Apps are now must-have for broadcasters and MVPDs alike-but how to extract the most significant value out of them? Like the other elements of a TV business, high-quality analytics are essential for all aspects of day-to-day and long-term decision-making for your apps. Analytics should be part of the build process of your app - the data engineering team should not build it as an afterthought - and should be integrated with the rest of your TV data strategy.
There are multiple ways that app data can be used to drive key subscriber analytics around customer acquisition, retention, upsell and engagement. However, before you start looking at this kind of predictive analytics you have to get the basics in place first.
When advising customers on how to build out their mobile analytics, we usually look at a hierarchy of downloads, users and events (including sessions and views). In this article, we're going to look at each in turn and understand how best to use them.
While the classic metric for analyzing apps is how often customers download the app, this can be a misleading metric (although it is often the first thing that management teams generally want to know). Users using multiple devices will typically download the app onto all of them, meaning that the number of downloads does not reflect the number of users, although the app stores do provide some recompense for this by stripping out downloads from a user who has already downloaded your app to another device.
The metric that is useful is the number of users who are actually using an app, as the difference between the number of downloads and the number of apps users is typically large, but calculating the number of users is harder than the number of downloads.
An interesting metric for us at Deductive is the number of times an app is uninstalled and the reasons why. There are a multitude of reasons why a user might uninstall your app, and these factors can be instructive. Some reasons - a user has run out of memory, competitor activity - are mostly outside of your control. Others - too buggy, too many ads, not engaging enough - are not. Metrics which the company can do something about are always going to be more useful than metrics which it is harder to see how to change.
Apart from this uninstall metric, what our data science consulting team see as the most critical download metric is the cost per install; i.e., how much is the company spending per install or, even better, per actual customer. While this is an easy calculation for cost per device downloaded, it is important to keep focused on the number of customers and not the number of devices, even though this is a more difficult metric to calculate.
In focussing on users, it is very useful to differentiate between a new and a returning user. Given that many users try an app and then abandon it (typically about 50% within three months according to some benchmarks), it is useful to know how many of those who abandoned the app will then return; we consider it likely that any app will have a sufficient number of returning customers for the metric to be worth measuring.
Other usage metrics we recommend companies embark on as a priority include:
• User flow, modeling the path taken by an average user in completing a task on the app • Path analysis, examining the relationships between the variables that may occur when a customer uses the app. Both this and user flow are important for gaining a rounded picture of how a user engages with your app. • Funnels, exploring the steps a user takes to reach a goal, such as becoming a paying customer. By understanding each step a customer takes to get to the goal it becomes easier to encourage new users to follow the same path. Goals are a set of activities which have been defined by the app provider, such as that a customer watches x hours of video or shares x articles with their friends or a goal could perhaps be something as simple as creating a playlist. These goals need to be tightly defined while building the app, and specifically its analytics, and then monitored closely.
Events, sessions, and views
The next step is to try and calculate how your users are actually using the app. Events, sessions, and views focus on what users are doing with the app.
• Events are any kind of event happening as the customer uses the app and are particularly important because they help drive both customer retention and customer engagement and are an easy way to understand how the app is used by real customers, as opposed to what the app provider intended. Typical examples of events include watching a video or buying something through the app; though merely viewing a page would not be classified as an event. Unlike with views, the app provider needs to define these events within the analytics process where they can be beneficial for gaining a picture of how customers are using the app, including which parts are popular and which aren't.
• Events are any event happening as the customer uses the app and are particularly important because they help drive both customer retention and customer engagement and are an easy way to understand how the app is used by real customers, as opposed to what the app provider intended. Typical examples of events include watching a video or buying something through the app; though merely viewing a page would not be classified as an event. Unlike with views, the app provider needs to define these events within the analytics process where they can be helpful for gaining a picture of how customers are using the app, including which parts are popular and which aren't.
• Sessions describe the interactions between the user and the app, typically made up of multiple events. Trying to understand why users enter and leave the app is always a useful metric. When does a session end? Either after a certain period (say 20 mins after activity ceases or at a particular time, e.g., midnight, or when the user closes the app). Closure is a useful session metric albeit an even more interesting one is noting what is called a campaign change, where a user enters the app via a campaign, leaves the app and then re-enters via a different campaign.
• Views refers to screen views, which is any view of a page in your website accessed through the app. This kind of tracking is generally built within the apps internal analytic system, so nothing has to be done other than ensure you are using the dataset which contains this data for all your app downloads. It is, however, useful information as it shows what consumers do with the app. If you can match this info against, say, which users don't abandon the app within three months and those who do it is possible to build a picture of which activities are making this a popular app, and which aren't, information which can be used to develop the app, perhaps adding new features and withdrawing unpopular ones.
And then there's revenue. This underpins your success or otherwise, and can then be used to calculate cost per acquisition. Finally, we would recommend that you track ratings from those app stores which offer your app. Comparing these to the ranking of your competitors (i.e., those with similar apps) is essential. If you aren't satisfied with user engagement with your app the first question to ask is whether your competitors are doing any better. If they are, you are in the right app business but need to tweak your app to make it more attractive.
App analytics is an entire industry in itself, and there are almost limitless things that you can do when you start reporting and modeling the data your app produces, we recommend the above as good starting steps.