What is going on with the SL Search ???

Oh well, I'm not actually talking about that. What we actually did to increase our rank was nothing more than keyword stuffing on the parcel description which I had eschewed until now as crass and ugly. I don't think User: pages on the wiki are even indexed, for example, they don't show up in the search. Well, anyway, it was worth a try, maybe it is helping.

That's what I meant - keyword stuffing. The links from your wiki page go to your parcel's html page.

While I've been playing with the 'furniture' ranking, it occured to me that most or all of the higher ranked pages will end up doing it, and the pages will become less and less useful for users, even becoming totally useless in the end. It concerned me enough to write to the LL search people, stating what I see as being a major future problem, and suggesting what I think is a very good solution. I got the email address from Jeska, but I didn't send the email yet. I'm not 100% on sending it yet.

Yes, I'm ! Trying to find some answers trough patterns.



Yes, Phil ! that's the point ! Obviouslly sometimes some1 really changes something.But usually ppl dont' change so much things and fall down anyway.And sometimes someone pump up with no reason too.As I said: It's a pop-corn cooker

Henry. I have to say that I haven't seen anything like you described in the few weeks that I've been involved (in detail) with it. I see people making changes that affect their ranking in one direction, then making a similar change that affects their ranking in the other direction, but I can usually see what effect their changes are likely to have even before the index updates. They are groping in the dark with the changes because they don't really understand what they are doing.

I even saw one person change the page and then, when the search system returned the changed page, the person thought that it hadn't worked, so they changed it again. But they hadn't waited for the first change to be indexed, so they got it wrong. There is a large part of every day when the pages that the All search returns are *not* the pages that are being ranked.

I'll expand on that...

There are only 4 hours a day when clicking a listing in the All search results guarantees to return the actual page that is being ranked.

There are 2 parts of the All search system - the GSA itself, and the SL system that provides the parcel pages for the GSA to index and rank.

The GSA index is updated once a day, at around noon SLT (it can sometimes be very late).

The SL system (the parcel pages) is updated twice a day, at around 4 a.m. and 4 p.m.

When an All search is done, the rankings are returned by the GSA, according to the pages in the current GSA index. When clicking on a ranking, the parcel page is returned by the SL system and not by the GSA. The GSA merely provides the URLs in the results and not the parcel pages themselves.

Approximately 4 hours after the GSA index updates, the SL system updates the parcel pages and, from then until the next GSA update, 20 hours later, the parcel pages that can be seen via the All search results are not the ones that the GSA has currently indexed - they are not the ones that are being ranked. Most parcel pages won't have changed, but some will have changed.

So, if you are looking to see why a page ranks where it does, you only have those 4 hours (approx.) a day to be sure of seeing the page that is being ranked. I.e. you need to look after the GSA index update and before the next SL system parcel page update.

ETA:
If you followed that, you'll realise that after the 4 a.m. SL system parcel pages update, you can see the pages that will be ranked following the next GSA update at around noon. Once you see them, it's too late to react to any changes that people have made, of course. However, if you see changes after the 4 p.m. parcel page update, you do have time to react and make changes to your page, as long as you do it well before the 4 a.m. parcel page update. But it's best to have a good grasp of what you're doing before reacting with changes to your parcel's page, or you may react to something that wouldn't have hurt your rankings, and your reaction could be the thing that hurts them.

What people might be experiencing is a bit of what Jack Linden said they'd be starting
"In the next few months, we will be making both technical and policy changes to the way relevance and ranking works in Search. The "traffic" score will be only one aspect of the ranking logic, and it will be scrubbed and weighted to account for gaming vectors. Our goals will be to (1) provide an informative and delightful search experience for all residents, and (2) increase fairness and reliability of the Search service for people who place listings. So Traffic will remain as part of Search ranking. Stay tuned to the blog to learn more about Search improvements and to join the discussion."

Rankings are going to change even if the ranking procedure is stable. People change things or new parcels get flagged for inclusion. But ... it's not going to be stable for a while.

If grid/search monkeys are banging on things, they would be wise to
- try something and review the results.
- revert the something
- think about it
- try something else
- etc.

Changing more than one thing at a time only introduces confusion.

With that approach, it should be expected that we'd see a patten of noticeable swings followed by reversion to "normal".

I would posit that if everybody begins to employ the same techniques outside the parcel page to 'enhance' their ranking, then the results would be more useful rather than less useful to users. While only a minority of people use the techniques, this can not be said to be useful to searchers. It is only useful to the parcels.

If the gaming effects are balanced out to a great extent, the differentiator left to ranking is the actual parcel page content. This is assuming that page games such as keyword stuffing are either ignored or simply balance out.
The only people seeing this as less useful (to themselves) would be those who were early adopters of the techniques.

The most sensible thing that LL could do would be to allow parcel owners more space to describe their offerings in detail. Let the indexing take it's best shot at ranking based on that.

Inbound links within SL simply can't have the same weight as they might out in the wider Web. This is because they in essence internal pages of the parcel 'website'. They don't add anything of value for the searcher except possibly as overflow for relevant content that is impossible to fit within the existing parcel pages.

I read that in the blog at the time, and didn't give it much thought. I suppose it's possible that something of what is seen in the All search could be caused by that - something did change a while back - but I think it's unlikely. That change was before the blog that Jack wrote, and he wrote about the changes as being in the future, but it's possible they could have tried a little test or something.

It's difficult to know what Jacks means by that. He could be talking about the All search, the Places tab search, or both. I favour the Places tab search because traffic plays such a tiny part in the All search and "scrubbing" it will make such a negligable difference. It could be the reason why they didn't decide to change the content of the Places tab search at this time.

I think you missed where traffic is only a part of what is being tweaked.

So they can be tweaking how it rates picks, items on parcels, keywords, etc

I can't say that I understood all that. It sounded a bit like you were talking about something different to what Elan and I were talking about. We were talking about page content (at least I was), which is definitely getting worse for users and, imo, will continue to deteriorate until the pages become allbut useless for users - except to look at the pictures and TP to the parcels.

I didn't miss it. He mentioned traffic a couple of times in that quoted part, which seems to indicate that traffic is something that's influencial, albeit only a part of the whole. But traffic in the All search is almost nothing, so there was no need to make a big thing about scrubbing it and such. I fancy he was talking about the Places tab, and something along the lines of a new search system coming to it.

Also, the blog post itself was largely to do with the Places tab - that's the only reason for traffic bots etc. But he could have meant the All search, or both.

I don't agree - at least not in the way I think you mean. I think you mean in the Name and Description fields. More space would just mean more space to make useless for users.

In the end, we are talking about the same thing, and in agreement on it.
If the page-external techniques are levelled out, then all that is left is the page itself.
What would make the page content useless is the use of the limited space to try and influence the technicalities of search indexing rather than try to describe the offerings in a way that is useful to a human reader.

For instance, it is insane to imagine that a parcel name or description stuffed with keywords makes that parcel somehow more relevant to the keywords as far as *the user* is concerned.
It is therefore inappropriate for the indexing to rank on repetition.


The longer term answer within SL is to
- expand the space available and make the parcel page almost exclusively the determinator of ranking.
- explicitly outlaw the classic in-page gaming techniques, and mitigate their effect as far as is possible.
- explicitly state that that the list is not comprehensive
Take ARs if necessary, but minimise their possibility by auto-detecting abuse attempts.

I do have what I believe is a good solution, but it's not like any of that, and it has the advantage of not requiring the outlawing and policing of anything.

I've cracked 'search all' again! (for about the 4th time).

As previously reported, my main store had slipped from the top rankings down to page 34 overnight on keywords like 'prefabs'. No obvious reason to this. Fretted and tinkered for 3 days then yesterday it occured to me. The solution I mean.... and today, the rankings are updated and back on page 1 baby!!

Who's da man? (yes that's right, it's me, Porky Gorky )

You da man!


(Now if only you could get a name that suits being da man. Porky just doesn't sound quite right )

You can just call me Mr Gorky young Phillip

Yes sir

I've just sent my idea to the search team, so I'll explain it briefly here - in case anyone is interested.

The problem with the current All search is that everyone can see everyone else's html pages, so we can all take tips from them and try to move up the rankings. But the tips consist of keyword stuffing, and it's becoming a case outstuffing each other. I see it happen almost daily. But if we couldn't see each other's pages, we would have no way of knowing why some pages rank above us, and no way of knowing how a place managed to suddenly jump over us.

So the idea is to have a new field, sufficient to hold about 500 or 600 words. Parcel owners can optimise that field to their heart's content and use all the keyword stuffing they like, because nobody else will see it. When the GSA request's a parcel's page for indexing, it is given the content of that field. When a user requests the parcel's page, by clicking on it's listing in the search results, a different page is returned - the normal "Objects found here" page. It's simple cloaking. The GSA ranks according to the new field, which only the parcel owner can see, so nobody would be able to copy a higher ranked page, because they could never see the page (field) that got the ranking.

It has the disadvantage of parcel owners simply filling the available space with keyword repetitions but, because of the way the GSA works, that won't be sufficient to get to the top. Even so, the idea could be done the other way round so that the GSA indexes and ranks the "Objects found here" page, and the new field page is used by parcel owners to 'sell' their places - like an ad. Users cannot see anyone's "Objects" page - not even their own - so objects can be used to optimise, just like they are now. I prefer it done this way round. It would mean abandoning the orginal idea of making a nice list of the objects, but that's pretty much going out the window now anyway, and it will go out the window more and more over time, until the pages are useless for users.

The whole idea is to prevent users from seeing the pages that are being indexed and ranked, and to return nicer pages instead. Either way round, the parcel name will have to be included with the GSA's page, for use as the clickable link in the results, and with the viewable page, so that would still get optimised. But the parcel description need not be given to the GSA, and it will be used to nicely describe the place again.

That's my idea/possible solution which I've now emailed to the search team.

Interesting...

I've got an even better idea. Give business owners the ability to choose just five keywords from a predefined list.

Pep (Saves a lot of time for everyone)

Not if that shop makes tons of different products...why should it be penalised?

I'm giggling in RL.
That idea is so.............. wonderfully ........ predictable




Yeah. That 's a bummer.

Game Search by running bots - everyone can see it an copy it, and do more of it
Game Search by buying Picks - - everyone can see it an copy it, and do more of it
Game Search by keyword stuffing - everyone can see it and copy it. Due to limitations of space. They can only do more of it by making everything a string of keywords.

Gadammit!!
With all these people copying gaming techniques, it tends to reduce the benefit of gaming for the gamers.
The next thing you know, people will be reduced to using the space available to actually describe their offerings clearly rather than chewing it up with stunts.

You want to facilitate stealth keyword stuffing?
How does keyword stuffing benefit the searcher?

Hint: "It helps them to find my stuff before they find other peoples' stuff" is not a benefit to a searcher. They can find anybody's stuff.

You don't have a grandfather called McCarthy, do you?

My thinking is that we parcel owners who want to be high in the rankings can see how others got high in the rankings and we can copy what they do. The main method that we see, concerning page content, is keyword stuffing. Already, higher ranked pages are pretty much useless for users and I can only see it getting worse. The pages are no longer nice, informative lists of items on the parcel, as LL designed them to be. My idea is nothing to do with fighting against what can't be fought against - keyword stuffing. It's simply to make the pages that users see worth seeing and useful.

So, to answer your question, keyword stuffing isn't useful for users. What would be useful for users is seeing good, informative pages instead, which is something that they can't do even now with the higher ranked pages. That's what the idea would produce. What goes on behind the scenes would be irrelevant as far as users are concerned, and the idea would be aided by the fact that nobody who would want to emulate it can see what to emulate - they would be groping in the dark, but that wouldn't matter because it wouldn't be stuck in front of the users' faces. Users would only see nice, informative pages, which is the whole point of the idea.

If you have a better idea, let's hear it, but remember that the problem I am talking about is the deterioration of the html pages for users.

Apart from what Rene said, we are talking about the All search and that can't be done with the All search.