Request a call
  • Hidden

Business Overview

Machine Learning (6)

Business Overview

The client has a mobile app retargeting product and data management platform for global brands. The product provides a holistic marketing funnel to retain, retarget and re-engage mobile app users, increasing conversion rates and delivering a personalized customer experience. The app retargeting product is built on machine learning (ML) algorithm, and for automation, integrated with all major attribution partners like Adjust, Appsflyer, Cochava, Singular.



The NIX team processed massive datasets from global media auctions that sell mobile advertising services and user data. The data contains information about the users—the applications they use, how often and when, their location, what advertisements they usually click on, what purchases they make, etc.

Our goal was to predict user behavior—specifically the user’s reaction to advertising before buying ads on the exchange and calculate profitability—thereby reducing expenses spent on advertising and better target audience. 

Machine Learning (7)

The NIX team used the LightGBM ML algorithm and feature engineering technique to audit data using market segmentation through cluster modeling.


  • 1

    Retrieve data from media auctions and group it by types of users and their activities during the last month

  • 2

    Calculate the average app open rate and in-app activities per user

  • 3

    Sort by app categories and user locations (games, business, utilities, etc.)

  • 4

    Group by installed applications in recent months

  • 5

    Sort by the time that has passed since the last purchase in this app or opening the app

The user opens the app, the request is sent to the app’s server, then to the platform with auctions, bidding, and predictions, and the winning ad is shown to the user.


On average, we analyze about 4 million user records per advertising campaign. Then we match this information with the bid at the auction and run a predictive model.

Data Management and Reporting Optimization

NIX data engineers are also engaged in maintaining and improving data management. Since the ML bidding system operates using large amounts of data, it was essential to set up an efficient data management process to streamline financial reporting and various metrics monitoring.

We also integrated the client’s system with supply-side platforms (SSP) and optimized the management of existing third-party applications.

As a result, the NIX team’s scope included the following activities:

  • 01

    Data aggregation

    Data was gathered from two main sources—Apache Kafka and S3—where it was aggregated into datasets using Athena and then transferred to MySQL, DynamoDB, and Redis databases. After that, the data was used as a foundation to enhance the effectiveness of bid request processing.

  • 02

    Data monitoring

    We used Kibana as the data monitor for logs from Elasticsearch and Apache Druid to assure the robust performance of systems for tracking both new and historical data.

  • 03

    Automated reporting

    We implemented the ability to create regular automatic and on-demand reports and added support for some external APIs to generate new reports for the end users and send them via email or Slack. We also used Redash for reports on historical data and Athena for ad-hoc reports and updated existing dashboards.

  • 04

    Custom module for handling errors

    We created a custom Scala-based module to automatically handle business logic related errors with Kafka and prevent financial loss.

  • 05

    Bidding process improvement

    We optimized the core logic of the bidding process using Python, Athena and MySQL, so now the system provides more data-driven predictions.


Machine Learning (8)


The NIX team delivered an ML-powered real-time bidding system that can process millions of data simultaneously. Moreover, we also streamlined data management, improved bidding process and automatic report generation.

This upgrade empowered the client with new ways to save up to 47% on knowingly non-performing advertisements, shift budgets to more promising campaigns and make better predictions.


8 experts (Project Manager, 3 Data Scientists, 2 Angular Developers, 2 Data Engineers)

Tech Stack:

Python, Spark MLlib, Scala, Elasticsearch, Redash, Docker, Angular, MySQL, Redis, LightGBM, DynamoDB, Kibana, Apache Druid

Contact Us