Linden Lab's Tom Hale clarifies Q1 results

You may recall that when we looked at the published Q1 metrics for Second Life, we were struck by the fuzziness of some of the material. It was clear that there was growth, but many of the figures lacked the sorts of definitions that would give them any apprehensible meaning. Graphs that show upward trends are nice -- well, usually, unless they're recording bad things.

Thankfully after our grumbling along that line, Tom Hale, Linden Lab's Chief Product Officer took note of the issues we and others had with the reporting and set out to add some necessary clarifications, quite a pleasant departure for this historically secretive company.

Downtime

Hours lost to downtime was one of the first stumbling blocks - since we didn't know what was being measured, only that it wasn't actually, you know, downtime. At least not in the conventionally accepted sense.

We have an automated process that looks at concurrency measurements every minute and compares current concurrency to what we saw during the same time over the last few weeks.

When this system detects an extended period of time during which our concurrency is significantly below our recent history, it generates an outage event and estimates the number of usage hours impacted by the outage (the outage impact is calculated by subtracting actual concurrency from the concurrency we had at the same time one week earlier).

Our System Infrastructure team reviews the system generated outage events to confirm the impact numbers and to categorize outages into "planned" and "unplanned" events. The calculation to derive "user hours lost" is length of outage multiplied by the delta between concurrency in the previous week and the actual concurrency during the outage.

So that's how we measure total user hours lost to downtime.

We wouldn't have guessed that they used this system. Coincidentally, it is identical in most major respects to the system we've used for the last few years to measure Second Life grid stability, however we made a few tweaks, because the system described above tended to miss certain regular problems caused by certain batch processing jobs.

If there's an increase in users getting online in the virtual environment though, that could well mask a partial subsystem failure from the Lab's reporting.

Repeat logins

Repeat logins are the number of returning Residents in a given month that are not new to Second Life in that month. To calculate this, we take the number of unique users that login at least twice, and subtract the number of new Residents who logged in for the first time during that month. Growth in this number indicates that our base of existing, repeat users is growing.

That makes quite a bit of sense. Primarily our confusion here seems to have been that two charts were published that appeared to hold the same data. Monthly repeat logins, and monthly unique repeat logins. Our conclusion ultimately is that the chart published on GigaOm was mislabeled and represents monthly unique repeat logins.

User-to-user transactions

For the last couple of years, the Lab's counseled us to ignore user-to-user transactions as a measure of growth or as an economic indicator of any worth. Part of that reason is that the figure is pretty trivial for any user to inflate (it's possible to make it jump a few million or billion Linden Dollars in short order - though that would likely trigger the Lab's ire if they noticed. And has -- this has happened before), and that an increasingly large percentage of the figure seems to represent 'idle' money, that is, money that is going from place to place, but isn't a committed part of a commercial transaction.

Eg: You sell me something for five dollars. I give you a ten, you give me five dollars in change. In user-to-user transaction terms, that's 15 dollars. In actual economic power, it's 5 dollars. There are fewer incidents in Second Life where change or refunds for inadequate payment are given, but they do occur. Likewise money is transferred between holding accounts (such as Xstreet SL), and the transfer each-way is added to the user-to-user transactions total, even if the money did no actual work.

Because of the Lab's, well ... frankly successful attempts over the years to convince us that there were good reasons to avoid any discussion of the user-to-user transactions metric as a valid statistic, and because the nature of the statistic seems not to have changed over time, we'll have to keep this figure filed in the junk drawer, where Linden Lab argued that it belonged.

Even so, without that figure, the remaining definitions bring a far more assuredly rosy picture to Second Life's Q1 growth. Growth appears to have been continuous for some time (if a little erratic), but the rates shown in Q1 are really quite impressive.

	Are you a part of the most widely-known collaborative virtual environment or keeping a close eye on it? Massively's Second Life coverage keeps you in the loop.