Every so often I see someone using statistics to back up a viewpoint they have. It is very common among Warmistas and Deniers. We also see politicians misusing statistics. The most common misuse of stats is to select the data segment and use it to PREDICT the future. If that worked, everyone would get rich playing the stock market.
So, recent articles on stats got me wondering what was up. I used to write an article each month on SL statistics. Unless stats are showing change, they get pretty boring. It’s been 16 weeks since I wrote about Second Life Statistics (Sept. 2011).
Rod Humble, the Linden Lab CEO, seems to have been spinning a bit when he said, “…and over the holiday period we had over 20,000 new people sign up a day.” Well, take a look at the stats. I get the images from Tateru’s site, Dwell On It. I composite them in Photoshop to provide a longer period. You can click the image below to see a larger version.
It looks like for the month of December the average could be over 20,000 per day. So, the spin, like all good spin, has truth in it. But, since then the sign ups have been in a steady decline.
Depending on what period one wants to look at things have been changing. Or they have been in a gradual up trend if we use all the data I have. That may change soon, if the recent trend continues.
There is also a down trend in signups at this time last year. It seems to start in late December or early January. We may be seeing that again this year. But, there are not enough stats to say this is a seasonal occurrence.
There are probably other factors affecting the rate of signups. We know that when they changed the signup page things improved. We also know via way of Viale Linden that more effort was going to be placed into the signup process to get people to Adult Verify. I amy dig through the meeting minutes to see if I can find when that was.
Web designers know that every field and button a person has to click on a forum decreases the likelihood that the form will ever be submitted. That is one reason the 1-Click purchase became a prime goal of design for commercial sites.
I don’t keep track of the signup page… been there done that a few times. But, not recently. So, whether or not they have broken the magic of the sign up page or whether something else has changed… I don’t know.
The Lab’s advertising for a time has sucked. Several have blogged about the poor ads the Lab has been using. Recently those have changed. It’s too soon for the change to show in the stats, if it does. My point is things are changing and we don’t know all that is happening. So, saying the stats prove anything is probably more wishful thinking than fact.
For many of us this is the make-break stat. I started keeping a trivial bit of data on concurrency. I have a little script in my house that shows the current number of logged in users. It keeps a max and minimum value too. I reset it once a month. It gives me an idea of the flow of people in and out of SL. Every 10 minutes it updates.
I’ve learned sign in and outs run anywhere from 50 to 1,000± per ten minute frame. The peak concurrency is right around 2 PM SLT and the minimum around 2 AM SLT. The minimum is typically 31,000 and the peak around 69,000+. Last month when a bank of servers went down at a co-location facility 17,000± people were knocked off and concurrency dropped to 21,000±.
For the data I have one can easily say there is still a down trend, the yellow line. The green line shows alternate changes in the trend. It depends on how one wants to look at the data.
While I question whether we have a seasonal factor in the data, I doubt there is one. Tateru has done more work on deciding whether there is a seasonal factor. There is an article about it on her site. Basically, there are no seasonal factors in concurrency.
We know the Lab is doing research and working on retaining users. Rod said, “We managed to grow the new users significantly: they bumped up by well in excess of 40 per cent…” But, 40% of what? If it is 40% of signups, that is huge. I’m not seeing the world flooded with newbies. So, it unlikely it is 40% of signups.
If the 2 or 3% of new signups staying in Second Life after signup increased 40%, I can probably believe that. They would have bumped the number from 2 or 3 to 3 or 4.
Empirically I do meet more noobies now. But, that doesn’t mean much.
Rod has told us that since the release of mesh login habits started changing. I know my habits have changed. I spend more time out of Second Life learning modeling and texturing. I also blog more. I am sure others working with mesh are spending more time learning to build with the new tools.
We have more new tools coming; Pathfinding and Experience Tools are in open and closed Beta respectively. We are told more new things are coming. So, it is reasonable to assume Second Life would be used differently and that a different use might have a different use patterns. For my own use I can make a case that I spend more hours on Second Life related tasks but less time in SL.
Summing It Up
Don’t accept stats as proof of much of anything. They tell us what has happened in the past. They tell us nothing about what today’s changes are doing.
When we do not know all that is happening we can’t know what is changing the direction of a trend line. And much of how the data is interpreted depends on who draws the trend lines.