantage, because migrated listings don't necessarily count as recent listings, even if the products have done well in the past. However, simply fixing migration issues has become such a problem (why yes, I am bitter) that wondering about how search works in marketplace becomes the least of your concerns when you're faced with just getting it to work at all.
In-world search is far more complex. Because the userbase is so diverse and scattered amongst various interests, search was divided into categories, which became subcategories. Since there's three different possible maturity ratings in Second Life, search became even more complicated still, needing to change results based on those parameters.
Each category has its own issues. For example, People Search will only return 100 results. If you need to find a specific person, you had better have an exact match for their name. Just searching by last name will only bring up the same 100 results over and over. Since the search has no fuzzy component, if someone is using an odd spelling or you manage to make a typo in search you will never find the person you're looking for. It will only return exact matches, not close ones.
However the search tab that's probably gotten the most attention (and not of the good sort) is that relating to stores and merchants. Since that's my focus, it's what is most problematic for me. The commerce forums and general Second Life blogosphere have been awash in questions and angry commentary about search (or the lack of it) for the better part of 2010. Since being a content creator can be about very real money in Second Life, the lack of a properly functioning search system is an obvious cause for concern. But we should probably go back to the beginning...
Traffic, Picks and Bots
It used to be that search in Second Life was a rather straight ahead affair. In-world locations were largely ranked on the basis of traffic. The more people you had at your location, the higher that location ranked in search results. However, in the quest for ever higher traffic numbers in order to boost search listings, various schemes were concocted to artificially manipulate that number.
First, there was camping. Camping was done by people sitting in a location for hours on end, earning small amounts of money for doing so. As long as their avatar remained in the camping area, the time it did so was recorded, and a script would pay that person a small amount per hour(sometimes half hour) for contributing to the overall sim traffic number. However, things began to get out of hand really quickly, when people discovered that with some clever scripting, bots could be created to do the work of an avatar without having to actually pay one. A bot is an avatar controlled by a computer- not a person. However, because it still counted as an avatar for the purpose of generating traffic numbers (and didn't need to be paid), bots soon took over as the camping device of choice for people looking to boost their traffic numbers and search rankings. Unfortunately they also contributed greatly to lag as well, as the more avatars you place on a sim, the more work the servers have to do to compensate for their presence.
Lucky chairs were a way of doing the same thing, except that instead of awarding cash, they award prizes, based on the first letter of your name. An entire culture has developed around lucky chairs, with freebie hunters stalking them for hours on end rather than spending money on merchandise. Though the quality of chair prizes vary, merchants who give out high quality items soon gain a reputation for doing so, and traffic follows.
In late April 2009, Linden Lab banned the use of bots to artificially game the search system. In addition, they banned camping for that same purpose- so you can still have camping on your land, but you can no longer include your land legally in search if you do that (making camping all but a waste of time to begin with). With that out of the way, they turned their attention on the picks system.
The Law of Unintended Consequences
One of the other ways search rankings used to be determined in SL was through the use of picks. For those unfamiliar, within your SL profile are a number of tabs. One of them is called picks, and it was originally designed for people to highlight their 10 favorite locations within Second Life, giving them both a search boost and a word of mouth boost as well, for those who like to examine the profiles of others.
However this became gamed quickly early on too, as merchants began granting both cash and prize rewards to people who were willing to give up a picks slot in order to promote their store. The real problem though, came through the Law of Unintended Consequences. Though initially intended to be a highlight for places within Second Life, residents quickly began using picks slots for other things- like people, or an extended place to give further information about themselves. This threw the potential results from using picks as a basis for ranking into disarray, making it ultimately unhelpful in determining rankings.
Speculation and Anger
In spring of 2010 however, not long after the release of viewer 2.0, search for all intents and purposes broke. The exact date varies depending on whom you ask, because not everyone in the merchant community experienced the fallout at the same time. Reports of something being seriously wrong began as early as late March, but really got going by April, when this report in the SL JIRA was filed. That JIRA report is still open, with no sign of resolution and others have been filed as well, relating to the same issues. Further, the search blacklist began doing strange things, like banning the term "lol", reported in August, while also permanently blacklisting lolita fashion in the process. Parcels began to randomly disappear from search that were once ranking prominently, coming back and disappearing again, with absolutely no rhyme or reason to explain why.
Jack Linden, helpful as ever, has refused to talk about how relevance in SL search is calculated, claiming erroneously that no one would ever give out such information. I know, all you SEO people out there might be sitting there gobsmacked right about now. Don't worry- so were all the merchants in SL.
Speculation has run wild about larger parcels on privately owned islands being given a boost in in-world search priority. I will say from a personal perspective that if this is true, I have not seen evidence of it. I mention this because if it *were* true, we'd have noticed- Dare Designs is a *VERY* large store on a private island, and yet the search malfunction hit it hard in June. If it were true, then Damned Good Design SL wouldn't show up with an appropriate keyword search (as it's only on a 30m x 30m parcel), and yet it shows up just fine, and has ever since I fixed the keywords in item descriptions in the store to increase density. I'm completely willing to listen to evidence that this is large private island scenario is true, but in my personal experience it's just not the case. However the real issue to me seems to be that *no one* knows what's going on - just that search is broken and that relevancy is completely screwed up in ways no one can predict or identify, and Linden Lab is deeply invested in keeping the whole process a huge secret, defying the very transparency promised to Second Life residents at the Second Life Community Conference back in July.
Search has also gone wonky depending on the maturity ratings you have selected at the time. Even if a location should appear no matter what ratings levels you're choosing, various in-world locations will drop off of search for no apparent reason, depending on the combination of maturity ratings selected. There is no explanation for this.
Heightened suspicion turned into a frenzy after Sea Linden, the head of the SL search development team, suddenly deleted their Twitter account, after what may have been an intended direct message gone awry. This has led to speculation that search results are being intentionally manipulated by the Lab to favor individuals (on what basis, I can't say). As for me personally, I'm not yet ready to jump on that bandwagon, mostly because it seems like a colossal pain in the ass. But the fact remains that the lack of information being given by the Lab on this topic continues to fuel the fires of rumor, none of it pleasant.
Tinfoil Goes On Sale
However all of these various problems may be indicative of a larger issue. Looking at when the problems began, in spring 2010, puts it right in line with these various other interconnected projects. It is not outrageous to link these issues together. The problem lies in the different needs of the current userbase, vs. the projected needs of the one that Linden Lab is out to capture.
There's really a need for three different types of search. The first is in-world search, for regular users of Second Life. The second is for Marketplace, which is much more purpose driven and less general. The third though is a tricky combination of the two, having to do with Skylight, the recent effort of Linden Lab currently under testing to put a viewer in a browser. Skylight users having both limited functionality and the potential for a completely different mindset than more traditional Second Life residents will change what "relevancy" means in terms of results. It is entirely possible that in-world search has malfunctioned whilst trying to accommodate the needs of that project, in the hopes of finding a system that would work for both ends of the spectrum (note to Linden Lab- it won't- stop that, it's damned annoying.)
It remains to be seen as to how this pans out- but without a highly functional and relevant search engine for each of these three branches of the Second Life existence, all of the interconnectivity in the world will ultimately fail. You can't do a thing unless you know that thing exists. Until Linden Lab gets the search issue fixed, that isn't happening.