Gamasutra: The Art & Business of Making Gamesspacer
View All     RSS
October 23, 2014
arrowPress Releases
October 23, 2014
PR Newswire
View All
View All     Submit Event





If you enjoy reading this site, you might also want to check out these UBM Tech sites:


 
User Acquisition For Mobile Games: The True Cost of Getting an Engaged User
by Riz Virk on 04/04/14 04:12:00 pm   Featured Blogs

The following blog post, unless otherwise noted, was written by a member of Gamasutra’s community.
The thoughts and opinions expressed are those of the writer and not Gamasutra or its parent company.

 

In 2009, I was part of the team at Tapjoy which came up with the first mobile app offer wall. We were tackling two of the most important problems for free to play mobile game developers at that time: distribution (how to get more users for a free to play game), and monetization (how to make money from free users).

At that time, most ads were sold on a CPM basis or a CPC basis.  But most app developers didn’t really care about CPM or CPC, they wanted to know the true cost of getting a user.  So, many app developers were forced to calculate eCPI (effective Cost Per Install) by backing out the impression and the conversion numbers.

Eventually, through the success of Tapjoy and other ad networks, including Chartboost, Flurry, among others, the industry standard for app advertising shifted to CPI – or Cost Per Install.   Even display networks like AdMob (which was sold to Google) and what later became Apple’s iAd had to shift to providing enough data so that advertisers could calculate eCPI numbers for comparison.  Today even Facebook has been forced to provide CPI rates for app developers.

Now that CPI has been the standard for a few years, mobile user acquisition is still a major issue for most app developers.  In fact, the rising cost of CPI is often the #1 issue for a new game or mobile app – many apps have an LTV (lifetime value) that is less than the prevailing CPI rates.

To complicate matters, the cost-per-install quoted by a network isn’t always a reflection of the quality of users.  Not only is it often expensive to get installs (on iOS the numbers have been astronomical and they’ve been rising on Android as well), but different networks charge different amounts and result in differing quality of users, so a simple CPI comparison isn’t enough.

While CPI numbers are tossed around almost casually now, I believe that most companies still haven’t truly figured out the cost of getting a real user.  By real user, I don’t mean an install – in the mobile world, particularly for free to play games, most users who install a game never play it after the first day!

I was speaking to a very knowledgeable developer recently who was fed up with CPI – he wanted to know the true cost of getting an engaged user.  Since most games have less than 50% day 2 retention rate (i.e. less than half of the users who install the app return the next day), the true cost of getting a user back on day 2 was what really mattered to him. If they came back on the second day, then there was a much better chance they could convert into long-term players of his game.

To get the cost of a day 2 returning user, you have to divide the CPI by the day 2 retention rate for your app.   Seems simple? Not really, because users that come from different advertising networks will have a different day 2 retention rate.   Which means you have to figure out and track the retention rate on a channel-by-channel basis.

As an example: if CPI is $1 but the day 2 retention rate is 50% from users from channel A, then the true cost of getting a day 2 user from this channel is $2.  If another network, say an incented network, channel B, charges only .20 cents for a user, but their return rate is much lower, say 20%, at first glance it seems like you might want to go for the higher quality, higher cost user - after all, 20% day 2 retention is much lower than 50%. 

But, if you do the math, you’ll see that in channel B, the cost of getting an engaged day 2 user would be only $1 (CPI * Day 2 retention).  So in this example, channel B is not just less expensive to buy an install from, but less expensive to get an engaged user, even though on average the quality of users might be lower.  Now if the retention rate of channel B was only 10%, then the cost of a day 2 user would balloon to $2, and there would be no significant quality difference with channel A.

However, the cost of a day 2 user is only one measurement of a the cost of an engaged user.  How about the cost of getting a user that reaches level 5?  Is that user more valuable than a user that just installed the game and never came back?  Of course it is.  

For a free-to-play slot machine game, it might be that users who have spun the wheel 30 times are much more likely to become engaged users or paying users.

We call these key “engagement points.”  Developers want to maximize the number of users that reach these.  For a card battle game, a key “engagement point” might be when someone joins a clan, or enters a weeklong event or tournament.

I was speaking to another mobile developer recently and he lamented that free-to-play games only have a few seconds to make an impression on a user.  It’s pretty difficult for users to really get to know a game in only one play session.  Many good games will lose users because they didn’t hook them in the first 30 seconds or 1 minute.  Because of this, it’s important for mobile game developers to find ways to get a new user back into their game after the first play session several times before they make a decision whether to keep playing this game or not.  This is a type of “re-engagement” that can be critical in improving the engagement metrics of a game.

At Midverse Studios, we started out as a game developer and faced all of these challenges in promoting our games.  After trying out pretty much every ad network / user acquisition source available, we took the best elements of what we had seen and implemented our new CPE+ model for our new AppEngage network.  We were fed up with CPI costs and the varying quality of users – when what we really wanted was an “engaged user.”  We also wanted to be sure that any users we paid for would log into the game multiple times, and we wanted those users to return on the second day -- if they did this, we knew we’d have a much better chance of hooking them.

CPE generally refers to Cost per Engagement, or Cost Per Engaged User.  The industry has been experimenting with CPE models over the last year or so, paying them lip service, but most ad networks have only half-heartedly implemented them and haven’t been able to provide any decent volume.

With CPE+, we recognized that a simple install is not as valuable as a user that completes the tutorial and actually plays the game a few times.  A user that plays the game a few times isn’t as valuable as one that levels up and comes back the next day.  And a user that comes back the next day may not be as valuable as a user that enters a tournament or a clan in the first week. And so on.

Event

Value of user

Install app

Low

Complete tutorial

Low

Play game on day 1

Medium

Level up on day 1

Medium

Launch game on day 2

Medium-High

Play 5 rounds on day 2

High

Level up on day 2

High

...

 

Reach level 10

Very High

Join clan or Tournament

Very High

Table 1: The relative value of users in the engagement funnel

Table 1 shows some sample types of engagement points and the relative “value” of users at each step.  You can think of this as an engagement funnel, where each step consists of a higher value but decreasing number of users who achieve that step.

In fact, you can think of the engagement funnel as a parallel achievement system to your own in-game questing or achievement system. 

It is a “new user achievement system” – you want to guide a newly purchased user down this engagement funnel.  Moreover, in an ideal world, the cost of a user would vary depending on how far down the engagement funnel they got, and developers could bid on how much each engagement point is worth. 

This is a long-term vision and a completely different way of thinking about user acquisition.  Would you like to buy a user that reaches level 30 of your game? How much are you willing to pay?

In order to fully implement CPE+, a new model in user acquisition, the existing format of ads won’t really work.  Existing formats – banner ads, offer-walls, full screen display ads, videos - are insufficient to provide a coherent user interface to guide users to try your game, log in again, and guide them to important activation points on subsequent days.  Moreover, rewards can encourage users to try new games and travel down the engagement funnel, increasing their play-time and their “evaluation” period.

To fully implement CPE+, what’s really needed is a cross-game achievement system.  Many companies have built custom integrations between games based on achievements. For example, a farming game might give a custom reward for someone that tries out their new aquarium game and reaches level 3.  The industry is full of experiments like this, usually by a developer in two of their own games.  I know anecdotally that many of these experiments have been successful at some of the best-known mobile game companies out there.  The problem is that they are usually implemented in a one-off way and cannot be re-used across games and with different achievements and different rewards.

In order for the industry to fully embrace CPE+, it would need to:

1.Give users a way to try new games that they might be interested in.

2.Reward users who try new games to encourage volume.

3.Provide a generic way to represent events and achievements

4.Guide users towards activation points and track their progress

5.Give users increasing rewards for reaching further down the engagement funnel

6.Get users to come back to the game after the first day.

7.Provide a way to charge app developers based on the quality of users delivered and not simply based on “installs.”

In my experience with mobile and social gaming, I have learned that users love two things in their games: rewards and achievements.  That’s why almost every game has some questing or achievement system.  There’s nothing like seeing goals ticked off presenting how far the user has come and how close they are to their goal, which is usually some kind of in game reward.

 

Figure 1– The AppEngage VIP Rewards dialog, implementing CPE+ 

Figure 1 shows Midverse’s specific implementation of CPE+, called AppEngage.   In our implementation, users earn rewards by trying a limited number of new games.  As they play, they get checkmarks for completing each “action” in each game, and are rewarded based on how may “checkmarks” they collect.  The rewards come at increasing levels - bronze reward for trying out different games, silver reward for getting to first level of engagement in those games, and gold rewards for high level of interaction/returning to the games often.  Users must engage not only in new games, but in the current game to reach each tier of rewards.

This is a very different way to think about user acquisition – it goes beyond simple measurement of how many installs can be provided at what cost.  In fact, we believe that this CPE+ model can replace CPI as the dominant way to acquire users, but it will take some time since the whole industry is just recovering from being stuck in a commodity-like cost-per-install world.  In other words, the refrain of the day has become something like: “hey buddy, how many installs can you give me and for how much?”

But installs aren’t fungible commodities like oil or gold - the quality of installs varies greatly.  Many app developers are coming up with models to measure the quality of users they are getting from different user acquisition channels.  I believe that engagement may be the most important measurement, and that’s why we came up with CPE+ as a model and started to measure the cost of an truly engaged user.

As the industry recovers, I believe that a shift to more “quality” users is inevitable, and CPE+ is a great way to get there.

About the author:

Rizwan Virk is co-founder of Midverse Studios and was previously co-founder of Gameview Studios, part of the original team at Tapjoy, and an angel investor in many gaming companies.  To check out Midverse’s new CPE+ network, called AppEngage, visit www.midversestudios.com.


Related Jobs

Square Enix Co., Ltd.
Square Enix Co., Ltd. — Tokyo, Japan
[10.23.14]

Programmers
Infinity Ward / Activision
Infinity Ward / Activision — Woodland Hills, California, United States
[10.23.14]

Senior Sound Designer - Infinity Ward
Treyarch / Activision
Treyarch / Activision — Santa Monica, California, United States
[10.23.14]

Multiplayer Level Designer - Treyarch
Nexon America, Inc.
Nexon America, Inc. — El Segundo, California, United States
[10.22.14]

Localization Coordinator






Comments


Kenneth Blaney
profile image
In light of a different article here taping about avoiding cloning by providing player engagement mechanisms ( like meta games) I think you are on to something great here. That is, reward players not just for sticking to your version of a game, but also reward them for engaging in your version of other games also.

Valve, incidentally, has sort of doe this with Team Fortress 2 by awarding special items to players who complete tasks in other games. I imagine that a specifically ambitious indie developer could use similar methods to draw attention to a back catalogue even outside the context of free to play or, I suppose, use a free to play game as something of a loss leader for traditionally sold games.

Matthew Fundaun
profile image
I have to agree with the cloning article's relevance. One of the interesting things pointed out was how important having a community is.

That people only talk about things they care about in some fashion is an obvious statement, but it's still a significant one. Tactics on fostering community seems like a natural extension of getting players engaged.

Ryan Christensen
profile image
Great information and pretty insightful from someone with experience in the trenches. I agree that engagement will eventually be the most desired measurement but sometimes you have to let the game do the work as well. Games good enough should get the depth of play but I like the rewarding mechanic.

Eventually I can see this going to identifying whales and offering them things that draw them away from other games, almost like high rollers in Vegas get hooked up. Gets into some gray areas but it is inevitable. Nothing wrong with drawing people to your arcade and giving them some tokens if they try other games.

Ross Kravitz
profile image
I think companies already account for a channel's user quality in the obvious way, which is by computing the average lifetime value for users in that channel. In theory*, this is roughly equivalent to your methodology, because it bakes in the effect of users dropping out after a day or two. For example, if your two day retention rate were to drop by 50%, this would have the effect of cutting average lifetime value by 50%. Of course that example is rough, but the point is that user quality is already used to inform LTV, which is part of the calculation of how attractive a channel is, along with CPI.

*My explanation breaks down in too crucial ways, the first I think is implicit in your article, and the second you don't consider First, using a straight LTV calculation ignores the synergistic effect that people can have on each other, i.e. my alliance membership causes you to spend more money than you otherwise would have. At the end of the day, this probably means that the LTV model should be refined, but it doesn't necessarily mean that the CPI metric needs to be refined with downstream information. My point is that it's largely equivalent to modify either of these metrics.

The second phenomenon, one that I feel is totally ignored in general, is that estimating LTV is a super super noisy proposition. From my own tests, I believe you need a user base of 20,000 users to be able to estimate LTV to within +/- 15% relative error. For the nerds, I actually bootstrapped different sample sizes of players, to get this estimate. What this means is that with smaller channels, it's extremely difficult to estimate LTV, as there is just too much randomness in how much players spend. The skewness/whales/small % of payers are the primary reason why this is the case. Therefore, optimizing around retention (instead of monetization) is potentially extremely useful, because it can serve ultimately as a less noisy predictor of longer term monetization. Given that, your model is definitely a step in the right direction. At the same time, however, you could still just consider vanilla CPI and augment LTV with a target which is based on retention goals.


none
 
Comment: