The right analytics in a mobile application

    The vast majority of mobile applications come out without integrated analytics frameworks and without their proper configuration. This is very sad, the application is not a website where you can stick Google Analytics in 10 minutes and collect statistics.

    I see four reasons to think about analytics in a mobile application BEFORE its release:
    • The first time you never succeed in making a successful mobile application. After the release, an analysis of the behavior of real users in the “battle” is required.
    • Evaluating the effectiveness of advertising campaigns is impossible without analytics.
    • Two weeks are spent laying out iOs applications, when you catch it you have to wait for analytics for about a month.
    • Data from the past cannot be returned.

    Under the cut, a set of analytics tools and recommendations on how to configure them are accepted in the Touch Instinct and valuable comments from LinguaLeo and the Mobio mobile advertising agency.




    User Behavior Analytics


    By analytics of user behavior, I mean the collection and display of such data as:
    - the number of daily audience (DAU);
    - user paths in the application;
    - How often are application functions used;
    - achievement of goals (registration, purchase, three visits in a month, ...);
    - frequency of audience return (retention rate).

    The web typically uses the free and powerful Google Analytics to retrieve this data. Google Analytics has a great mobile version . To make it work, you need to build in a special library of iOS , Android .



    Analytics Mobile setup guidelines

    Track the opening of each screen. In Android, we call the tracker in onStart and onStop for each Activity. In iOS, we inherit the ViewController from GAITrackedViewController and prescribe the name of the screen .

    Track the click of any button, interactive element on the screen, or search query. In Google Analytics, Events are for this purpose. In the handlers of the elements we are interested in, we call sendEvent . Think about what action parameters will be required for analytics: the text of the search query, the id of the object on the card of which the button was pressed, ... and transfer them to the Label.

    If the application has built-in purchases, do not forget to configure their tracking. ( Android ,iOS )

    Analytics is powerful and free, all analysts know how to work with it. Almost perfect tool. If not for one fat BUT, Google Anayltics is not able to track the source of installations for iOS. Able only for Android. This sharply crosses out all its advantages, so I recommend using free Flurry .

    Ilya Krasinsky , Lingua Leo
    1. It is much worse that GA works on sessions. This often leads to errors in the calculation of unique ones, i.e. his uniques very often lie, especially if in some products, for example, after registration / authorization, for some reason, the session breaks (indirectly, you can track it for a large number of 1-3 second sessions) or compare it with other analytics

    2. Out of the box knows how to do cohort analysis, you have to use variables and record the date of the first session of the user.

    3. GA does not calculate conversions on the basis of all the data, it constantly uses only a small% of the data and on the basis of them approximates the results of the entire sample, especially when you look at the behavior of the metric for a month or two.

    Last fall, mobile version 2.0 was released, some bugs were fixed, but still, from time to time on the application, you encounter outrageous bugs in the data


    To use Flurry, embed the library in the application ( iOS , Android ).

    In Flurry there is no concept of the screen, there are only events (logEvent). Never use a single pageView event with parameters . This is then impossible to use. For each screen, trigger a separate event.

    Alexey Pisarevsky , Mobio Mobile Marketing Agency
    When setting up your Flurry events, start with the most basic. For example, setting up Flurry events for a series application with paying for the series with money or watching a commercial.

    The key events in the application will be:
    • purchases;
    • video views;
    • authorization;
    • sharing in social networks;

    Basic setup of events and parameters for measuring these events:

    Ilya Krasinsky , Lingua Leo
    I recommend using verbs with additions in the event name: Success Purchase, Press plus button, Create Task, Watch movie.

    This greatly facilitates the construction and analysis of funnels - you read a coherent text.

    In addition to GA and Flurry are popular : MixPanel (the most powerful tool), Localytics and Apsalar . I recommend starting with Flurry.

    Installation Sources


    On the web, you can accurately track where each user came from thanks to the referer and the ability to pass arbitrary parameters to the url (for example, utm tags ).

    In the mobile world, things are more complicated. After clicking on the link to the application, the user leaves the browser in the application store and the chain breaks. That is, without additional gestures, you cannot evaluate the effectiveness of marketing activities .



    Google has made support for referer and utm tags for Android. Just add a couple of lines to the manifest and you will see traffic sources in Analytics. Apple, unfortunately, is not so generous and it is impossible to understand where the installation came from. Therefore, Google Analytics does not address the issue of mobile analytics.

    Third-party services help: AppsFlyer , Distimo , Mobile app tracking , Adxtracking and ... Flurry.

    These tools work heuristically and do not always correctly associate the installation with a link. It works something like this:
    1. A unique link is created for each channel (ex. Iphones.ru, google admob, habr, banner on the site, email newsletter ...)
    2. User opens link on smartphone
    3. The server collects the maximum open data about the smartphone and redirects the user to the AppStore
    4. After installing the application, the method from the SDK tracker is launched. It collects data about the smartphone again and compares with the data from paragraph 3.

    Ilya Krasinsky , Lingua Leo
    MAT (Mobile App Tracking) is generally the most reliable. It is believed that in 80% of cases it accurately determines the source.

    The tracking tool in flury appeared not so long ago and so far I have no data on its reliability.

    MAT is good in that it can wrap the link in flory and thus there will be 2 tracking systems.

    Flury is good because on the basis of tracking you can build segments and analyze the audience that came from different channels.


    Alexey Pisarevsky , Mobio Mobile Marketing Agency
    I recommend that you build in a tracking system that allows you to buy ads using the CPI model (cost-per-install, payment for installation). Such systems inform the ad network of the fact of installation. There are several CPI networks in Russia: Clickrocket , Unilead , Mobioffers), ... they work on the same principle as CPA networks on the web. Set a price for installation (for example, $ 3), the offer is sent to network partners. The sites accept this offer and advertise. Upon installation, the network pays the sites a fee equal to your rate minus the network commission. To use CPI networks, you must send her information from which site the user came from. You can embed a library of each site or tracker that is integrated with a large number of CPI networks. Flurry is integrated with few networks, and Google Analytics is not able to send information at all. They are not suitable for this purpose.

    The Most Powerful Mobile App Tracking Tracker. It is integrated with the vast majority of advertising sources. Unfortunately, it’s expensive: 0.2 cents for each measured event, including clicks and events. Other popular trackers are AdXtracking and Appsflyer. Their price is more loyal. Appsflyer only charges 5 cents per install. Organic installations are free.

    I recommend Appsflyer + Flurry to my clients. The first is to evaluate the effectiveness of promotion, and the second is to analyze user behavior.


    In my opinion, in most cases Flurry is the perfect solution. This analytics of user behavior and tracking the source of installations for iOS / Android and even free. Although, for my taste, Google Analytics is prettier and works significantly faster.



    Push analytics


    Push alerts are an important tool for engaging users. It is important not only to correctly send push notifications, but also to constantly monitor their effectiveness and adjust the strategy.

    Push analytics can be divided into two parts: analysis of user behavior from push messages and analytics in the distribution service.

    Behavioral analysis is simple. When you start the application with push notifications, we send the event to Flurry, indicating the type and text of the message. In standard reports, users who open the application from various push messages will appear.

    With the mailing service, the story is a little more complicated, if you have not decided yet, I recommend using Parse, it has excellent reports. If you are creating your own push notification admin panel, collect the following statistics:
    • quantity sent;
    • the number of application openings after a push message;
    • percentage of subscribers to push among all users of the application.

    Push analytics in the mailing service allows you to conduct a / b testing. For example, before mass mailing we do a test for 20% of users, look at which text has the higher conversion to openings and send it to the remaining 80% (thanks to Maxim Ilyakhov for the idea ).

    Heatmaps


    This rather pleasant addition, than the necessary tool, is useful for identifying problems in the application interface. It looks something like this. I recommend the Heatma.ps service , although it costs money: $ 18 for applications with a small number of users, $ 90 for applications with 10,000 active users per month. All applications that Touch Instinct has been launching since August comply with these rules by default. And we teach customers how to use all this. It is very interesting to hear your opinion! What should be of analytics necessary? Maybe there is something better than Flurry? Or are you adding unusual events like “scrolled to the second screen”?









    Subscribe to our habr-blog (the button in the upper right). Every Thursday, interesting articles about mobile development, marketing and the mobile studio business. Next article (August 29) “Layout of Android layouts without pain” practical tips on how to make up a layout designer for Android, which will adapt to all permissions, correctly stretch elements and scale fonts.

    Also popular now: