Today there’s a large and growing number of UA partners and solutions to choose from. Some of them manage user acquisition campaigns manually, while others use machine learning powered algorithms. In this article we’ll talk about the latter.
Hopefully you find insights on how to set up your UA campaign in a way that is both representative and cost efficient. We’ll provide ways to adapt your pricing models and make your campaigns even more profitable, while minimizing wasteful spending.
You can find more information on UA pricing models in our previous article CPI, CPA, CPM or RevShare: choosing the right UA pricing model. Feel free to check it out!
Let’s start with the basic process of setting up programmatic user acquisition campaigns. Generally it includes 2 stages that are closely connected: learning and optimization.
Learning and Optimization
Advanced user acquisition solutions use machine learning algorithms to get as many insights from your user acquisition campaigns as possible. This information is applied in order to narrow targeting and figure out your most profitable and best performing traffic sources. These algorithms also make sure that if a traffic source spends your money but doesn’t convert users, it’s automatically turned off. The solution responsible for that is called the predictor.
It targets segments of users that are most likely to engage and become paying users. But in order to do all of that, the predictor needs to “learn”.
The learning stage is one of the most important parts of your UA campaign. In order for the predictor to be able to reach KPIs it needs to have correct targeting, clear goals, representative metrics and a sizable amount of traffic to collect information from. Otherwise you waste money bringing in traffic that won’t convert or become paying users. So as a result you’ve exhausted your resources only to start from scratch.
Applying the Learning Stage to Each Particular App
Depending on how well you know your app, the learning stage can look different and have different goals. Here are some possible cases.
Let’s say, you’ve already chosen your app’s pricing model and preferable rate, plus have sophisticated analytics and the right metrics in place. Still, to adapt, optimize predictions and target high-value users, machine learning algorithms (just as it’s stated in its name) need to learn by testing different traffic sources and user segments. Generally this stage takes a couple of weeks.
For recently launched apps this stage can have different goals. Instead of targeting high performing users it will likely aim for wider reach. This helps acquire as many users as possible, providing sustainable data on user behaviour so that you can properly determine desirable traffic costs.
So, do you stick with one pricing model throughout all your campaign? Or do you have other options at different stages?
Transitioning to a New UA Pricing Model
Gradually shifting from one model to the other can be a logical step in scaling your app. As you are growing your audience and get more data insights.
For example, say you know that CPA is a conventional choice for your app type. But you don’t have enough data to make CPA campaigns profitable and don’t want to risk your UA budget. So, using the CPA model might not be very cost effective.
Let’s take a recently launched app as an example one that doesn’t have a sizable user base or historical data for analytics. So your priority is to collect information on users, have stable retention rates and test monetization. As it’s still too early to talk about LTV predictions, you want to find out how much you are ready to pay for an install with the help of retention and ARPPU data, while continuing to build your user base. The CPI model can work just fine for this purpose.
Switching to CPA Model
Once you can understand and predict user behavior, with CPI you might start losing profit over time. At this stage you can already find out the desired target action and its cost, increasing the possibility that CPA could be your most rewarding pricing model.
For example, you’ve recently launched a mid-core game and you need to get more traffic to clearly see the game’s performance and make adjustments. CPI is a good solution here. It gives you a bigger volume of traffic at a smaller cost. Once you see that your essential metrics are good and your users are making in-app purchases, you can figure out your ARPPU.
The CPI model brings you a lot of traffic, but the majority aren’t paying users. You already have a clear vision who your paying users are and how much you’re ready to pay to acquire them. At this point, it makes a lot of sense to switch from CPI to CPA, providing the opportunity to pay only for paying users.
Moving from CPA to Revenue Share
So you’ve successfully applied all the steps above and have a prosperous app with its own loyal user-base, stable user acquisition channels and sophisticated analytics. Now you face the obstacles of scaling.
For well established apps, target action stops being the main UA campaign goal. Acquiring whales becomes a much more desired goal. If you use a CPA model, you pay for all converting traffic regardless of the users LTV you end up with. CPA proves to be profitable, but is still potentially wasteful since paying users are no longer the main target.
In this case, a revenue sharing model might provide more profit potential than CPA. Even though the cost of the traffic can be seemingly higher. How? Quite simple. Revshare is tightly connected to the actual LTV of each user. Non-paying users cost you nothing, you’ll be paying only for those who are actually spending money in your app.
So, the shift from CPA to a revenue share model seems to be a very logical one. And a perfect fit for apps with big numbers of active users and in need of new traffic. Such apps process a lot of payments from a large number of paying users, and have good install-to-payment ratios and LTV metrics.
Revshare targets whales and is risk-free for a publisher. The only sensitive point is trust. For it to work publisher needs to share the info on all the payment that users make in the app. In order to reach this level of trust it makes sense to start with a CPA model during your learning stage and then switch to revenue share.
For revshare model to work, a publisher should have a good understanding of its audience and an exact idea of what their user acquisition goals are.
How Should You Choose a Model
To set up a UA campaign that won’t end in disaster, you need to understand how well you know your audience and its behavior in your app. You also want to make sure that you’re clear on what to expect from the learning and optimization stage. Then simply decide what goal you want to pursue with your user acquisition campaigns.
- Do you want to test your app and acquire as many users as possible on a certain budget?
- Or you already know your audience behavior, can predict it and know the amount of money you’re ready to spend for a specific targeted action.
- Or you’ve passed that point and are focused on finding whales to propel your app business’ growth.
Once you’ve decided on your goals and know your UA process, the best model and when to change should be clear. Just remember that if you aren’t losing money, that doesn’t mean that you aren’t losing profit.
What’s Next?
We hope this blog post has shown you a few ways to conduct user acquisition campaigns depending on your goals, the stage of your app’s lifecycle and the amount of data you have on your users.
Machine learning UA campaigns were designed for you to unlock more users with high ROAS metrics. But predictions need some time to learn and optimize, to be able to show their top performance.
That is why we’ve created AppGrowth, a solution that specializes in acquiring high quality users. It lets you benefit from a number of pricing models, like CPA, revshare and CPI and uses machine learning algorithms to ensure that you’re making the best possible use of your UA budget. Get in touch with us if you’d like to see our machine learning UA in action.
App growth isn’t possible without good old business intelligence. Your business needs as much user insight as possible to scale in the most profitable way. With our DataCore solution we align your user acquisition and monetization data and provide actionable insights for your particular business.
