Discussions of the artistic success or failure of different expansions/phases of the history of the World of Warcraft usually focus on subscriber counts. This makes a certain kind of sense. In order for subscribers to leave, they probably need some source of dissatisfaction. Thus, major declines in subscriber numbers clearly point to something that people dislike.
This chart shows the subscription numbers up to the eve of Draenor. The story is a familiar one: strong growth through the vanilla and Burning Crusade periods, followed by slower but still positive growth during Wrath of the Lich King. Then with the release of Cataclysm at the end of 2010, the game begins a long and steady decline.
I think this picture is both incomplete and misleading in important ways. After all, this plot does not represent a closed population of players. Instead, it deals with a dynamic, open population with people leaving and entering at various rates throughout. This is a huge interpretive problem because people do not enter the World of Warcraft for the kinds of reasons that drive departure. People enter because of marketing or positive word of mouth, while they leave because of burn-out, slow development, or changes in content or mechanics that they dislike. Thus this graph conflates the success of marketing with the degree of engagement produced by the actual MMO.
No data source will be immune from this problem, but given the issues it is useful to look at the health of the World of Warcraft in more than one way. This chart uses Google Trends to show the degree of online interest in the World of Warcraft over time. While subscriptions perhaps measure the breadth of interest --- how many people are interested enough to at least not cancel their subs? --- these results combine breadth and depth. How many people are reading a lot about the World of Warcraft, not just logging in for raid nights?
The two charts tell different stories. In terms of subscriptions, the peak of the World of Warcraft was at the release of Cataclysm. In terms of interest, however, the peak moment for Azeroth was the release of the Burning Crusade. On this metric, the Vanilla period was a phase of rapid, almost vertiginous growth, but (setting aside the launch peak) the Burning Crusade years were also a period of (slower but still positive) growth. Three months before the launch of the Burning Crusade, interest in WoW was 80% of its peak. Two months later, interest was at the same level. By the end of the Burning Crusade, three months before the launch of Wrath of the Lich King, interest was at 85% of its peak. Overall engagement with the World of Warcraft was thus stable or even growing slightly through the first expansion.
Wrath of the Lich King pushed interest from that 85% level three months before launch to 91% during launch month. Two months later, as with the Burning Crusade, the bump had dissipated, and interest was back to 85%. By the end of Wrath, however, interest had fallen radically to 43% three months before the launch of Cataclysm. The Wrath period thus burned off roughly half of the overall level of engagement with the World of Warcraft --- even though it was a period of subscription stability.
Cataclysm's launch saw a huge surge of interest, from 43% three months before to 63% at launch. Unfortunately, two months after launch the level of interest had fallen to 39%, and by three months before the end of the expansion interest was at 20% of peak. Cataclysm thus destroyed more than half of the reduced level of engagement produced by Wrath.
Mists of Pandaria saw interest jump from 20% three months before to 25% at launch. While this is a small move in the overall scale of the time series, in relative terms it is not a bad launch result. Interest increased by a quarter of its prelaunch baseline, a larger relative gain than Wrath had and basically the same as the relative gain for the Burning Crusade (although smaller than the relative gain at the launch of Cataclysm). By two months after launch, interest had fallen to 22% --- lower than the launch peak but for the first time in the history of WoW higher than the interest level three months before launch. By the end of the expansion pack, interest had fallen to 18%.
Thus, Pandaria burned out 10% of the overall engagement that it inherited. This is clearly a worse performance than the radical growth curve of Vanilla or the more gentle positive growth of the Burning Crusade, but it is by far the best performance since.
Finally, Warlords of Draenor saw a large launch-window burst of interest in WoW, jumping from 18% of peak three months before launch to 26% at launch. This is almost as large a surge in relative terms as the launch bump for Cataclysm. Two months after launch, the (as yet not totally complete) data suggest an interest level at 19% of peak. While this 73% retention of launch-window interest is better in relative terms than the dramatic drop in interest after the launch of Cataclysm (62% retention), it is a notably worse short-term performance than any other expansion.
How do these data revise our understanding of the history of the World of Warcraft? First of all, they reinforce the huge success of the early years. Second, they provide an important clue that the Wrath of the Lich King period was not the strong success that subscriber numbers suggest; instead, it was a period that dramatically reduced overall player interest in the World of Warcraft and set the stage for the game's long-term decline in subscribers. Cataclysm remains the period of biggest failure in the history of the game, although these results suggest that it should share the blame for the overall decline with Wrath. Mists of Pandaria, by contrast, appears as a reasonably successful expansion, at least in comparison with the other late expansions of the game --- an evaluation that we can usefully triangulate with the relatively stable subscriber numbers through 2013. Warlords of Draenor clearly brought a lot of attention back to the game, relative to recent years, for a brief moment around launch but is showing no early signs of a major, transformative impact on overall levels of player engagement.