In web analytics, there is often debate about how “Unique User” metrics overcount the true number of unique users. People delete cookies, remove flash beacon trackers, and visit websites from multiple devices. It’s an interesting debate, but not that meaningful to online gaming companies.
This post is part of a series provided by Global Decision’s Online Gaming Analytics practice.
In online gaming, we have the beneficial prerequisite that you have to login to start a session. For that reason, we can ignore the debate from http-ville and focus on the confusion that the words “unique users” creates. For starters, there is no such thing as a generic unique user. Because users can play your game repeatedly, good game / MMORPG analytics always specifiy a duration of time with the words “Unique Users.” Monthly Unique Users seems to be well-liked, although the more immediate results from counting Weekly Unique Users or even Daily Unique Users can add value more quickly.
Once the time-dimension is included in the specification of unique users, we can address the fact that unique user counts are not additive over time. Consider the following simple example: in the table below there are four players, and an “X” indicates a week in which a player logged in to the game one or more times.
You can see that the number of unique users each week varies from 1 to 3. Our core data (the “X”) has weekly granularity, so our column sums will produce “Weekly Unique Users.” Let’s suppose this data is for February 2011. It’s tempting to say we had 8 Unique Users in Feb 2011 (adding Weekly UU row). However, the example shows the pitfall of this approach: many users login multiple times in the month. The true number of Monthly Unique Users is 4. Using weekly data will result in a count that’s 300% to 0% too high.
If you happen to have the situation where each user only logs in once per month, then your Weekly Unique User count will equal your Monthly Unique User count. Since this situation represents poor loyalty and probably poor engagement, it’s certainly not a desired state of affairs. In the other extreme, where all users login every week, using the weekly data to generate a Monthly Unique User count will overstate the actual total by 300%, assuming a 4-week month alignment.
Key takeaways: First, never utter the words “unique users” without knowing the granularity used to generate the number. Second, use separate data queries to generate “unique users” for different granularity; they are not additive.