I mentioned yesterday my surprise at seeing an ad for the California Senate race -- specifically, an attack ad on Barbara Boxer from a "Protect Life" group -- when I was checking out a Spanish-language news site based in Paraguay. Plus an English-language ad for giant salamanders, left. The explanation for at least the Carly ad turned out to be that the Paraguayan ad-server detected that I was coming in from California and so delivered a relevant commercial.
A reader quickly responded that, in fact, most online users like me were true naifs about the extent of ever-evolving profiles of our online identities. He writes:
Around 8 years ago, I started a company in the Internet advertising industry. In the last 8 years, the one constant has been the growth in how much we know about you and all the other Internet surfers out there. It used to be that we could just map your ip address to a location and tell what time of the day it was. Then we started getting some demographic information like age and gender. These days there has been an explosion of data, to name just a few:
- Look alike data: Other web surfers with similar surfing habits like yours are known to perform activities A, B and C, therefore you are likely to do so too.
- Behavioral data: You recently researched a Honda Accord so perhaps we should show you ads for Toyota Camry
- Social data: You share these interests and therefore are perhaps interested in these products. Your friends share these interests, so perhaps you might be interested in them too.
- Offline data: Your credit score, your zipcode (of where you live rather than where you are at this moment), the wealth percentile of your household... Contextual data: Deep, real time contextual information about the page you are on at this moment.
All of this data is anonymous, so it's not a privacy risk (at least not yet), but it is scary how much data is out there. What is scarier is how much data is out there that nobody (at least not officially for moral/PR reasons) is using yet. Depending on the type of the data, this data is usually auctioned off to the highest bidder in real time or shipped to any one who asks for a fee.
Agencies then craft campaigns that pull together various combinations of these targeting criteria to decide whom to target with which ad and how. Companies like mine excel at using huge mounds of this data to model advertiser performance (dollars spent versus revenue generated) to ensure that every ad that is shown has the maximum impact on advertiser top line/bottom line.
My guess is that the industry has just got started.
I wrote back asking, But what about the giant salamander? And after the jump, the explanation -- along with another reader's experience with ads showing up in unexpected venues.