Blog  Engineering  

Shining Light into the Black Box

by Ric Ren, November 22, 2017

Imagine you are a campaign manager who wants to promote holiday sales on your brand’s website. After spending weeks researching marketing and advertising trends, you finally launch and are eagerly awaiting  for visits, quotes, and purchases to start rolling in. After a few quiet days, you’re disappointed to realize that no one new is coming to your website. You add some new content and make a few adjustments, but soon the holiday season is halfway over and your campaign still isn’t delivering. It becomes apparent that you will need some help in the form of targeted market research and advanced insights in order to turn your failing campaign around.

Amobee’s platform solutions are specifically designed to efficiently solve problems identical to this. We know that the digital marketing industry is constantly evolving, and that’s why our platforms do the same. Our RTB platform contains a highly complex distributed system consisting of a comprehensive targeting pipeline, sophisticated bidding algorithms, and a real-time budget control mechanism. It makes flash decisions on tremendous amounts of bid requests according to specifications from campaign managers. This means that for that holiday campaign that just isn’t performing, Amobee’s platform solutions are able to provide instant feedback on what’s working and what isn’t, so campaign managers can adjust their tactics accordingly.

However, because of the complex system and highly dynamic nature of RTB, it can be difficult to provide realtime insights and sufficient feedback. When troubleshooting issues or trying to improve spend and performance, campaign managers didn’t always have an easy way to get logical guidance. Amobee realized that this optimization was inefficient and not scalable. That’s when we started our latest project.

Once our technical solutions team was able to get a thorough understanding of why campaign feedback wasn’t providing the insights it should be, we started our project to bring campaign managers the insights they want to know, exactly when they want to know them. Specifically, we wanted to reveal a few things:

  • First, we wanted to quantify how restrictive a campaign’s targeting is. To do this, we measured the percentage of bid requests that qualify for each targeting rule, and the percentage of bid requests that are discarded.

  • Second, we needed to expose how the algorithm evaluated bid requests. We decoded the following question: Is the algorithm still learning and collecting performance data, or is it bidding with confidence?

  • Third, we had to capture any changes that happen to a campaign. For example, if inventory traffic suddenly changes or a campaign’s bid price rises or falls over time, we wanted to capture fluctuation or trends with time series data.

Acquiring the answers to these questions wasn’t easy. In order to achieve our goals, we had to implement a data collection framework capable of collecting time series data 24/7 from the RTB platform.

Here is how that data collection framework works: First, data is extracted from various components, including the filtering pipeline and bidding algorithms mentioned above. Since it isn’t practical to store 4M QPS data at the event level, data is then aggregated locally on each component into hourly means.

Next, a centralized server collects hourly aggregations from various components and joins them together. The server then inserts the combined data into Amobee’s reporting platform to benefit from its huge data storage and fast query speed. Lastly, a newly constructed user-friendly UI tool helps users retrieve the diagnosis data from a specific campaign off the query server. All a user would need to do is simply provide a campaign ID and click a few buttons. Within seconds, the campaign diagnostic tool will retrieve data from the reporting query server and export it into hourly or daily view charts. The instant insights provided are invaluable.

As an added benefit, reporting APIs support integrations with the console and algorithms. In another project, we seamlessly integrated targeting exclusion reports into the console campaign intelligence dashboard for self-service debugging. Our team was also able to develop more advanced algorithms based on these new insights. The platform diagram, found below, helps conceptualize this process.

Since the release of the campaign diagnostic tool last year, Amobee has seen very positive outcomes. Campaign optimization on the Amobee Platform is now more transparent, more data driven, and most importantly, more efficient than ever before.

 


 

Read Next

All Blog Posts
Perspectives

Beyond the Box: How Programmatic Video Measures Up

More and more consumers in every demographic are watching video online, which means every brand campaign has something to gain from well-executed programmatic strategies. Here are a few ground rules for putting programmatic video to work for your brand.

September 28, 2015

Engineering

Amobee Sponsors VerveCon 2018 Tech Conference, Supporting the Advancement of Women in Technology & Engineering

To support empowering women in technology and engineering, Amobee is proud to sponsor the inaugural VerveCon 2018 Tech Conference in Santa Clara on February 9th. VerveCon is a convention for women in…

February 2, 2018

Perspectives

How Viewability Fits in a Successful Campaign Strategy

It seems like a total no-brainer: No one wants to pay for ads that can’t be seen. But viewability is just one factor in an effective campaign. Focus too narrowly on it, and you can compromise targeted reach, conversions and brand lift, and you risk overspending on a parameter that’s not generating nearly enough value.

April 14, 2017