ProkofkyNeva: Metrics and Understanding How Agents, Malware, & Shadows Effect Them
Posted in 3D Social Networks, Attention, Memetic Engineering, Second Life, Security, Web 2.0 by wayne.porter on November 1st, 2007 Prokofy gives me his take on CSI:NY Episode regarding metrics. Now I will give mine.
Wayne Porter has some SL CSI:NY metrics, but they are merely taken from the Lindens’ own pages, and don’t look like anything new, not even the “baby avatar” increase –
Correct. Exactly. DId I allude otherwise?
and I don’t know where he even gets 50,000 concurrent as I was watching assiduously, refreshing and keeping a record, and I actually saw it go to 41,000 then down to 38,000 over 3 hours. To be sure, a few days later it crept up to 56,000. Sales of Lindens are brisker, and went up to a whopping L$70 million today, but that’s to be attributed to Halloween.
What was the rate of your refresh? Was it consistent? Did you take any breaks?
The real numbers to look at for CBS and the Sheep are their actual sales sites, shop.onrez.com and the special CBS site
— which I can’t even find in Google right now.
What is your background- out of curiosity? There are a wide range of metrics to look at and if cookies were employed in tracking far more accurate ones. Sales on OnRez? If you want to go deep into this I am happy to do so. While that is an interesting metric, the total dollars are not huge. Far better types of data in a properly configured system would be more valuable to CBS.
I think it was clear the metrics I quote and gave counter suggestions as to what they might mean are some (quick) betters ways to measure a number of traditional 2web metrics. Some could include network class, location, medium, bounce, TOS, AVPPV, LTV, AVB, Basket Abandoment, etc, etc, .I was in-world during the show and your numbers hit around what I saw when I bothered to glance up. I quoted their metrics directly from the blog at the time…and offered what I felt would be more useful more metrics. For example:
What I said and I shall clarify:
- Total sign ups continues to climb but they are like unqualified leads. Throw it out the window unless you want to develop an attrition factor or “sleep” factor.
Summary- so what? Poor GUI, and poor indoctrination of AVs (avatars) into the complex physics and even vertigo often experienced and I can probably run an attrition rate on those numbers and the longer term “conversion” (how do we want to define it? And would be lack luster… )
- More interesting would be Change in Concurrent users per minute or login delta. Especially Login Delta of Accounts created on the Day of exposure. Login Deltas of New Avatars and Login Frequency are really two critical health metrics that must be watched.
This would be interesting just to judge the media’s impact and curiosity at large as well as grid health.
- A longer term metric of great value would be conversion of Free to Premium and also factoring in cookie duration and attrition through deletion and chace attrition to normalize. (I believe negative in September for awhile.). Frankly this is where LL really needs to work or some leadership. A 512k lump of “server space” and a tiny stipend of L$ really is NOT a compelling factor for premium pay. Start by doubling groups! Hint.
I stand by this. The “right” to hold a virtual 512m plot really isn’t that great of deal and the tiny stipend is laughable. Start with advanced communication options, perks, double the groups, Up the IM cap. I mean get real…communication is very inefficient as I am sure you are aware.
- 51k New signups is like a small baby avatar boom, although in early April and a couple of other dates post there were similiar spikes, although not as high numbers.
In the end they had a nice rush but we I think we should calculate attrition q month and give the prolonged burst q 3 months. So let’s hold our breaths. Even so prime time TV had the largest spike, there were other similar spikes prior this year without the peaks and valleys for a prolonged period.
I did not set around refreshing an agent over and over. I do not think that would even give you an accurate number. How did you do this? What was the time? Was it fairly accurate?
- Peak Concurrency of 50k. Actually nothing incredible here. I believe we saw numbers like this in September.
That was per their blog. It has been hit before, and who knows it may have hit 50k for a nanosecond.
- Grid Stability Index of 1.22. Fine. GSI, I believe, is based on open-ended scale, starting at zero (stable) and increasing as indicators show variance from a stable situation this includes both planned and unplanned outages, service interruptions, login failures, and other grid-wide performance issues.The Grid held up.
As for metrics. I have been developing my own. Some new ones that are quote interesting and telling of a mixed story. When I release them, or how- I don’t know yet.
ALEXA ANALYSIS
Let’s get some other analysis corrected.
Even using Alexa numbers to judge is ridiculous, sits use it evaluate eCPC and eCPM due to lack of better tools. I have proven, as well as other e-commerce and security experts that Alexa can be gamed. I have seen Chinese malware inflate (try to inflated they claim) the numbers. In reality I know, at least at one at one point you can you use a serious of 4-5 pcs and a couple of auto-refresh programs set to certain time and inflate the numbers easily. The same with laundered onBlur traffic, hidden iFrames, or Traffic Slams or other methods. The numbers have improved, but they are a crude thumbnail and with such a small differential (the numbers don’t bake well until you get under 20k or lower) it makes your conclusion premature.
IN GOOGLE
I have to ask which inde did you you check against and with what agent and did you use a VMware environment or were their any external changes like .hosts tampering? In terms of the now morphine “SEO” industry one sjhould understand that Google has a thousands of distributed servers (I have plumbed them using various methods- sorry no need to say hi Google) and that not only are the results and databases radically difficult in both SERP ranking, but SERP composition- e.g. media format. They are effected by the fact you are logged into a Google Service or not, Your geographic location. The time of day, random tests and other factors. While there seems to be a “cluster” of searches of SERPS on a many servers they tend to show pages or media chunks with a similar mode, still this is not always the case. Further testing under disparate conditions would be needed to make a blanket statement and test away- we would get different results based on a number of variables up to and including where,who, how, agent, time, and other variable we could never ascertain. Sometimes very similar, sometimes radically different.
This will continue as Google launches their “dark fiber and server” infrastructure play against facebook and continues to use algorithms that are guided by individual behavior and other variables and people bite into search history and that AI.
The same for Yahoo! which is launching Search Intent.
Skewing SERPS from scholarly to commercial via an ajax slider.
SLEX and STABLE
Lastly, ” Slexchange.com shows a pretty stable traffic for that range in the last month.”
Unless you audited their server logs, and it would depends on the technology they use, this is again a “Support the Troops” kind of statement. There are simply far too many ways for sophisticated games- both SEOs specialsts, botnets, rogue agents, malware, and I have even seen virtual botnets created through the use of the triple-obcuscated Action Script injected into mainstream banners that are triggered to do their work when the IP filtering attacks they use allow them to dance around US based QA teams. It happened to MSFT x2 and they had a stellar record. They have hit You Tube and they will hit anyone. This happens ALL THE TIME.
These numbers are hazy reflections, albeit better then what LL provides now for traffic (and what I am trying to solve). Their numbers are probably meant to try to wrap engagement, curiosity, etc by factoring in time, maybe activity, etc. They do a poor job of it, perhaps, like you, it is not a core competency. I have some beta success with that, and that is one place Second Life could shine if they understood these emergent metrics. Heck- properly setup systems that could take the strain would allow us to model WOM.
I find the 2L economy all together “backwards”, but that is to be expected I guess- new. However, in my eyes so many solutions are very, very clear and not that difficult to fix. I share your frustration on that part- like sand-boxing no mod scripts for security hygiene. But I don’t play politics I try to solve problems.
Thanks for posting your thoughts and the reference. What solutions do you propose?
3D social networking attention Memetic Engineering Second Life Security web2.0Popularity: 6% [?]
Second Life News for November 1, 2007…
Identity Hunting In Second Life Lingerie is easy - but where do you find anthro-style purple cat ears? Like so many people, I come to SL to be myself, but more myself than my RL body will allow. Problem with that is for me that requires more than jus…
I’ve already answered this on my blog, but I’ll answer it again here in more detail.
If all you’ve done is take LL’s statistics which are widely reported as broken when you use the aggregate from the daily/monthly stats on the front page (NOT the green dots), then obviously you’ll be called on that. That’s all. It’s not the end of the world, however, because, as I’ve explained and explained, like the flawed and widely-knocked Alexa, even if it is fallacious and full of problems, comparing two sites measured with that even flawed measuring stick still tells you something — and watching the Linden front page figures do tell you something — but not everything — over time.
You are so hidebound by all your “knowledge” — formalistic and ritualistic recitations of received memes — that you can’t see the forest for the trees here. It’s pretty simple. the CSI:NY caper didn’t do so well. There were no million as predicted entering SL — not even close to anything like that. We saw no more than about 4,000 — and more like 1,000 — concurrent log ons at these sims. Today it’s more like 100. The Electric Sheep reduced the number of sims from 420 to 28 this week, because they had no takers — for the same reasons many people bounce off SL, it’s too hard, the email to confirm never arrives and is caught in spam filters, the graphics card isn’t good enough, etc. etc.
The idea that CBS wouldn’t be interested in the sales figures of shopping.onrez.com, the Sheep’s site for avatar clothing, vehicles, houses, etc. is just plain short-sighted. Of course they’re interested. The Sheep could get away indulging themselves in adding on special buttons to deliver the clients CBS bought for itself with this tie-in because CBS understands that the only way to make cash on these projects is to have stuff for sale for avatars (ROI is a more complex subject). It’s a slow way to make a lot of money, that’s for sure, it’s micropayments, but it is part of the experimentation they must do to get into the Metaverse and follow their customers who have Tivo’d their ads and fled even from TV itself into virtual worlds and social media spaces.
50,000 is what they had at the point before the TV tie-in, and it simply didn’t go up significantly. I watched as it went DOWN on the day of show. Yes, I refreshed the page every 15 minutes for 3 hours. No, I didn’t take breaks. Yes, I can read a number and write it down and watch it go from 38,000 up to 41,000 and down again — and also watch the amount spent inworld not change drastically, either.
An agent wouldn’t have to be “refreshed”; only a web page was being refreshed. As for avatars — the streaming world does that for you. You go to the island, you use the mouse, which normally has a hover text show up of the number of agents on the sim. Yes, as we’ve told you a million times, this measurement of green dots is completely reliable, and always has been, and the very remote possibility that it wouldn’t be, let’s say through packet loss, is so problematic that if you suffered from that, you wouldn’t even be there counting anything anyway.
The mouse-hover-text function was broken a few days before/during the CSI show, and I wondered — by accident? It was described on the company blog. Some people got it working again. Experience in SL can be very, very different depending on all kinds of factors, location of log-in, ISP, type of graphic card and computer, location of the server (SF or TX) etc etc. For example, some people will swear up and down that the upgrade to premium function is broken, wiether from the website or inworld. Others swear it isn’t broken and prove it. A Linden keeps testing it and yes, finds it is buggy, and yes, some people experience it. That’s what happens in a 3-D, real-time, interactive world that is simply experiencing a different kind of load and dynamics, with constantly dyamically changing content, far different than a web page, which is what you are used to measuring.
That truth simply has to sink in to you somehow — and it isn’t, so let’s try again: you are not measuring a web page; you are measuring *a world*. Yes, *a world*.
Fussing about bots, malware, Alexa problems, blah blah blah just belies the obvious truth: Slexchange.com has been a more stable site, has been broken less, and with better traffic as watched over time — as many have watched it. shopping.onrez.com is a newer site that grew out of slboutique.com and has had more problems. The OnRez site also developed download problems on the day of show.
Of course you play politics, and you place yourself squarely in the middle as some kind of ostensible power-monger.
No sale *shrugs*.
Um, I don’t need to post “solutions” when I provide recipes for fixing all kinds of social and technical problems on my site constantly. A complaint about the sheer nonsense and hype floated around the CSI:NY show merely needs to be debunked so that people are more sanguine about dragging millions to SL to prop up the flagging numbers. This isn’t ’support the troops’ but rather an expose of how there are fewer troops than everyone imagines. It’s merely about reporting the truth of a story riddled with hype for corporate reasons — the solution is that the corporations rein in their hype, there isn’t something that somehow has to be done with metrics or servers.
In fact, it’s hugely political, as what the Sheep have been doing along with other companies serving large corporations has been to knock the inworld “traffic” metric which you don’t even reference and didn’t even bother to track. Traffic measures number of avatar visits — but it’s a formula that is tied to total number of places visited and total log-ins that day, it used to be used to give out incentive payouts, but that was gamed and removed.
It’s still a very valid metric once you go past the first few gamed slots, just like Google constantly fetches up Wikipedia or AOL shopping sites these days, so you have to push down a few to see what is really worth it.
Traffic on these sims show 4,000, 5,000, 14,000 at the most. To give a contrast, imagine my rentals or my sponsored infohub, without any TV tie-in or corporations or anything of the sort — they get 1500 or 2000 or 5000. So that’s the problem — there just isn’t the traffic everyone imagined.
And it’s that very poor showing in terms of traffic, a valid inworld metric and one understood by everyone, that is making the corporations pressure the Lindens to remove it. And remove it they will, they are already making some other socially-engineered metric that will be the returns in a new search function they are building.
Prok,
Thanks for the reply. Let me digest this all and reply.
best,
Wayne