【Prediction】How Payment Prediction Can Help Make Your Operations Ultra-Precise - Huawei Developers

1. Why do I need to predict user payment potential?
Payment prediction refers to using machine learning to determine audiences that demonstrate low, medium, and high payment potential among active users from the previous week. These predictions are based on in-app user behavior and user attributes.
Successful products are those that consistently garner revenue, and thus, boosting revenue is a major priority for any product operations team. But in an era with a fully mature Internet, it can be costly to acquire new users. Therefore, extracting value from existing users is crucial for achieving sustainable success. For audiences generated by payment prediction, you can formulate targeted marketing strategies that address relevant users, for example, offering discounted plans, to boost payment conversions.
2. How can I encourage user payments?
Due to the high cost of acquiring new users, maximizing the value of all existing users throughout the entire lifecycle has been a major priority for enterprises, as well as guiding users toward paying for products or services. A user who is more willing to pay is also more loyal to your products. Loyal users can help you find more potential users, setting a virtuous cycle in motion.
A broad range of apps have pursued in-depth product operations strategies, in order to determine the conversion model that best suits them. This is especially true for e-commerce and game apps.
(1) E-commerce apps
In order to encourage reluctant users to make payments, e-commerce apps often launch promotional activities, such as time-limited discounts and flash sales that appear and vanish in an instant.
(2) Games
Most payment conversions for games are from item purchases. In addition to providing item purchase preferences, games can also push payment-relevant notifications at key moments to encourage payment conversions.
3. How can I leverage payment prediction to grow my revenue?
Profit models vary widely depending on the app category. This article will use game apps as an example to show you how payment prediction can help boost product revenue.
As mentioned earlier, games earn most of their revenue from item purchases. Video ad clicks in games are also an important revenue source. We will talk about how payment prediction works under these two scenarios.
How does payment prediction work?
Simply put, payment prediction finds the various probabilities of players paying over the next week, based on their in-game behavior and attributes, such as the app version and number of sessions over the previous week.
The system predicts three audiences by default, those with high (>70%), medium ((20%, 70%]), and low (≤20%) probabilities of paying over the next week. Of course, you can also customize probability ranges based on the characteristics of your game and players. For example, you can define the medium-high probability range as 50% to 70%.
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}
* Payment prediction details page
After the audiences are generated, you can conduct refined operations using AppGallery Connect's grow services.
Focus on in-app revenue for the high-probability audience
Users in the high-probability audience are more likely to top up in games than any other audiences. Messages related to time-limited top-up discounts and limited items should be pushed to them. In these messages, keywords such as "discounts" can help encourage hesitant users to pay. Keywords such as "time-limited" and "limited" are also highly useful, as they create a sense of urgency, helping promote payment conversions.
Currently, the high-probability paying user audience generated by Prediction can be directly applied in HUAWEI Push Kit and App Messaging. For example, when a user views the details of a card drawing activity, you can use App Messaging to show a pop-up to notify this audience of a time-limited promotion for drawing rare items. Since the users in this audience are more willing to pay, the time-limited activity is likely to lead to more payments.
Push Kit can also be utilized to send notifications about top-up discounts to targeted users.
Message pushing by audience is quite easy to use and cost-effective. Audiences can be directly selected when messages are sent.
* Available audiences for Push Kit
Focus on ad click revenue for low-probability audiences
It is difficult to get low-probability users to pay, and thus it's necessary to adjust your profit model to extract value from them. The most common way to earn revenue from such users is through in-app ads.
Many game developers are concerned that an overabundance of ads can undermine the playing experience, and that revenue generated in this manner may even fail to compensate for lost users who are turned off by the number of ads. Consider the following scenario: A user is about to complete a game level, but finds that they are about to run out of credits. This user is unwilling to purchase an item to extend the game, but when he or she sees a message that offers credits in return for viewing a 15s ad, they are likely to be enthusiastic rather than annoyed. This is correct way to place ads in games.
Ad delivery should take player psychology into account. Users with low payment probabilities should be encouraged to watch rewarded ads with keywords such as "gold coins" and "credits", to boost the ad click-through rate and earn more revenue.
You can apply the low-probability audience as a filtering condition in Remote Configuration to display specific ads only to this audience, without affecting the gaming experience of your other users.
* Prediction filtering conditions page in Remote Configuration
In addition to the usage cases described above, you can also leverage the payment prediction function to formulate more targeted and versatile operations strategies that meet your full range of needs.
Welcome to try out our Prediction service. For more information, please visit our website.
To learn more about DTM, click here.
For more details, you can go to:
Reddit to join our developer discussion
GitHub to download demos and sample codes
Stack Overflow to solve any integration problems
Original Source