1. In-house Adserver
    1. Object oriented model of traffic management for real time decision making
    2. Pretargeting configuration of traffic flows
    3. Advanced post campaign reporting and analytics
    4. Display, Video, Mobile Web, Mobile In-App, Native, Native Video, SmartTV
    5. Can offer ad
    6. Each ad request triggers auction with both internal and external bidders SSP

  2. In-house DMP
    1. Proprietary taxonomy of segments
    2. Target segments are formed based on both own and client data
    3. Cluster analysis is performed on elements of taxonomy first party data to reveal marketing insights
    4. Segments can be used for targeting in DSP solution as well as pushed to customer
    5. Real time profiling of third party data
    6. Look alike segments generation

  3. AdExchange / SSP
    1. Setup configurations ranging from direct orders to multi level auction
    2. Most of RTB protocols are supported for communication with bidders
    3. Header Bidding is supported
    4. Bid requests can be enriched by first or third party data

  4. DSP / bidder
    1. ML optimisation towards conversion of traffic
    2. Operations of advertising campaigns are fully automated
    3. Supports Display, Video, Mobile Web, Mobile In-App, Native, Native Video, SmartTV
    4. Embedded real-time decision trees
    5. Bidding strategies for 1-st and 2-nd price auction, direct deals, PMP
    6. Bid rate is calculated in real time based on conversion predictor
    7. 2019Q4 average bid-rate =9.7%, win-rate = 11,2% (400B bidreq served)

  5. Verification
    1. Viewability specification — IAB/MRC
    2. Fraud detection: SIVT & GIVT
    3. no MRC certification
    4. Integrated with MRC certified verifiers — DoubleVerify and Adloox

  6. Infrastructure, architecture, performance
    1. 80 bare metal servers, 120K qps scalable to 300K qps.
    2. System architecture supports working from multiple data centers, allowing for geographically distributed system to serve local users
    3. C++, proprietary components from frontend to backend, in-memory databases.

  7. ML (AI) infrastructure
    1. Snapshot of each event consists of 1000 features out of 100K total available features
    2. Models are trained on data from last 4B events
    3. Architecture supports any ML methods, including hybrids (ie decision trees, logistic regression, naive bayes, etc)
    4. Trained model is used in real time and needs < 40ms per bid request
    5. Each bid request results in XXX calculations of primitives on average

  8. Clients
    1. International advertising groups https://www.adriver.ru/agency/
    2. Supported exchanges, RTB partners https://www.adriver.ru/pro/dmp/
    3. International brands

  9. About company
    1. Contacts https://www.adriver.ru/about/contacts/