by Devin Guan, VP Engineering, Drawbridge
Advertising platforms are built to bring messages to the right people at the right time, and each platform has unique advantages. Cross-device platforms leverage a database of paired devices and insights to target audiences based on behavior across smartphones, tablets and desktops. Obviously marketers greatly prefer to run targeted campaigns, but the question of how the targeting is actually done remains a mystery for many of the players involved in launching a targeted campaign.
Let’s say I want to run an ad campaign for my fictional golf store, Devin’s Golf Shop and my ideal audience is US-based males in their 30’s who like to golf. If we generously assume that 40% of devices where ads can be served are in the US, half the population is male, a quarter of the population in is their 30’s, and 1 in 50 people enjoy golf – simple probability will tell us that only one in every 1,000 ads displayed, will reach a US-based male in his 30’s who likes to golf:
P(US, male, 30-39, golfer) = P( US ) * P( male ) * P( 30 -39 ) * P( golfer )
P(US, male, 30-39, golfer) = .4 * .5 * .25 * .02 = .001
But if I run a smart, targeted campaign that leverages cross-device data, I can reach the right audience with a high degree of certainty.
How does a targeting engine work?
Targeting can be approached much like an Internet search query. Search is a well-solved problem, so there’s no need to reinvent the wheel. Plus open-source software is readily available for search-related tools. When searching for something online, a user submits a query and the search engine finds the best result. In the case of targeting for my golf store campaign, the search query would be:
…or as a search query: q= +country:[us OR DNTC] +gender:[male OR DNTC] +min_age:[* TO 35 OR DNTC] +max_age:[35 TO * OR DNTC] +interest:[golf OR DNTC]&q.op=AND
In targeting, it’s essentially the same story as search, except inverted. Because ad campaigns cannot actively find users to show ads to, as ad requests (and the associated user data) come in, the targeting engine is searching out a campaign that matches. So if my ad campaign is targeting American males in their 30’s who like to golf, as soon as an ad request comes in from a user that matches those dimensions, the engine will find a campaign that’s looking for that type of user.
As these targeting engines get smarter, we can add more dimensions. If I wanted to, I could add filters to my campaign that block users that are already customers of Devin’s Golf Shop (“q= -customer:[ DevinsGolfShop ]”), block advertisements from my competitors on my site (“q= -block:[ MikesGolfShop ]”), or retarget users who have visited my site before (“q= +rt:[ devinsgolfshop.com ]”).
The single largest benefit of running a targeting engine like a search engine is the ability to scale. With the high traffic volume, massive amounts of data, and required response times in the nanoseconds, a rigid database solution would not suffice for these flexible queries. By mirroring a search engine, targeting engines can conduct complex but precise queries, such as the above example, that yield valuable audiences for agencies and advertisers.
Devin Guan joined Drawbridge in 2011 as a founding member of the engineering team, bringing over 15 years of experience as a leader, entrepreneur, and developer. Prior to Drawbridge, Devin was Chief Architect at Announce Media, where he built a technology platform that powered the business to secure $260M in annual revenue. Devin joined Announce Media from Netflix, where he was responsible for conducting Netflix Prize, a large-scale crowd-sourcing machine learning contest.
Like most Drawbridge employees, Devin had his heart set on mobile advertising early on; Devin was also as a founding member of AdMob’s optimization team. Devin has held various engineering leadership roles at prominent companies like Yahoo!, where he designed query expansion algorithms for Yahoo! Search Marketplace, and also oversaw the entire architectural design of Yahoo! Video and Yahoo! Video Search.