The price per unique paying customer drops 2.9X
Pixonic, a Russian mobile game developer, had a goal of increasing the number of paying users in Taiwan and Korea. To do this, they launched a machine-learning-based campaign using Universal App Campaigns. In the Taiwanese market, this helped increase conversion into paying users by 85%, and in the Korean market, it helped reduce the cost per unique paying customer by 48%. Read more about the methods Pixonic used and other results in this article.
- Increase number of paying users in narrow markets where campaigns don’t receive many impressions and attract few new clients
- Used Universal App Campaigns and a predicted events algorithm
- Taiwan: conversion from installs into paying users increased by 85%
- Korea: acquisition cost per unique paying customer fell by 48%
In October 2017 Google modified a mechanism of mobile app promotion. Universal App Campaigns (UAC) became the main solution for app developers. This product uses machine learning to analyse a large number of ad targeting configurations in real time, optimise them and find the most relevant audience for any given business goal.
Universal App Campaigns allow advertisers to access the relevant audience of most main Google products: Search, Google Play, YouTube, and the Google Display Network – all in a single campaign. According to Google internal data from June 2017, acquisition managers who use UAC in their campaigns receive 140% more conversions per every dollar than their counterparts who use other mobile app promotion products.
Unlike most AdWords campaigns, in UAC most ads are created automatically. To receive ready-to-publish ads in different formats, just upload text, pictures and YouTube videos. The system then tests different configurations of ad messages and optimises the bids to maximise installs or other target actions, such as sales, reaching a level in a game, orders, signups and so on. For setting goals, there are different UAC types: UAC Installs (for installs acquisition) and UAC Actions (for target actions acquisition).
While UAC has been shown to be highly effective, difficulties may occur: either conversion is not high enough, or there may be delays. These problems can be solved with event prediction at early stages, such as the payment stage. Afterwards, that prediction can be used as a conversion to base the optimisation on. Currently, this algorithm isn’t automated and must be created for each project individually. Predicted events allow high-probability predictions of which users will perform the target action in the first days after installing the app.
The predicted events algorithm consists of three stages:
- Creating the prediction model. To do this, we need to receive historic data from a partner, choose a method and then develop and program the prediction model.
- Automating daily data streaming. This requires us to prepare a data import from the client, complete the event prediction and then send the prediction information back to the client.
- Sending conversions to AdWords.
Pixonic tested predicted events by advertising in narrow markets in Korea and Taiwan. Before that, campaigns in those countries tended to receive few impressions and acquired few new users. Here’s how UAC performed:
- Cost per purchase fell by 3.3X
- Cost per unique paying customer dropped by 2.9X
- Installs increased by 85%
- Cost per purchase fell by 12%
- Cost per unique paying customer went down by 48%
- Installs increased by 43%
Pixonic’s success story shows that the predicted events algorithm – combining client data about in-app user behaviour and Google data – leads to high results, increases campaign metrics and attracts new users. Currently, Google is working to automate the algorithm and scale its use.
Pixonic is a Russian mobile game developer with additional offices in Berlin and Cyprus. Its most profitable project is the mobile shooter game War Robots, awarded one of the best Android games in 2016 by Google. In 2015 Pixonic was named one of the top ten most profitable mobile developers in the CIS region by AppAnnie.