How to choose an analytical system for your mobile game
There are lots of analytical systems and tools on the market. We’ve compared some of these systems specified for mobile games. I hope, this article will help you to make the choice you’d like (as there is no right choice, you’ll see it).
For those who do not want to read, here is a comparative table:
And now let’s take a closer look at every point.
Key metrics like DAU/MAU/WAU, retention, monetization metrics (revenue, ARPU, ARPPU, etc.) are those 20% of the functional of an analytical system that solves 80% of the problems.
It is not surprising that the systems are in complete order with the metrics and they all have the full pack of key metrics. The only exception is Flurry, where there is no monetization metrics and all the conclusions are made only on activity and retention metrics. On the other hand, Flurry has such useful metric as Frequency of Use or Sessions Per User, particularly relevant for mobile games.
Custom events and funnels
Of course, metrics provide the answers to 80% of the questions, but will you be satisfied with only 80%?
Mobile analytics is usually built on custom events. “Event” means any action performed by the user: click on the menu item, the level completion, killing the boss, inviting friends, raise the rating, and so on. Typically, each event has parameters: for example, what level was completed, how much time killing the boss took, what was the virtual currency balance at that moment, etc.
Of course, adding events and the ability to customize this functional makes analytics profound and extended. More exciting is that this analytical tool is limited only by your imagination.
Funnels are perhaps the leading MixPanel feature. Funnels can be built on any number of steps and for any period. The funnel is counted in a matter of seconds, even though retrospectively (the funnel is created today, and the statistics are shown even for yesterday). Restriction: each event in the funnel can have only one parameter. This is inconvenient, but for most purposes a single parameter is sufficient.
In general, the Localytics funnels functional practically repeats the same functional of MixPanel at the moment.
There is funnel function in the system, but with the following limitations:
- the funnel is counted for about a day;
- there is no possibility to apply conditions to the various steps of the funnel;
- it is impossible to set the parameters of the events.
A funnel is created within a few minutes, any change in the existing funnel is recounted. There is a limit of up to 5 steps per funnel, but that’s enough for most of the funnels.
The funnel is created within a few minutes, any change in the existing funnel is re-counted. You may apply filters to every step. There is a limit to 20 steps per funnel.
The funnel is created within several hours, and any change in the existing funnel is recounted. Filters are possible to apply to every step.
Segmentation is also an essential tool for mobile application analytics. Draw conclusions for all users at once is a wrong way.
The larger the community is, the greater is the differences between users. For example, the LTV metric, counted for all users at once, has a little to tell you. You can only compare it with CPI, and that’s all. But if you divide users into segments and calculate LTV for each segment separately, you would be able to make bette and more correct decisions.
Users may vary by country (compare the Japanese and, for example, the Italian market), on the socio-demographic characteristics, maturity in the game, the indicator of monetization (paying/non-paying), or the size of payments. In short, by anything.
The advanced version is the segmentation of the completion/non-completion of any event: for example, compare the users who passed the tutorial with those who missed it — who would have higher retention?
Segmentation is present, including the completion of events. Data is counted retrospectively. It works quite convenient.
Again, the functional is very similar to the same functional of MixPanel. The differences between these two are to appear later on.
There are segments but with very limited functional. Firstly, they are counted only from the date of the creation (in other systems, you can still see the segments retrospectively). Secondly, there is no way to change the segment after creation. Finally, Flurry has no option to segment by the event parameters; the maximum you can get is a pie chart of the values of this parameter.
Segmentation is available, including the completion of events. Runs retrospectively.
Segmentation is available, including the completion of events. Runs retrospectively.
GameAnalytics provides an option of setting segments on completion / non-completion of the event, and for all the other cuts (country, payments, device) only relevant reports may be built, but not segments.
User Acquisition Analysis
Which channel of user acquisition is the most profitable? How many users from each channel remain in the game after a day or a week? What is the average revenue the user brings in the first month of his life in the game from each channel? When will the paid traffic pay off? Analytical systems are able to respond to all these questions.
In pure form this feature doesn’t exist, you should use third-party services that would send data on installs of the application to the system. For MixPanel, those services may be Adjust, AppsFlyer.
Such a possibility exists for InMobi, Adwords, Facebook. To track users from other sources, you have to connect third-party services (AppsFlyer, Adjust).
There is a built-in mechanism for determining the source of traffic and analysis, which works well, even if not with all advertising networks (for example, does not work with Facebook).
This option is under 3rd Party Attribution, and third-party services are required to connect. For devtodev, such services are MobileAppTracking, AppsFlyer, and Adjust.
Perhaps we were not looking well enough, but we haven’t found this opportunity at deltaDNA. In the Acquisition section, they are talking about something else though very useful things (see “Benefits for game projects”).
The option exists but third-party services are required to connect: Adjust, AppsFlyer, MobileAppTracking, Apsalar.
Not the primary function of analytical systems, but rather a nice bonus.
Proper and competent use of push-notification can increase retention, return users from churn and, as a consequence, increase the revenue (usually slightly, but extra 10–20% to revenue are not redundant). And if the analytical system provides the ability to send and analyze push-notifications, it is not only convenient but also useful.
Push notifications are provided by all the systems described, except for Flurry and GameAnalytics.
Benefits for game projects
Do not forget that in addition to games, there are other services, and analytical systems may choose two directions in their development: either be a universal analytical system for all services or perform deep analytics especially for gaming projects. We are talking about games and as game analysts want an analytical system to provide tools, that would be useful just for us to analyze the economy of the game, game levels or anything else more.
devtodev proposes the section of Game Metrics, that would be useful for game economics analysis, it analyzes all the IAP, the balance of currency in the accounts of the users, the distribution of users by levels. One more notable report is Tutorial Steps, which is a step by step tutorial funnel, that may consist of several hundred steps — it is easy to track the churn of users on every step.
devtodev also offers the group of reports called Social Metrics. They allow you to evaluate the distribution of your players on social networks, as well as to fix the time of publication of the post in a particular social network and link it to an event in the game.
devtodev also provides a complete range of reports about paying users including the analysis of frequency and amounts of payments, recurrent payments analysis, report about periods of conversion into paying users etc.
There are two excellent reports at deltaDNA. LTV Predictor predicts how much revenue a user would bring you for all the time spent in the game. And Retention Predictor, in turn, predicts how many users would stay in the game in the future.
Ease of integration and consistency of SDK is, first, the subjective question, and secondly, always quickly solved: if you have issues with integration, it is of interest of the analytical system to resolve them as quickly as possible.
So we omit the question on the ease of integration. Let’s just take a look at the platforms supported by the analytical systems to date (December 2015, based on the available SDK from the official sites of the services).
Android and iOS, as we see, are added at all of the systems, and then everything is not so rosy. Note devtodev as a universal analytical system, eight platforms, including the web, Mac OS, and Windows.
Data accuracy and speed of the appearance in the system
In order to take relevant analytical solutions, we need to have the correct data. The faster the data generated by the user in the application appear in the analytical system, the more decisions can be taken per unit time. In all the systems described, the delay may vary from several seconds to several minutes, which generally is normal — the decision can be taken in one hour.
Flurry and Game Analytics are the only exceptions — the delay is not documented, but in practice, the data gets in the system in about 24 hours. For operational tasks is prohibitively long. Also, Flurry is often accused of the inaccuracy of data, which is likely caused by a long delay.
Let’s assume we have the game where all users are active and generate 5 events per day for 30 days. So in one month by one user is generated 150 events or data points, in the terminology of the analytical systems pricing. Let’s see what it would cost us to use the analytical system per month, if our game has MAU equaled to 1000, 10 000 and 100 000 users.
GameAnalytics and Flurry are free for use, hence the limitations in functionality. Among paid services devtodev looks like a cost-effective option: for projects with MAU equal to 100k, the price does not exceed $1000.
Now let’s talk about what did not fit in the previous parts.
If the fixed set of metrics is not enough, there is a section at MixPanel called Formulas, where you may create a new metric and describe the rules of its calculation. This may save time as you don’t have to upload metric values and count them manually at Excel.
Also, there is a section of Surveys at MixPanel, which allows to organize surveys for users and then analyze the results.
Localytics analyzes lifetime value well and offers a good tool to work with this value: LTV can be counted for different segments of users.
In particular, users with large LTV may be distinguished (the so-called whales) and their behavior may be analyzed separately.
Flurry offers the classification of users by interest on the basis of other applications they use, and by age and sex based on patterns of behavior.
Also, for this purpose Flurry offers reports Users in Common with this App and Users in Common with My Apps. These are useful features which can be of practical use, such as cross-promotion.
devtodev receives data not only from the application but also from the market, so, for example, a situation where the actual revenue of the application does not match the revenue displayed in the system, is impossible in the case of devtodev.
Noteworthy is the ease of work with reports and dashboards: at devtodev you may build reports on several projects and metrics simultaneously while the data is available in tabular form, and there is no need to upload anything at Excel. Reports and dashboards may be shared and the access might be limited at the level of metrics. For example, you can separate those who play on one device from those who play on several devices.
deltaDNA offers a fast and convenient tool for creating custom reports, it is called Slice&Dice. Users can be grouped as you like and you may review metrics by the selected group.
Among other interesting reports, we would like to note the report on time to the first purchase (the time of conversion into paying user) and Monetization Matrix, allowing to see how the paying users are divided by the number of payments and the revenue they brought.
GameAnalytics offers the opportunity to compare your project with the other projects from the industry. Without naming specific applications, the service just tells you what percentage of games are above you in terms of one or another metric, and how many are below. According to the official website, the conclusions are made from the analysis of 5200 games of 20 genres.
So, we compared the main analytical systems with each other. We tried to be as impersonal as possible. Based on the comparison table, we may give the following recommendations:
- If you want to get free analytics — choose Flurry or GameAnalytics. You will answer the main questions, but be prepared for delays in receiving data.
- If you have a small project with MAU less than 10 thousand users, then MixPanel, Localytics, deltaDNA will be free for you as long as MAU does not exceed the mark of 10 thousand.
- If you are ready to pay for the analytics, no matter of the scale of your project, choose devtodev. At large projects, notably, this is the best offer in the market among the premium analytical systems.
There’s no ideal analytical system and it’s only up to you which one to choose. We only hope that, among other things, you got our conclusion about the benefits of living with the analytical system than without it.