USER IDENTIFICATION PROBLEMS ON DSP SIDE IN TERMS OF ADVERTISING RTB - PowerPoint PPT Presentation

USER IDENTIFICATION PROBLEMS ON DSP SIDE IN TERMS OF ADVERTISING RTB AUCTIONS B. F. Trofimov, А. V. Arsiriy, Е. А. Аrsiriy

AGENDA ● Introduction ● Informational technology of AD delivery for RTB ● User identification on DSP side ● Profiles fragmentation in DSP Database ● Conclusions

INTRODUCTION ● Comparing to the previous year, the RTB market growth was 36% ● according to PRNewswire, RTB market will hit $42 Billion by 2018 ● Primary complexity and problems of advertising domain

User identification on DSP side

Retargeting and Cookie Origins Retargeting dsp_cookie RTB Auction ssp_cookie retargeting , is a signal from third-party partner web pages, that a specific end-user has a specific ● interest. impression , is a signal that an end-user has seen Ad; ● click , is a signal that an end-user has clicked on Ad; ● conversion , is a signal that an end-user has performed some highly expected activity, for instance, ● he bought promoted goods;

INFORMATIONAL TECHNOLOGY OF AD DELIVERY IN TERMS OF RTB AUCTIONS Strong isolation between advertisers and publishers ● Well-defined interaction protocol between DSP and SSP ●

USER IDENTIFICATION Piggybacking – async JS redirect to all DSP § servers with ssp_cookie as a query param Chance for DSP to link ssp_cookie_id with own § dsp_cookie ( retargeting again) Publisher Content Ad Server All bidders should Ad Server receives delivered sends signal to send their decision shows page that request to user Ad Exchange (participate? & redirects to the DSP ● price) back winner server makes piggybacking ● 80 ms 20 ms 100 150 170 200 210 280 300 350 400 1sec >>1 independent companies take part in this auction Publisher Site sends SSP (Ad Exchange) SSP chooses the User’s web page sends request to receives ad request and winning DSP shows ad banner ● response Ad Server opens RTB Auction and sends this with ad back toshow ad info back to Ad firing DSP’s 1x1 ● Server pixel (impression) Every bidder/DSP receives info about user: ~70% users have ssp_cookie_id ● this cookie aboard geo data ● thankfully to retargeting site url ●

PERFORMANCE

Profiles fragmentation in DSP Database

PROFILE FRAGMENTATION IN DSP DATABASE PROBLEM: Often neighboring profiles are the same user PROBLEM: Often neighboring profiles are the same user GOAL: Build complete person profile based on bridging information from different sources: GOAL: Build complete person profile based on bridging information from different sources: § § § § Mobile Mobile TV TV § § § § Desktop Desktop Social Social WiFi Location X WiFi Location X WiFi Location Y WiFi Location Y Time 10:00 am Time 10:00 am Time 8:00 pm Time 8:00 pm

GRAPH-BASED DEFINITION General task definition: § Build graph § Vertexes – user’s interests § Edges – bridging rules [cookies, IP,…] § Task – Identify connected components

CONCLUSIONS ● User identification at RTB auction ● Multiple profiles problem and business impact ● Development of models and methods of profile bridging based on intelligent data analysis (Data Mining and Machine Learning) allows to improve accuracy and pace of decision making by DSP. ●