This installment of "Genome Decoded" looks at how Genome's unique data management platform analyzes Big Data to fine-tune campaigns
Editor's Note: "Genome Decoded" is a series that tells the story of Genome, our data-driven audience-buying solution, from many angles, to bring you a little closer to Big Data.
When you hear the term "taxonomy," do you immediately want to call your accountant? Does "lift analysis" make you think of wait times on the ski slopes? Never fear---we're here to help simplify the complexity behind the Big-Data-busting power of Genome's sophisticated technology.
Big Data can overwhelm marketers, but they know it holds nuggets of information that can help optimize campaign performance. How do you find them? Genome's proprietary software brain, called the Open Segment Management (OSM) platform, was built to solve that challenge in two basic ways:
- Analyzing the largest amount of data from the widest range of sources possible, and
- Doing it in real time, since campaigns and the conditions that impact them are always changing.
The Ad blog talked to one of OSM's developers, Jason Lynn, VP of Product, Genome from Yahoo!, to find out more about the technology that powers Genome.
Yahoo! Ad Blog: How would you describe OSM?
Jason Lynn: OSM is the data "brain" that helps us deliver data-driven ad solutions to our clients. It's a very big, very fast data management and analytics platform that has some unique features specially designed for online campaign targeting.
For example, with 40 terabytes of RAM and growing, OSM is one of the world's largest commercial in-memory databases. Its architecture is designed to leverage a very large and diverse data set against campaigns very quickly. In many cases, an hour can make a big difference in campaign outcomes, and our platforms are built to deliver answers at that kind of speed.
Another example is the OSM's data taxonomy, which is designed to organize a continuously growing stream of data sources, types and uses---and do it efficiently. It currently has 1.5 million nodes, or unique data facts, such as "BlueKai Sedan Auto Shopper" or "Yahoo! Sports Enthusiast," which can be leveraged against any of our clients' campaigns.
YAB: Is there anything else like it in the industry?
JL: There are systems that have some of the same functions and use some of the same algorithmic approaches as OSM. But OSM has other parts that no other system has, like our information taxonomy, and no other system puts all these things together in the same platform like we do.
YAB: Where does OSM get all that data to optimize campaigns?
JL: It comes from a wide variety of sources. An enormous amount comes from Yahoo!, including data on audiences, their habits and preferences, media and campaign performance, and a whole host of other online metrics. More information comes from our relationships with over 25 top data providers, our ad partnership with AOL and Microsoft, and client marketing information, such as website or customer file data.
OSM allows us to leverage all that data against every campaign that we manage for our clients. It can actually go through all those 1.5 million nodes to determine the best-performing combination of data, media, and creatives for any given campaign.
It's important to remember that there's no "silver bullet" for maximizing campaign performance. Each campaign is unique and has different types of inventory, creatives and data that will work best for it---and all those factors are constantly changing. You have to have a platform that can test various combinations simultaneously. OSM collects data, keeps it organized and continually performs lift analysis to find out what's working—and what isn't.
YAB: I'm just a simple blogger—what's "lift analysis," and why is it important?
JL: Lift analysis is our technique for choosing which data to use in a campaign. There are literally tens of thousands of data choices, so how do you determine which data to use (and pay for)?
From a cost or time standpoint, it's impractical to test tens of thousands of facts, so we use lift analysis to make this decision. We do an individual lift analysis for each campaign using every data fact we have in OSM (with some exceptions for privacy and ownership considerations).
Our goal is to not miss any opportunity to use data to maximize a campaign's performance. If you do lift analysis correctly, you know which data will positively impact a campaign. OSM allows us to do that.
YAB: We often talk about Genome's ability to develop campaigns that scale---does OSM help here too?
JL: There's really only one solution to creating campaigns that scale, and that's to identify, analyze, and leverage as much targeting data from as many sources as possible. That's what OSM was built to do—and it continually tracks campaigns to check which data is working for a campaign, no matter what stage it's in, to ensure it's actually making a difference for our clients.