GDC is in San Francisco this week, just next to Flurry’s headquarters. By the size of the crowds, we (very scientifically) estimate that attendance should easily surpass last year’s record of 22,500. Having tracked the growth of mobile games for several years, we weren’t surprised to see more than 30 sessions during the week focused on smartphone and tablet gaming.
Here’s the big picture, based on our estimates: There are now over 1 billion active smartphones and tablets using apps around the world every month. And of all the apps consumers use, games command more than 40% of all time spent. Looking at revenue, games also dominate. Today, for example, 22 of the top 25 grossing apps in the U.S. iTunes App Store apps are games. Gamers spend money, and game makers are in love.
In this installment of research, Flurry studies how age, gender and engagement vary across key game types. Understand this, and a game developer can design a more engaging game that appeals to the right audience. In short, they can build a better business. In this study, we included more than 200 of the most successful free iOS games, with a total audience of more than 465 million month active users. For a better comparison, we organized these games by their game type (aka game mechanic) instead of traditional, less granular genres. Let’s start by looking at how different kinds of games appeal to gamers by age and gender.
The chart above plots game types by age and gender. From left to right, we show what percent of the game audience is female, with the far right equaling 100% female. The opposite is true for males. For example, 0% female equals 100% male. From bottom to top, we show the average age of the game type’s user base, between 20 to 50 years old. Putting it together, games in the upper right quadrant are preferred by older females. Games in the lower right are preferred by younger females. Games in the lower left are preferred by younger men, and so on.
Young Men and Their Competition
Inspecting the chart, the tightest cluster appears in the lower left; specifically, game types such as Shooters, Racing, and Action RPG skew younger and more male. Card-battle and Strategy games also skew toward younger males. The only genre that skews toward males over 35 is Casino/Poker, with pure poker games skewing even more male than the game type as a whole. This appears to leave a big gap in the market for developers who can create games that appeal to middle-aged and older men.
While men tend to gravitate toward competitive games, women gravitate toward games that are less competitive and tend to be played in a more enduring way. These include Management/Simulation games where players can build out an environment, Social Turn-Based games in which they can play over time with friends, and Match3/Bubble Shooters and Brain/Quiz games, to which users can frequently return when they have a few spare minutes. Slots and Solitaire are both solo-play game types that skew toward females who are over 40, suggesting that they serve as long-term time-fillers.
From a marketing perspective, mobile game publishers can also leverage this knowledge to design targeted campaigns appropriate for the kind of audience to which a game appeals. Flurry’s ad network, AppCircle, allows publishers to target specific demographics for efficient spending.
From Courting to Betrothed
We find that mobile gamers tend to prefer playing a few kinds of games and demonstrate highly predictable play patterns. In other words, they form relationships with their games. Savvy publishers understand these dynamics and use them to inform acquisition strategy, gameplay design, and both in-app purchase and ad-based monetization tactics.
In the chart below, we map game types by usage and retention. On the y-axis we show the number of times per week consumers play different game types. On the x-axis we show how long different games retain their user (i.e., Flurry defines Rolling Retention as the percentage of users that return to the game 30 days after first use, or any day after that). To see how usage and gender work together, we’ve also colored-coded game types by whether they are more male, female or neutral in appeal. Let’s take a look.
The chart above reveals that different strategies should be employed for different kinds of games. It also shows, loosely, that women are more committed while men are more fickle. Sound familiar in life? Hmmm. For the gaming industry, the universal take away is that to optimize engagement, retention, and monetization, developers must tailor their mechanics and messaging to match their ideal target audience. Let’s take a tour of the quadrants.
“Players” try a lot of different games, play for only a short time and tend to be found in highly competitive games (e.g., “Player vs. Player style games). While fickle, they tend to have a high willingness-to-pay in order to progress faster in a game or increase their ability to compete at a high level versus other players. Attracting the right users through targeted acquisition can pay off, as those that stay will pay. Notably, Card-Battle games have very low retention but off-the-charts monetization, extracting enormous revenue from the small number of users that stick. One implication is that these games need to be highly polished at launch with updates ready to go, as gamers will discard games quickly and move on if the game fails to resonate. From a design standpoint, these games should offer immediate opportunities for users to advance by purchasing upgrades and boosts. For the users that might not spend, a lucrative option is to offer in-game currency for watching video ads.
“Going Steady” game types are found in the lower right quadrant of the chart. Usage is less frequent but retention is very high. While these gamers don’t play as often, they are loyal. This group of games tends to be easy-to-learn and easy-to-return-to even after a lapse in playing. They lend themselves to quick play while in a “wait state” (e.g., waiting in line, taking a bus or perhaps checking out of the meeting or class they’re in). Since these are not particularly immersive or competitive games, they are less likely suited to in-app purchase. However, they can generate significant ad impressions over time, and can be designed to show banner or interstitial ads without being overly disruptive to the experience. For games with larger audiences, publishers are utilizing mediation platforms that enable the use of multiple ad networks in one system, ensuring maximized fill and ad-revenue for each space.
“Committed” comprise of consumers who play games for the long-haul. As such, game makers should think of it like a marriage. Think about appoint mechanics like setting a date (hey, even married people need to keep it fresh). These games should be designed with deep content and not try to sell too hard to their users too quickly. From a monetization perspective, commitment-oriented games have great potential for in-app purchases since users of those games are likely to value such purchases and amortize them over long periods of gameplay. And while only the largest titles have achieved this to-date, this group of games are great candidates for in-app product placement. Additionally, with high impressions counts, it is worth publishers’ investment to implement monetization platforms that make ad spaces available to real-time bidded ad exchanges, ensuring they reach the brands and advertisers that value their audiences.
“Infatuated” consumers have fallen hard and fast for their games, but the candle that burns twice as bright burns half as long. They have crushes, and show binge behavior. During the “crush” window, the developer needs to work hard to extract as much revenue as possible. As such, developers must provide vast amounts of content to the users, consistently, in a short-window. Matching monetization to game type, the competitive nature of Strategy games, and Slots users’ incessant desire for in-game currency, make a solid in-app purchase strategy paramount. Sales, events, and purchase opportunities timed with key moments of emotional investment can drive significant profits for publishers.
There’s a Pebble on the Beach for Everyone
In gaming, there are a vast number of game types that attract distinct audiences. And these different consumer segments display very different usage patterns, which have direct implications on monetization strategies. As in life, where the richest relationships are borne from knowing oneself and his or her partner, game companies must also understand both. Only then can you get the most out of the relationship.
Just as a company might look to metrics such as their Net Promoter Score or individuals might look to their Klout Score to judge their social media influence, app developers want benchmarks to evaluate how their apps are doing relative to other apps.
To provide benchmarks, we studied apps by their retention and size of user base. We also compared these two dimensions to see how they relate to one another. For example, do apps with more users have stronger retention than those with fewer users due to network effects? Do apps with smaller audiences see higher retention because they focus more on the interests of a particular segment?
Apps By Number of Users
We started our investigation by identifying the apps that Flurry tracks that had at least 1,000 active users at the start of November 2012. That eliminated apps that were being tested or were no longer being supported. We then split apps into three equal-sized groups based on their total number of active users. To be in the top third of apps, an app needed to have 32,000 active users. To be in the top two-thirds, it needed to have 8,000.
Apps By Retention
We followed a similar process to categorize apps based on retention. For this analysis, retention was defined as the percent of people who first used an app during November 2012, who also used it again at least once more than 30 days after their first use. To be in the top third for retention, an app needed to have at least 37% of those who started using the app in November do so again more than 30 days later. To be in the top two-thirds, 22% of new users in November needed to use the app again more than 30 days later.
Combining User Numbers with Retention
Having classified apps into three groups based on both active users and retention, we then compared how the two metrics relate to one another. The proportion of apps that fall into each of the nine categories that result from considering retention and active users jointly is shown in the table below. If active users and rolling retention were completely independent, then approximately 11% of apps would be in each of the nine categories. As shown in the table, the mid level categories for each metric follow that general pattern, but the categories in the corners of the table don’t. The differences between what the distribution across the nine categories is, and what it would be if the two dimensions were completely independent, is statistically significant.
Fifteen percent of apps are in the enviable position of being the top third for active users and also in the top third for rolling retention. We refer to those as Superstar apps since they perform well on both dimensions. These apps are best positioned to generate revenue regardless of their monetization model. Another 17% of apps are at the opposite extreme: they are in the bottom third for both user numbers and retention. We refer to that category as a Black Hole. Apps in this “cell” could be relatively new apps that are still trying to establish a user base, old declining apps or apps that are of poor quality.
Possibly the most interesting apps are in the bottom right and top left corners of the table. We refer to the 6% of apps in the bottom right category as Red Dwarfs because they have a relatively small user base yet are doing well on retention. Those are likely to be successful long tail apps. In the opposite corner from that are 6% of apps we refer to as Shooting Stars since they have a lot of users, but may fade away quickly due to poor retention.
Time Spent by Retention and Active Users
Unsurprisingly, the average number of minutes per month users spend in high retention apps is greater than in low retention apps. This can be seen going from left to right in each row of the table. For example, Superstar apps have almost twice the average number of minutes per user than Shooting Star apps, 98 minutes versus 50 minutes. This correlation between average time per user and retention is statistically significant.
Average time per user per month is also positively correlated with the number of active users. This can be seen by looking from the bottom to the top of each column in the table. For example, users spend more than 50% more time in Superstar apps than in Red Dwarfs. Once again, this correlation is statistically significant; however the correlation between time per user and retention is stronger than that between time per user and active users.
Retention, Retention, Retention
These results imply that developers need to make retention their top focus. Developers can impact retention by shaping and modifying the app experience. It’s within their control. Furthermore, the association between retention and time spent implies that retention drives revenue. More repeat usage means more opportunities to generate revenue from in-app purchase and advertising. Finally, the more useful and compelling an app, the better it retains users, making acquisition efforts more efficient. Acquiring aggressively before an app retains well can be a costly mistake. On the flip side, an app that retains well can generate powerful word of mouth, which is the ultimate (and free) promotional machine. The more a developer masters retention, the better their chances of turning their Red Dwarfs into Superstars.
Suppose you’re an app developer who wants to ensure that your app is optimized to function well on 80% of the individual connected devices currently in use (e.g., my iPad, your Windows phone). How many different device models (e.g., Kindle Fire HD 8.9" Wi-Fi, Galaxy S III) do you think you need to support? 156. Maybe you’re okay with having your app optimized for only 60% of active devices. That still means that you need to support 37 different devices. Even getting to 50% means supporting 18 devices, as shown below. If you’re a large or particularly thorough app developer, reaching 90% of active devices will require supporting 331 different models.
The dominance of iOS and Android platforms has obscured the proliferation of connected device models. During January, Flurry detected 2,130 different device models with active users (defined as having app sessions during January), including 500 different device models with at least 175,000 active users.
20% of Device Models Is Still a Big Number
Using the 80/20 rule, the market for devices might even seem concentrated: just over 7% of device models account for 80% of active users. Still, the large total number of device models in use poses challenges for developers.
It’s obvious that different apps are required for different platforms. Developers can choose to serve only a portion of the app market by developing apps for only a subset of operating systems (and consequently a subset of device models). Even having made that choice, though, adaptations may be required to accommodate different versions of the same platform (e.g., iOS 6.x versus iOS 5.x, forked versions of Android, etc.), smartphones versus tablets and the increasingly wide variety of screen sizes and aspect ratios in which those devices are now available.
Developing apps on the device models that represent the majority of devices currently in active use has become an expensive and time-consuming process. Not optimizing or testing apps on devices being used by even a minority of people exposes developers to negative user experiences and potentially to buying expensive devices to troubleshoot problems as they arise.
Is the Market for App Development Ripe for Consolidation?
This fragmentation has the potential to change the app ecosystem by making it harder for small developers to compete since they are unlikely to have the resources to support the growing list of device models currently in use. They may also be disadvantaged in economies of scale in promotion (including word of mouth) if their apps are not available or do not work well on most device models. Scale is likely to be increasingly important when it comes to app development and that may lead to consolidation within the app development industry.
Developer surveys, such as Vision Mobile’s, consistently show that the revenue distribution for app developers is highly skewed: only a minority of developers make more than $500 per app per month. The increasing need for scale to ensure full functionality on the full range of connected device models in use may help explain why. The growing challenge of discoverability in an increasingly crowded app market is also likely to be part of the explanation.
So what is a small developer to do? One strategy is to focus on the device models used by the greatest number of people. Surveys consistently show that developer commitment to iOS is disproportionately strong relative to the market share for iOS devices. Our results suggest this trend is probably a consequence of developers seeking efficiency (the most users for the least work) because device models running on the iOS platform average 14 times the number of active users than device models running on other platforms. This is shown in the chart below in which the average number of active users for device models running on different operating systems are indexed to Android (where Android = 1).
It’s difficult to fully disentangle platform from manufacturer and comparing devices made by Apple to devices made by the three other device manufacturers with the greatest average number of active users per device model tells a similar story. This is shown in the chart below – this time indexed so the average number of active devices per Samsung device model = 1. As shown in the chart, on average Apple device models have more than seven times as many active users as Samsung device models and more than four times as many as Amazon device models.
App Sessions Are More Concentrated than Active Devices
Of course, some people use their devices more than others and many developers prefer to target heavier app users. So what about app sessions? They are somewhat more concentrated than active devices. As shown below, for developers to ensure they were optimized on the devices responsible for 50% of app sessions conducted during January, they would have needed to support only eight different device models and to cover 80% of sessions they would have needed to support 72 different device models. That’s still a lot of device models, but it’s less than half the device models required to reach 80% of active devices.
In addition to having more active devices per device model than other platforms, iOS device models average more app sessions per active device than device models running on other platforms. This is shown below, again using an index for which app sessions per active Android device are set to one. This further clarifies why developer support for iOS is disproportionate to iOS’ share of the installed device base. Developers can reach more active devices by developing for a smaller number of device models on iOS and they can also capture the attention of very active users. People who have iOS devices tend to have more app sessions, creating more opportunities for in app purchases, advertising revenue and paid app purchases.
Viewed at the manufacturer level, Apple device models average more sessions per device than device models made by the other manufacturers previously shown. This is shown below, again indexed so that average sessions per Samsung device = 1.
The App Development Company
With competition in the device market heating up, manufacturers seem likely to fill and expand product lines with an increasing number of devices intended to differentiate themselves and address the preferences of specific types of users. That implies that it will only become more difficult for developers to optimize, test and support their apps for use on all device models. And yet doing exactly that is likely to be increasingly important for app developers given the market for apps is also becoming more crowded and more competitive, making negative user experiences more damaging. Promoting apps and leveraging that investment in promotion across as many potential users as possible will also become all the more critical. Putting all of this together, we expect a future in which app developers are less frequently individuals with a creative idea and a laptop and more frequently, companies designed to develop, produce and distribute apps at scale.