The Discussions on Traffic Reform with the Lindens

Except that what I am specifically advocating is the elimination of traffic numbers. It is the those traffic numbers on the right hand side that determine the order in which those stores will be listed on the search. Hence, yes they are being woefully gamed.

As far as your categorization concern, on this website we have a "New Products" forum that is aptly able to neatly categorize all of the new products being released into Second Life. It is very basic in its implementation, yet achieves its purpose quite well. Certainly a bit of tweaking could achieve the same results in a categorized search system .



Yep, I am glad you agree that the current search system it is being currently gamed. And no, with the category search that I am advocating they would not result in higher ranking because they would be listed in a prioritized and categorized order. As stated in my prior post.

Why not have a category pulldown for objects themselves when they're marked as "include in search"? That would provide a nice subset of search data for any particular parcel.

It all seems so simple when looked at with an open mind, doesn't it Chip?

I thought it was pretty much agreed that traffic needs to go away. Except for by Phil and some other holdouts. If we take traffic OUT of Places, as has been previously suggested... then we won't have your traffic gaming issue.



Oh I get it. So if Merchant X with 200 varieties of shirts lists his stuff in Places, then he'll get a higher listing that Merchant Y with only 10 varieties of shirts? Doesn't sound exactly fair. Isn't that what you're going for here?

First, I don't think it's "unfair" that a store that specializes in the one item I want comes up ranking higher than a store that sells the item I want as one of many.

Second, though the prioritized keyword system could be "gamed" to a certain extent, the "gaming" is much more relevant to what I want as a searcher. My "sex bed" search turned up, in the top spot, a store that sells sex beds, and sex beds only. It didn't turn up the store with the most bots.

Third, "sex bed" is a very narrow search field to be dominating. That store doesn't have a great deal of advantage over every store selling a wider variety of products.

Fourth, you underestimate the searcher's ability to evaluate searches and make refined searches. Let's say I do my search, and all I get is a bunch of store listings that sell only "sex beds." I can figure out that there are stores that I may want to see that sell other things. Now I do a broader search- maybe "menu driven furniture", excluse "sex beds," and I have a whole new search.

Fifth, you could include options for searchers to include other things besides the keyword-ranking-only system. That's the real beauty of it. Any searcher who did think there were better ways to get a search could click the options for them.

What this system does lack is an easy way for a merchant to have everything. It doesn't give any merchant an easy way to dominate a wide range of searches.

But if you make a system that allows a merchant to easily dominate a wide range of searches, that's why you open up the gaming. That's why there is gaming. It's a minimization of work required to gain the maximum effect.

The problem is that many merchants want some sort of system that will protect their means of dominating searches, and exclude what they consider to be gamers.

I haven't constructed a perfect idea. However, it's based on starting off all the merchants on the same playing field, and forcing them to make choices that are the same choices that will be helpful to searchers when it comes to search results.

Every system that's designed to have some hidden mechanism to benefit certain "worthy" merchants will be gamed by those who spend all their time to solve the puzzle of the hidden mechanism, and frustrate honest merchants who are trying to place good ads but can't figure out the mechanism.

Welll.... what if Phil has his stuff listed as "Sofa - Brown" "Sofa - Red" "Sofa - Green" etc etc and he's got 50 sofas for sale? If someone did a search in Places for Shopping/Furniture and then typed in "Sofa Green" in the search field, they'd see every single green sofa listed for sale in SL. Is this what you meant? In that case, why wouldn't the searcher use Search/All which already does this?

It's worse than that. Someone reselling one style of freebie shirt and listing it 200 times gets better ranking than someone with 10 original shirts, 10 original pants, and 10 original dresses.

Actually, I took the time to suggest a keyword-ranking system because someone in this thread had challenged that people come up with alternatives rather than just criticizing what's there.

But there are three principles that drive what I think would be a good system:

1) It's transparent. Everyone knows how to influence the searches.
2) It's egalitarian. Everyone who wants to place the ad has the same tools to influence it that everyone else does.
3) It has options. That allows the Searcher- the sole expert on what a Seacher wants to see- to influence the way results are displayed so they are meaningful to the Searcher.

Systems fail that:
1) Are mysterious. They reward the lucky, or those who over-invest in solving the mystery. The punish the honest people who want to create a good listing but can't figure out the system.
2) Weighted toward certain merchants. Not only to such systems protect the "old guard" from newcomers, but they also protect the gamers who figure out how to manipulate the system the best.
3) Limited options. With everyone telling me, the Searcher, what results I should care about, I never get the results that I actually care about.

There is no system that cannot be gamed in one way or another. Let's get this pipe dream out of mind right now. What we need is a system that provides relevant results and is as resistant to gaming as it can be. I am of the mind that the new search does this pretty effectively, and that gaming it effectively through picks camping costs money to do, and is going to be prohibitively expensive at some point.

Traffic is a metric that should be visible in About Land, but not the basis for search. I am fine with it as it is used in the new search, since it is minuscule and may serve as the equalizer to a copycat that copies the search terms/description of a competitor's parcel.

Like Lias, I have just come back form the first of the meetings, and I don't have any heart to discuss these search methods any more. I think it's wasted effort. Showcase is pretty much a done deal, as Lias said, and I doubt that they'll take any notice of suggestions like 'make it non-profit only' I did plug that idea though. That was the first part of the meeting.

The next part was about what metrics parcel owners want. That was good, and they'll likely try to produce some, but I fear they will use the new metrics for search rankings as well. In other words, they'll increase the Places ranking factors and make more ways to game them. That's just what I think though.

The last part was about traffic, and I've no idea what will happen to it. They are not planning on removing the Places tab though as was thought.

Great. Can't they see what's hitting them over the head???

I give up.

Someone else can work it out.

My own sentiments entirely.

It all depends on how you want to prioritize your items. The number of shirts you list is not a factor in were you fall on the listing. It is, in essence, an alphabetical directory. Save for their alphabetical order of listing, all of those listing 'shirts' as their main priority item will all have the same priority listing. Which is higher than all those who list 'shirt' as their second priority item.

The most important factor about this categorized search system is that under this system everyone has a fair chance to be seen as indicated by their priority offering, or that merchandise that best exemplifies their business. In other words, you will no longer have "Acme Designs," which is primarily a dynamite selling store, coming up first on search for, say jewelry, or for each and every tagged category in their description solely because they have that named merchandise tag in their description even though they may only have one jewelry item on display in their store, and in fact received all of their traffic for dynamite sales only.

But why are you making me repeat myself over and over again?

You won't find a Linden who doesn't describe traffic calculation (# minutes/avie) as dwell (dependent on the % of time you spent on a parcel, with an upper limit) which hasn't existed for 2 years now so if you accept things a Linden states at face-value you're going to be led astray at times. There's nothing wrong with being critical, but apply it to both residents and Lindens.

As far as what Phil says about traffic in the new search: you have the "Popular Places Directory" which consists of 12 distinct pages that are nothing but a sorted list (by traffic) of all parcels in SL, capped at a certain amount (Top 10, 20, 50, 100, 200, 500, 1000, 2000, 5000, 10000, 20000 and 50000 - or 12 total).

If you manage to get into the top 10, you get an extra of 12 incoming links. If you're #25,636 you get 1 extra incoming link and if you're #51,254 you get no extra incoming links based on traffic.

If you get 12 extra picks (assuming everyone is listed in the new search), you get the same benefit as you would being #1 through #10. Make it 13 and you have more of a benefit in the new search than you'd ever get in it through traffic, which is why traffic is fairly irrelevant in the name search as a form of gaming. It's much easier, cheaper and least of all sneakier (noone can tell you're cheating at first glance) to game incoming links.

Linden confirmation would be nice, but it wouldn't change anything. You can see the links with your own eyes, they're all equal.

Well, the reason I've spent a lot of time talking about traffic in the forums, but little time at Linden meetings at such, is because I've always suspected that the Lindens probably don't care what I have to say about anything.

Sounds like I was right on that.

That's the procedure at Linden Labs.

(1) Decide what we want to do before we ask for opinion.

(2) Ask for opinion to pretend that we care about the customers.

(3) Highlight the opinions that just happen to agree with what we'd already planned to anyway, so we can say we listened.

(4) When the idea fails, we'll point back to those opinions that agreed with what we did anyway, and offer that as cover, "everyone agreed with what we were doing at the time."

Because you can't tell me whether you want merchants to list individual items for sale, or to select a category in which to place those items.

Besides, it's all moot now that they're not going to take traffic out of Places. Knock yourself out trying to improve that.

In the first few weeks you could use the link: operator to see which pages linked to a certain other page and how keywords ended up either matching or skewing the results, which is how I found out that there are 12 popular places pages to begin with.

Correct me if I'm wrong.... (and I didn't attend the meeting).... But my understanding:

-There are no plans to remove the "Places" tab.

-The status of "Traffic" being used in the Places tab, and elsewhere.... is up in Limbo.


If the above is accurate, that doesn't mean traffic wont be taken out of places. Or that it will be. The only thing we know is that the "Places" tab is staying, right?

The store will be listed based on a general "business model" category. Much like the ones avalable now, but much more refined. That is not to say merchants cannot go into detail as to what they offer in their description. But they will be found in search based on the generalized category that best describes their business model as prioritized by the respective merchants.



I don't think they've decided on the traffic numbers yet. Or maybe that memo has yet to cross my desk for approval. Ill have my people keep an eye out for it.

Search / Places will be replaced with a shiny new search HTML page that uses the new search.

*has no illusions they'll bother fixing the in-world browser bug that keeps people from being able to scroll it so the net effect will be no usuable search of any kind*

So then you mean "list your categories" as opposed to "list your items."

I didn't notice that, Kitty? That's a big problem with fast moving chat between 30 people - it's hard to follow. I'll read the chat log

Jeska said it in response to your question even .

And people say it's the Lindens who don't listen, tsk... (teasing ).

I pretty much agree with you. I'd be satisfied if they just stopped using traffic or picks as metrics for listing order and went only by relevancy to the search terms. Listings that have the same relevancy score (or even within a bracket of scores) could be listed in random order. That takes away much of the incentive to game the system.

Well random or by whom paid more for their classified, the cost of a classified is an easy to understand metric and would actually put some worth back into the classified system.

However I doubt they're going to abandon picks as links.