Vizury’s Technology Platform Overview
We are in the business of behavioral advertising and show personalized advertisements (ads) to users who drop off from e-commerce websites across other websites (publisher) on the internet. These personalized ads help our clients (advertisers) get the drop-offs back on their website. For the uninitiated, this is Retargeting. As a result, three primary entities interact with Vizury’s platform – the advertisers, the publishers, and the end users (via advertiser and publisher websites).
Publisher interface: Whenever our publisher has inventory for us, they insert our ad tag on their web pages. When a user visits such a publisher website the result is a call to our ad server. Our ad server looks at the user’s behavior data and identifies products to be shown in the ad using proprietary recommendation algorithms. Our ad server creates a personalized ad by combining the results of the recommendation algorithm and the feed data (for the price, image etc) and is served on the user’s browser as a flash or html5 banner. When the user clicks on the banner, the browser is redirected to the relevant product page on the advertiser’s website. The predominant inventories used for showing ads are Real Time Bidding (RTB) and non-RTB. The publisher allocates non-RTB inventory for Vizury ads. In RTB, publishers ask for bids for the inventory from various DSPs in real time and the winner gets to display an ad.
Data and analytics: There is a huge amount of logs generated by different components of Vizury’s platform and we have two components on our analytics platform that handle this data. The first one is an OLAP reporting system used by different stakeholders to analyze the performance of the campaigns and contains standard reports. The second is an ad-hoc querying system, where one can perform any query on the generated log files and is used to find various insights that are fed back to algorithms and operations to improve the campaign performance.
What are the technological requirements and the challenges involved?
Vizury’s platform handles hundreds of millions of user data and receives billions of ad requests every day. The analytics system has to process multiple terabytes of data every day. Creating systems that can scale to those numbers is both interesting and challenging. The scale has to be thought through on every feature that gets implemented. Performance (or latency) of the systems are also of utmost importance. All the end user related processes have to be completed in single digit milliseconds so that the user experience on the advertiser and publisher websites are not affected.
To run a service at internet scale, a very reliable and efficient infrastructure is a necessity. On the technical side, it should be able to support the scalability and performance. The end user experience will depend on both the processing at the servers and the network latency from the user to the servers. All RTB publishers impose a time limit of 100ms between the sending the bid request and receiving the response. Hence, the infrastructure has to be designed with these constraints in mind. On the financial side, there are costs involved for servers, storage, and bandwidth. Any feature designs have to consider the costs into account.
Through this blog, I have made an attempt to present an overview of the technology and challenges involved in Vizury’s retargeting. In future blogs, we will look at some of these components in detail, their technical challenges, and talk about some of the technologies that we use to address those challenges.
Keep watching this space. Say hello to us at firstname.lastname@example.org