33Across: Five Billion Daily Events, One Near-Real-Time Dashboard

2020-12-30

33Across is a tech company that assists their clients in capturing consumer attention in the era of automated advertising. 33Across's programmatic ad platform helps premier publishers like Warner Bros. and USA Today place less intrusive ads, optimize yield, and gain actionable insights.

We worked with 33Across to provide these publishers with state-of-the-art dashboards that aggregate raw key performance data from five billion daily ad-serving events, performing calculations within half a second.

Share:

Service:

  • Data Integration, Analytics, and Activation

Got a project?

Harness the power of your data with the help of our tailored data-centric expertise.

Contact us

Problem

The 33Across platform helps publishers make more money from their ad space by connecting with a variety of supply-side platforms (SSPs). However, these third-party SSPs weren’t driving as much revenue as they could. 33Across needed a way to connect to demand sources directly using their own SSP in order to increase profits from 40–50 percent to 60 percent, improve user experience for client audiences, and cut the costs of third-party SSPs.

ილუსტრაციები-1.jpg

Solution

Publisher Manager, one of our solutions, had many of the features 33Across was looking for to solve its SSP challenges. We built 33Across an in-house SSP by integrating Publisher Manager into their existing infrastructure and customizing it to match their unique needs. We also built a creative audit tool to guarantee ad quality without impeding auction speed.

As we worked, we faced considerable data integration challenges, dealing with peak traffic of 20 million events per minute, which meant that reports sometimes had to process five billion rows of data. To speed up data visualization, we built a data aggregation pipeline. Within this pipeline, the most popular reports had their own custom, precalculated aggregates that contained only their own dimensions, while custom reports could tap into generalized data tables of all possible dimensions. This pipeline was deployed into a cluster of sharded databases. As a result, approximately 200 million rows per shard could be queried and visualized in less than half a second.

Carousel 2.png
Carousel 6.png
Carousel 3.png
Carousel 4.png
Carousel 5.png

Results

33Across saw 8 percent better margins on ad placement sales compared to bids through Google AdX, AppNexus, Criteo, and other networks. Our solution also helped 33Across increase effective cost per mille (eCPM) by 14 percent in just a few months. Additionally, report generation went from taking minutes to less than a second. As a result, 33Across can now provide publishers with information on performance in near real time.

ილუსტრაციები-4.png

Tech Stack

NodeBackbone.jsSpring BootReactJavaGoogle CloudPythonAnsible