By now, most people have noticed how a specific advertisement can follow them from one website to another. It’s a strange feeling: the first time you’re “remarketed” to. The remarketing ad is based on something you looked at on a previous website, but now it’s following you on other sites trying to get you to buy it.
How is this happening? It’s simple: You have a cookie in your browser.
The developers of the online shop placed JavaScript code (or an image on the site URL) in your browser that requests an advertisement management system show you something specific. The cookie identifies you for that system. It acts as your personal marker. When you visit another website connected to the same advertising network, that website identifies you by that cookie and makes a decision to show you an advertisement, too—something that corresponds to the interests you showed on previous sites.
This process is called remarketing (or, “retargeting”). But beyond particular individual sales and identifying cookies, the value of remarketing might have much higher potential in terms of how a DMP system work.
So what exactly is a DMP?
DMP stands for Data Management Platform. Imagine that instead of a cookie from an advertising network, the code of a DMP from the shop who’s site you visited was put in your browser. And all other sites you’ve visited also used that DMP code. All the information would accumulate, rather than remain specific to one cookie.
For example:
- Say you visited a sports website and looked at bicycle accessories.
- Then you visited a shoe website and searched for women’s shoes, size 35.
- Finally, you read an article on a forum for young mothers.
This entire sequence of selections is collected by a DMP, classified by demographics and analyzed to identify your likely personality type. DMPs could deduce the person looking at bicycles, shoes and mother articles is:
- Female
- Interested in sports
- Foot size 34—36
- Age 25—35
“Wait a minute!” you might say.
“Those general categories could be applied to any one who surfed those sites. But not all of those people will match those demographics.”
And you’re partly right, but with DMP data, we’re dealing in the accuracy of assumptions. And believe me, as the sequence of selections on the list grows beyond three sites—to four, five, ten, twenty—the accuracy of the assumptions gets greater and greater, exponentially.
It’s scary to imagine, but when DMP systems operate on social networks, imagine the full picture they can paint using only statistical assumptions. After all, any social media account will likely have much more than three to six data points contained within it. And sometimes the information is extremely specific, personal, and unique to individual users.
So why do DMP systems need that personal information?
DMPs use it to help other systems explore specific user groups for data trends. For example, advertising networks can use data provided by DMPs to help advertisers find their target audience. For big sites and big systems, it’s an opportunity to analyze user groups interested in content in order to make the right decisions about what other content to show.
And Demand-Side Platforms (DSP) have a special interest in this data, because they don’t have publishers like advertisement networks do, and they don’t get straightforward requests from browsers, either. DSPs get requests from SSP systems, which don’t have much information about specific visitors. The DSP gets all kinds of user data from a DMP, which helps make a decision on prices and bids at ad auctions.
Orbit DSP works the same way. It uses DMPs like Aidata and VisualDNA—and soon it will be integrated with one of the top players in the market, BlueKai. This will create better and better pricing for our advertisers, plain and simple.
Big Brother
Beyond using data for commerce and bidding, society must decide about the limits and rights of individual privacy in our online lives.
But it’s not really “big brother” watching us. Instead, we create data profiles ourselves—and on social networks, we do it very willingly. Those profiles are based on a list of choices and assumptions that can be figured out, not by individual surveillance, but by statistical sorting.
In a counter-intuitive way, we identify ourselves through a series of choices that can only be unique to us, not by individual observation of our individual actions.