We have a Goon, boviscopophobic
, who has done an amazing job these past months running analytics on Star Citizen using data mined from the game's public facing data (the same place where the funding numbers
come from). If you see a chart in any of my blogs or missives (e.g. my latest
), chances are, he did it. Below is his latest masterpiece.
JAN 2017 RSI DEMOGRAPHICS UPDATE
About 6 months and $25M have elapsed since the previous demographic snapshot of the RSI forum population. Updated funding-related statistics are summarized below. The methodology is mostly the same as in my previous post, so you can refer to that for details as well as an explanation of the meaning of each graph.
First off, the forum account vs RSI account rate discrepancy has stabilized at about 3.5, meaning that 3.5 RSI accounts are currently being created for every forum account. This ratio obviously spikes during free fly events.
For the funding-related graphs, the basic funding assumptions remain the same as last time, but there are three updates to the methodology. The first relates to closed accounts. Thanks to the wave of refunds post-Streetroller, I learned that is possible to determine with reasonable confidence if an RSI account has been closed, which generally indicates revocation due to refunds or possibly other misbehavior such as hacking. This latest set of summary graphs includes only accounts that were "alive" at the time of the snapshot. The previous set of graphs included a certain number of "dead" accounts, which affected the accuracy of the title counts.
The second change is that I've aggregated all titles not associated with a funding level into an "OTHER" title, except for a small set of user titles that I deemed to be CIG-related. These titles, namely "Staff", "Developer", "Creator", "QA", and "Game Master", are assigned the aggregate title of "CIG". Some other user titles that are arguably CIG-related, but which I did NOT include in the CIG set, are "Bug Moderator" and "Moderator". Note that some developer accounts may mark themselves as such as such only by their account name; these would not be included in the CIG count.
Finally, to counteract title churn from people changing their title, I look at each account's titles over a number of forum snapshots and use the one that implies the highest funding level. Since user titles can be "understated" but not "overstated", so to speak, this should be a reasonable procedure if user funding levels are non-decreasing. Thankfully, since CIG almost never grants partial refunds, an assumption of non-decreasing funding levels should not introduce too much additional error. (Note: if no funding-related titles are available, then CIG-related titles are prioritized over "OTHER" titles.)
The contribution of concierge-level backers (High Admiral and up) has slightly increased -- 56.4% under the mid funding scenario, versus 53.6% last time. If we compare the graph of estimated average user spending by quarter of enlistment with the previous version, we get the following average per-user increases:
This indicates that accounts of all "ages" are continuing to put money into Star Citizen, possibly more so for pre-2014 accounts. However, note that an increase of $10-15 or so is a small fraction of the likely average transaction amount -- recall that this period includes Gamescom, Citizencon, the anniversary livestream, and the holiday sale, which featured pricy concept ships, cash-only sales, capital ship sales, etc. Depending on what you think the average transaction amount is (which I have not attempted to estimate), you could translate this into an estimate of the size of the current paying backer population.
Another longitudinal view of the backer population can be obtained by constructing contingency tables at various time snapshots. For example, the following is a comparison of highest user titles achieved through early August 2016, versus early January 2017. Because of how highest titles are computed, this table contains some unknown fraction of users "leveling up" through spending, and some users simply adjusting previously understated titles upward. I believe that the dominant contribution is leveling up, especially when looking at movement between the higher tiers, but I have not attempted to quantify this.
Note that we have two new pseudo-titles: "DEAD", indicating that the account died off (refunds etc.) between Aug 2016 and Jan 2017, and "UNBORN", indicating that the account was made between Aug 2016 and Jan 2017. So for instance, we can see that of 92 completionists as of Aug 2016, 1 of those accounts got a refund. Of 193 wing commanders as of Aug 2016, 31 were promoted to completionist and 2 got a refund, etc. Notably, 22 CIG accounts "got a refund", which most likely means they left the company.
As a rough measure of the propensity of backer subpopulations to level up, we can construct a matrix of outflow percentages. In this table, the number in a particular row/column indicates the percentage of the population with that row's title that advanced to get the corresponding column's title. So for instance, 16.06% of all Wing Commanders in August became Completionists by January. Similarly, 0.37% of Civilians became Freelancers/Colonels, etc. The hottest cells consist of concierge backers (High Admiral and up) moving up one or two levels, and CIG accounts moving to the exits.
If we are interested in inferring refunds specifically, then we need to look at pairs of snapshots that are closer together in time. Otherwise we can miss salient developments -- for instance, if a Civilian in August became a Wing Commander in November then got a refund in December, it would only show up as a Civilian refund in the above table. Using a set of several snapshots I derived the following counts for account deaths per highest title. I also noticed a large number of newly established Civilian accounts showing up as dead. To exclude possible low-effort banhammered trolls from the refund counts, I only counted Civilian accounts if they were confirmed as being alive for at least 45 days in at least one historical snapshot.
- Completionist: 5
- Wing Commander: 4
- Space Marshal/Lieutenant Commander: 19
- Grand Admiral: 27
- High Admiral: 70
- Vice Admiral: 47
- Rear Admiral: 58
- Freelancer/Colonel: 116
- Bounty Hunter: 67
- Mercenary: 61
- Scout: 32
- Civilian: 448
- CIG: 26
- OTHER: 74
Since this is a small and very much non-random sample, the likely accuracy of the funding scenario assumptions (already not that good) is probably far worse for refunded accounts. On the one hand, Civilians are assumed to have a low average contribution partly due to the proliferation of free accounts; however, a refunded account would obviously not be a free account. On the other, high-value accounts may not be refunded for anywhere near their nominal value, due to grey market transactions.
If we go ahead and apply the min/mid/max funding assumptions anyway, we get refund totals of $407,420, $674,587.50, and $941,755, respectively. For another estimate, also problematic, we can consider the self-reported refund amounts from /r/starcitizen_refunds. From reading through the posts that stated actual refund amounts, I arrived at an average per-user refund of $1366.10. Applying this to the 1028 non-CIG refundees, we would get a total of $1,404,350.80. These estimates are of course only for the refunded forum population. The multiplier to get the total amount of refunds in the entire RSI population would likely be well less than 2.5, which is the ratio of all RSI accounts to all forum accounts.CONCLUSIONS
All previous caveats about the accuracy of these estimates still apply. In addition, there are particular problems with trying to estimate refund amounts. Nevertheless, I think we can conclude that the refund outflows, while CIG certainly would find them annoying, are probably small enough in total that they can be easily compensated for with an extra concept sale (if we don't account for increases in engineering debt).
There are indications that funding is leaning even more heavily on concierge-level backers; this might be a good topic for follow-up analyses. Account age does not appear to play a large role in incremental spending.
Previously I speculated about a soft per-user average spending ceiling around $200. This now seems to be more of an artifact of the bounded time window the backer populations have had to spend their money in. As that time window lengthens, fresh spending continues apace and it remains to be seen when there will be a large-scale change in backer purchasing behavior.