Polymath Snuggler

I said it in the last thread and I'll say it again ~ since people are so intently focused on DEMO items. ( I can't really blame them~ I'm focused on being able to exclude them as well ~ HOWEVER!!! ) There is a substantial fraction of marketplace listing / visibility that is maintained exclusively through the popularity of the products demo listing. Let me give you an example. Let's say I'm a vendor selling VeryAwesomeBoxes. I have a 7 listings for my VeryAwesomeBoxes. They are as follows. Yellow VeryAwesomeBox : Sold 2 units. The ASSOCIATED DEMO for my Yellow VeryAwesomeBox : Sold 79 units. Blue VeryAwesomeBox : Sold 9 units. Red VeryAwesomeBox : Sold 11 units. Green VeryAwesomeBox : Sold 2 units. Black VeryAwesomeBox : Sold 23 units. White VeryAwesomeBox : Sold 15 units. Now~ since overall Relevance ranking is tied to prior item sales and since each individual color is it's own search listing. As far as the system is concerned there's no way of identifying that this bunch of listings as a *product group* has sold 62 Units. So when someone searches for "boxes", the varying colors of VeryAwesomeBoxes will be tremendously low in search ranking due to their low individual sales. But the DEMO is near the top of page 1. The demo is what keeps this product being visible and finding buyers. A demo which I may remind you is only tied to ONE LISTING, which has sold 2 copies because I was a stupid merchant and made my Banana Yellow VeryAwesomeBox my primary listing instead of my White one. But in this scenario, which mind you is very very real the DEMO ~ ironically is the lynchpin that keeps the listing alive. So ~ as you can see demos and colors are very closely intertwined in terms of how they need to be handled. Which is why both are out of the scope of this project. Both involve going back through the item database and recatagorizing and reattributing sales populartiy from one item in the database to another as well as grouping massive quantities of products based on a new ranking scale. THIS IS NOT A TRIVIAL TASK. Searching for "NOT DEMO" presently removes items titled "DEMO" from the search results. If you need to use it it's there. But I can understand LL's hesitation to implement it as a standard. Edit: Sorry Dakota ~ was writing this post while you were posting yours!!

Boolean searches for NOT [four letter word] are now properly excluding wildcard-type results such as **[four letter word]** and *[four letter word]* Thank you for the fix ! ♥

When a search of any type is performed it asks the database for a result and the database answers with a list of what it thinks are relevant items. ( not to be confused with sorting by [Relvance] ). Returned search results are determined by: [ This list is not exhaustive, just the ones that I am aware of at the moment, and it's worth pointing out that LL is adjusting this constantly, that's the whole entire point of this entire beta Marketplace project. ] Associated Keywords : How the item is marked. Listing Title : If the listing is titled your search term~ It's likely to show up in the search. Store name: If the store name had a possible connection to the search the term the results would show up. Approximate keyword / Fuzzy results: This is a new critera that the new search seems to be doing, If you enter Dutchie, it'll pull up all of Froukje's store first ( YAY BIG IMPROVEMENT ~ way to go LL! ) and then some other things like Dutchy, Dutch Oven etc etc. followed finally by listings for Hutches. First 150 characters of the listing description.: Used to have some bearing ~ I think~ maybe? Now we're getting into voodoo-guesswork ~ [Relevance] : Is the default ordering for the marketplace search results. "Relevance to search term." The returned results are (intended to be) a balanced result tuned to take into account the following factors:: Sales popularity: How well this particular item has sold. Novelty: how 'new' the posted item is ( IE was it recently posted ) Reviews: How much positive / negative feedback a specific listing has recieved. Older algorithms that I don't know if they've been reimplemented used to include: Overall store popularity: How well this stores other items sell. If a popular store has a listing relevant to the performed search, even if the item itself is not a best-seller, it'll filter up a few notches in the search results. [best Selling] : Is the ranking tab that's meant to organize a search result by EXCLUSIVELY what sells the most frequently. Excluding most other factors, specifically Novelty, possibly Reviews as well.

Mesh composition of an item should not be at the forefront of search. I agree 100%~ I'd agree 1000% if I could. Adding a searchable "Mesh" catagorization is completely superfluous and ultimately detrimental because the "percentage of mesh" as a metric of a build is so utterly divorced from actual build quality that it's absurd. Mesh is not quality, Mesh is not efficiency. Mesh is a misnomer for "modern". "High Quality Mesh" is a misnomer for high poly count. I hate having to buy into it as a creator~ and mind you~ I say this as a designer who works almost exclusively with rigged mesh and materials. The mesh / not mesh catagorization of items is a loose~ flexible and catastrophically misleading metric that 99% of customers will never be able to comprehend. Making it searchable brings that flawed system to the forefront and creates a problem where one doesn't presently exist.

Suggestion & Bug Report BUG REPORT So ~ since Boolean Keywords have been implemented I decided why not just search "NOT DEMO" So I did that~ and it worked~ sort of!! All products I have listed titled as "DEMO" were not showing up in the search results ~ Woo!!  However the results were also scrambled beyond recognition as though the Relevance ranking algorithm was not applying to the actual search results yeilding apparently entirely randomly ordered results (This would be the BUG). Compare it to an empty search field "Relevance" search of my store and you can see the ranking difference. Note:: This behavior seems to occur all over the marketplace for boolean searches, when using the "NOT DEMO" term results dig up a completely different prioritization of results compared to an empty search field "Relevance" ranking.  Part II :: SUGGESTION It was specifically stated that "Demo listings as a separate category" was "out of scope of the project" We don't need Demos to be their own catagorical listing~ The notion simply exists that enabling customers to easily be able to remove Demo results from their search would improve the shopping experience overall. ( less things to sift through ~ less duplicate postings, more variety etc etc. ) In the interest of usability and convenience, and since not all people are savvy with the art of boolean search terms, nor will they remember they exist. I propose the addition of a little radio button below the search box that will append a "NOT DEMO" to whatever the entered search request is~ I was thinking~ maybe something like this? ( I just photoshoped this image together~ )  Set it up so that hitting "Search" with a configuration like this would perform a search for "Toothbrush NOT DEMO". Edit : I'm aware that this is a slightly hacky solution to a more complex problem, but in the interim it would improve usability I think. 2nd Edit: Further fiddling has revealed that other stores who list their items as *DEMO* and **DEMO** do not get caught by the boolean search. Perhaps it would be possible to wildcard search "DEMO" surrounded by non alphabetic characters? 3rd Edit~ Okay so searching for "Insanya" yeilds serach results of :::::::insanya::::::: (which is good) but searching for NOT DEMO does not exclude **DEMO** ( which is confusing )

Current Best Selling vs Relevance So ~ you switched shops to Best Selling as default ranking. Which is awesome! Only problem is current best selling is a bit..... inaccurate?  But "Relevance" ranking now orders results how I would expect "Best Selling" to do. Mind you the ranking is flawless for Best Selling criteria~ except this is "Relevance" ( I know the text is blurry. ).  Soo~ somewhere something got confused.

They are quietly working on it indeed. Many of the searches I've been doing have been performing similarly to the live MP at the moment. Slight variatons etc ~ but functional. It seems that they're in the process of fiddling with all the knobs and dials. I've been told we'll get a formal update when they feel there's a substantial update in the overall behavior.

Oh I'm not belittling the importance of marketplace needing to be done right. 85% of my sales come from MP which is a substantial percentage of my monthly RL income. Which is WHY these problems need to be addressed *correctly* in a calm sane manner. The narrative this has taken these last few pages has mainly been ranting and sputtering and very little productive chatter. So~~ instead of sitting here and commiserating about how our doom is inevitable ~ and everything's a 'done deal' because LL is clueless whilst screaming "YOU'RE FIRED" at random people ~ cause you know ~ that's productive.... Why don't we pick out the details of what we need~ what we want, and how we'd like to see it done? So with that ~ let's get on with the technicalities. What priority should be assigned to sales frequency / popularity for overall relevance? If the review system is put more front & center / overhauled ~ how much relevance should that sub-system pull? How do we deal with keywords? Some people suggested reducing the number of allowed keywords. Since reducing the number of allowed keywords will make older listings "more valued." as they will turn up in more searches~ is this acceptable? Another person pointed out that they should have their relevance "tuned down some." To reduce the exploitation of them. What should take their place? How do we filter store names out from listings when we as creators stick our brand names on everything we make? Do we try and compare store names to listing names in an attempt to exclude store names from listing somehow? Is that even possible? Are we willing to re-list stuff? Are we willing to edit our existing listings? Is there some sort of automated handler LL could implement? ( is that even feasable? ) How about listing naming conventions? How do we handle the Demo situation? Do we delist them? Do we give them their own unique status on MP listings? If we delist them~ How do we deal with retroactively processing the existing marketplace database so that Demos are able to be associated with more than one listing. Is there a way to accurately automate delisting of the thousands and thousands of existing demos on the marketplace? ( Both associated & unassociated? Will the delisting of those items negatively impact sales of the merchants who own them or will the sold item gain the ranking of the associated demo. Will the sales relevance of the existing demo be added into the ranking of the existing product? What sort of "marketplace merging tools" should be created for managing our listings? What if that demo *should* be associated with multiple products? If we don't delist them and opt to remove them from search results via a tickbox.~ How do we deal with the fact that for many merchants their demos currently have far superior listing relevance than the item they're selling? How do we deal with demos that are presently "unassociated" namely ones for gatcha items or for expired listings? How do we condense mutliple listings for color variants. What sort of visibility priority would a multiple "variant" sales listing gain over a dozen individual listings? How do we compound various sales statistics from a multitude of slight variant listings. If there's an automated "variant association algorithm" available ~ do we trust it? Are we willing to put up with reassigning miscatagorized products? How much work are we~ the creators willing to do for a possible improvement in visibility? Are we willing to pay added premiums for a professional marketplace environment?? There's a lot of very important unanswered questions. And a lot of discussions that need to take place. It requires conversation between us ~ the merchants and the commerce team. So please~ Because this IS important.

Yes there's a substantial fraction of items labeled "DEMO" that are associated with a single item. ( Huzzah ~working as intended ) There's also a substantial number of items that aren't associated with a DEMO as well~ since it's restricted to 1 item-demo link. And as was mentioned prior in this thread, there's literally 1000's of demos to removed listings, just floating around on the marketplace. As well as original demos posted for people who do Gatcha resale. Those demos aren't linked to any specific listings either~ they're just there as a conveneince for people. Bottom line: SL is a user created entity, even it's infrastructure. Improving on something so fractiously generated is NOT a simple task. To further compound that people compose work arounds to an issue which generates a spiderweb of related problems. Multiple color variatons of listings is a nice idea~! I'd love to see something like that implemented. Easy to implement ~ but difficult to retroactively determine. IE all the old multi-color-listings are going to still have to be dealt with in a way that doesn't compromise visibility. This means somehow creating extra listings that point to the database statistics from 15 different listings. Perfectly doable mind you for standard listings~ a little more difficult for gatcha resale which is 100's of the same item posted in various shades. I'm sorry I'm raining on the pitchfork festival with logic and reasoning. I sincerely hope this discussion moves forward towards problem solving as I think we can all agree that marketplace search is a multitude of complex issues that all NEED to be addressed.

Okay~ I get that people have been dissappointed in the lab before. That has been made inescapably clear by the blind cynicism and ongoing negativity. My point I've been trying to make the last 3 posts is a lot of the requests people are putting forth here haven't been thought through with a "feasability" mindset. You can sputter with indignant rage all you want that you're not getting your golden hen that lays golden eggs of golden golden-ness, and "god damn LL is worthless at solving the lack of golden hens problem". But that narrative is not solving the fundamental underpinning issues that a lot of what we're suggesting be fixed ( and what genuinely NEEDS to be done ) has some very serious technical issues that are unresolved. Pretending that those don't exist and posting "is it fixed yet, is it fixed yet?" isn't solving anything ~ nor is it even helping. On that note I'm going to try and redirect this discussion towards something more productive. Problem #1: Creator & Store Names Removing creator names & shop names from search results has a compounding difficulty that a substantial portion of our items we sell as creators are labeled "[MY SHOP NAME] GoldenChicken". Or {MYSHOPNAME} GoldenEgg. We habitually do this in the hopes of creating some sort of brand name recognition. And it's an important part of that. But it throws an absolute monkey wrench in the idea that such names need to be excluded from the search results. Keep in mind we do this for listing titles, listing names, listing keywords, item names ~ each of us has decided to stick our store name in a different place. Potentially you could scan for similarities in shop names to item names and then try and match and compare that to the store name but then you wind up with incorrect comparisons against abbreviations as problem. But what if I'm selling items another vendor created that's a gatcha resale. Then suddenly the store name is 100% critical on what the item is. XYZ Golden_Egg [RARE] needs to be relevant when I'm searching for "XYZ" + "RARE" Now to solve this you could naming conventions for marketplace listings but there's literally millions of items that are for sale on the marketplace ( many of them fantastically made and still best sellers ) that are made by people who have left SL and are no longer managing their account for various reasons, in many cases those reasons are exceedingly valid, such as being deceased. Their store is their lasting legacy. So updates *MUST* be automated and accurate. If you have a solution to this sort of thing please post it. Problem #2 : DEMOS Existing listings. Again ~ an offshoot of the prior dilemma. A lot of people have chosen to label their demos differently. Not everyone with a marketplace shop currently is there to manage / curate it. Catagorizing existing listings into specialized "DEMO" classifications automatically cannot be 100% accurate. Not everyone's demo items are linked to a main item. Retroactively doing a decisionmaking process on this is not a simple problem. Things to consider on this: From the Labs perspective every purchased item on the marketplace that funds an inactive account is pure profit for the Lab. As a businses they have no interest in removing these listings from the marketplace. Ever. Nor should they. SL is a user generated world. People should be able to buy bizarre stuff from 2004. If you have a solution to this sort of thing please post it. Problem #3: Keywords Revamping the keyword system is difficult due to the fact that LL has catagorized and listed everything by keywords. Certain keywords are associated with different ratings and visibility. LL attempts to keep keywords secret in this regard in and attempt to reduce the number of people who exploit the system. Reducing available keywords (as was suggested) would hinder the visibility of newer products compared to the older ones overall which is the exact opposite of what people want to have happen. Removing them as a ranking metric entirely would effectively render all entered keywords as a negative impact on search listings as the only thing they'd impact would be G/M/A Visibility. These are just the 3 that popped into my head while I was half awake and pondering. Nevermind database management concerns~ Because I seriously haven't a clue about those. My point is this: It's NOT SO SIMPLE. So if you really really believe LL is as utterly incompetent at handling your needs, then guide their incompetence, rather than just complain. These aren't even coding issues I've posted they're logic problems.

Our feedback as a community has been basically "restructure your entire project." IF they take our advice~ I'd expect turnaround times on the order of MONTHS ~ because computer programming isn't something you just flip a switch on and "make happen." I refer you to XKCD. 

A mandatory review should not be instated. However a prompt to review should be attached / activated in some way after delivery is completed. (Sassy had an EXCELLENT idea on how this could be done ) All reviewer names should be recorded along with those ratings. Reviews should be able to be submitted without text, not everyone wants to take the time to write out a little blurb about the "why". More often than not "3-4 stars" is more than enough information to communicate "This thing I bought sort of met my expectations." If the shop owner is curious enough ~ they can then seek out the reviewer and ask the 'why'. Achieving statistical relevance is important in this. That is the goal. It's worth noting that the 1 star - 5 star dichotomy is not an exclusively SL related problem. Most review scores for other online marketplaces are predominantly 1 star and 5 star ratings averaged out to an overall 3.5-4.5 star rating. That's just how people answer these sorts of feedback questions.

YAY a reply! ♥ Okay ! Cynicism, dark humor and irritability aside: Here's a long winded in-depth analysis for you Commerce Linden people!:: The current Marketplace Sales Environment (non BETA) suffers from some serious but non-critical flaws. Here they are in rough order of priority. 1: Dependency upon customer reviews: Reviews are an IMPORTANT part of a virtual marketplace economy. (see Amazon.com, Yelp! Newegg etc etc ) The problem with the current iteration of SL Marketplace is that customers rarely if ever have any incentive to place a review for an item unless they're dissatisfied with it. I've sold roughtly 1500 units in the past few months and the number of positive reviews I've gotten without interacting with my customers has been 2~ I think? That being said, I have people who bubble and cheer effervescently in IM's with me over my work. But they don't stop to think "hey, I should leave this girl a good review" unless I specifically nudge them towards that direction. Properly solving this problem unfortunately involves developing new client-to-website code I believe, as the most assured way to have people review things they've purchased would be to have an inworld "rate me" popup ( typing optional ) after an item is delivered ( or maybe removed from the customers "recieved items" folder? ) I understand in-world to website integration is hitchy and difficult to implement at best. But this would fix most of the reviews problems as "revenge" reviews would be drowned out by the many valid ones. Likewise obvious review exploiters would be less impactful overall. 2: Inclusion of Content Creator and Store names in the general search field: This singlehandedly muddles the marketplace up for certain queries to the point of utter unusability. There's even a "Merchant / Store" alternate search field right at the top of the page for people to use that SHOULD DO THIS. But not the "General Search"!! It's like the idea was thought about, but never implemented!! Just to use Trompe Loeil as an example. ( Cory is an excellent SL merchant, she's a great example of what we aspire to ) But if I'm looking for Edo era Japanese paraphanalia, chances are I'm probably not looking for Trompe Loeil furniture. The problem is I can't find anything relevant to my actual search because I'm having to sort through her entire catalog of items before I find anything else. Here's where it's important to distinguish: Cory deserves every single bit of her earned search ranking. She deserves every single bit of her visibility from sales popularity. When I search "White Bed" she should have the #1 result ( actual return is like ~ #4 on the current system !! working as intended! ) Just not when I search for "Edo". 3: Lack of priority for newly posted items: Posting a new entry on Marketplace should provide some measure of visibility for a short duration of time. ( 7-14 days maybe? though I'm sure some people would find a way to exploit this ) It shouldn't outshine the tried and true best-sellers, but having your new products place in dead-last is always a tad bit disheartening. Perhaps giving stores the option to customize their own storefront so that their shop-main page can have a "New Releases" bar in the same spot where the "Featured Items" bar exists for the standard listing results would be a nice addition to the MP experience. 4: Keywords: Keyword spam / abuse. Keywords & Catagories used to be absolutely essential for a proper search, back in like 2007. The problem was in the exploitability of this. For a few years google tried various tactics to prevent abuse of keyword & tag spam but they ultimately gave up on trying to prevent exploitation of it, and opted to just not use them at all. Instead they scan the text of descriptions and actual website contents for parameters that they were looking for and used that in lieu of the tags. Since the SL Marketplace was designed around the keyword system though, I don't think removing them entirely would be possible as everyone would then have to redo every listing they own. If we learned anything from the VMM implementation, some people have literally 9500+ listings... I'm sure people will put forward recommendations on how to solve this issue in this thread. But this is probably the most difficult problem to solve. The best 'solution' may be to just leave it as is. 5: DEMO items: Demo clutter is not helping things. DEMO items should not be searchable. When listing a product you should be able to hit a "radio button" that sets the product as a DEMO, and excludes it from search listings. Upon implementation, retroactively any item with the word DEMO in the title should be dumped into that catagory ( pending shop owner approval of course ). In addition, currently only being able to link one item to a given "DEMO" listing is a tad bit crippling. Multiple listings should be able to be tied to a single "DEMO" item for organizations sake. ------------------------------------------------------------------------------------- That's all I can think of at the moment. A lot of this is paraphrased from other parts of this thread~ as well as feedback from people inworld etc etc. Things the MP currently does RIGHT. Listing things by sales popularity is key to the success of any online marketplace. It HAS to be. I've noticed what I percieved as a "store popularity" statistic as well ( sorry if that's some sort of secret part of the algorithm ) where new items from a popular store are more likely to fly up in the visibility rankings and stay visible based on a store's overall weekly / monthly sales than new items from an unheard of store that doesn't sell much of anything. While this sort of mechanic makes it difficult for new stores to gain a customer base #3 in the above essay may help alleviate that minorly. I believe having this mechanic is important though as top-end content creators are likely to continue being top end content creators. Thank you LL for taking the time to consider all of this. I think fixing your marketplace code will help both Second Life and Sansar to be stable profitable enterprises for content creators and Linden Labs alike. I seriously would not be where I am today without SL as a platform to create and market on.

Okay MP beta is obviously not useable but before you start putting up pitchforks at LL employees calm down and wait for their response. It's the weekend~ speaking of people who don't want to have their lives messed with while off the job~ LL staff is no different. Also ~ It's worth pointing out that when LL does something right no one even notices. Interest lists, Object occlusion, Server stability. GroupChat fixes. Experiences. Scripted Materials control. All of this stuff has been implemented recently and all of it has had a positive overall effect on SL as a whole. Sure it doesn't all work flawlessly~ yes there are still issues, but if take a step back and look at it from a programming aspect:: SL is 12 years of spagetti code written by 100's of people ~ a substantial portion of them are not actually employed by the lab. And it's all knotted together. They somehow make it RUN. Not flawlessly mind you~ but it RUNS. That's a minor miracle in itself. When Ebbe took the lead they looked at the SL project and considered revamping it from the ground up. It was then that they determined that the inherent structure of the SL platform prevented it from being expanded into a platform that will be relevant from 2016 to 2020+ Hence Sansar was born. But they're still making an effort to keep SL going and improving. So. It's our job as users and content designers to give them feedback on the improvements they're considering implementing. Things like this one obviously need to go back to square #1 , be outsourced ~ or be scrapped altogether. But seriously put the pitchforks away please. That's not helping anything. I've urged people to try out MP beta so that LL will have a nice sample of responses from a wide range of people. From a business standpoint that's the most impactful thing you can do. That being said~ I'm still gravely concerned that this will be implemented ~ as I've voiced prior. But starting a flame riot on the forums is just going to lose the meaningful feedback in a soup of anger.

You won't be shot for suggesting that ~ In fact I posed this exact question into the question pool for Ebbe yesterday. Come November 19th we'll see if he has time to answer it.

Followup question to ChinRey's: Has LL considered purchasing a 3rd party solution to help optimize the SL marketplace experience? Google & Amazon both offer extensive services that do search and ranking exceptionally well. Is it possible that in the long run it would be cheaper to simply adopt a 3rd party search and raise the 5% cut LL takes off MP revenues to pay for it?

I'm just going to assume this behavior is unintended behavior. As such I submitted a bug JIRA on the topic. We'll see how it goes. https://jira.secondlife.com/browse/BUG-10688

Is Relevance functioning correctly or even at all?? The current MP search ~ when I look for an item catagory, MP pulls up what I know are best selling items in that catagory. I see popular merchants known for unique work that I happen to know sell incredibly well. ( They're my compeitition and rightfully so! ) My items are right there with them. On BETA search all the most popular items by the current metric are DEAD LAST. Bottom of the last page. The front six pages? All varying colors of some derivative spammed object that I know for a fact can't possibly be outselling the actual goods that had time invested into them. This is way way worse than the current system. If this is indicitive of how the system will perform this is catastrophic. Good merchants that provide unique content will have their sales fall off a cliff if this goes live. I'm genuinely concerned. Just as a stark and obvious example : Trompe Loille is nowhere to be found in furniture listings? What is going on here?

omg! youre a lifesaver! TY TY!

Anyone able to help me? Tell me if this is happening to them too? I tried on both firefox and google chrome. Agni mesh upload questionaire page loads fine ~ just not the Aditi one. I'm sad = ( I have meshes I want to test and I can't. anyone help ? thx.