More and more online display inventory is being purchased via ad exchanges than ever before. On these platforms, advertisers utilize technology to bid on each and every ad impression in a real-time marketplace based on the value they put on the viewer. For example, retargeting has become a popular tactic as advertisers are able to tag (and then buy ads exposed to) users who have already shown interest in them by visiting their site. This technique has proven to be very effective, with higher than normal click-through rates (CTR), conversion rates, and other key performance indicators (KPIs).
Currently, more than 400 billion global monthly impressions are up for bid to online marketers -- that translates to about 150,000 ads each second during high internet traffic times!
Since the first banner ad was sold more than a decade and a half ago, it's no big surprise that publishers have aggressively sought ways to increase revenue by optimizing their ad inventory. Because so much of online inventory goes unsold or dropped into remnant channels for a micro-fraction of what can be made from presold, premium inventory, ad exchanges were an organic evolution in the ecosystem. A successful model was already in place with paid search where advertisers bid in a real-time, auction environment for ad impressions. As with paid search, exchange ad inventory is optimized by capitalist equilibrium -- some inventory is worth pennies, some is worth tens of dollars. Either way, ad price is determined by what the market is willing to pay.
Because most of the transaction is automated by technology, exchanges are very efficient for publishers to monetize previously unsold inventory without the need for robust sales teams, as well as for advertisers to buy direct inventory without middle men (such as ad networks) inflating costs. This has led some industry experts to predict that the market share of exchange vs. traditional online display buying methods will grow quickly in the next several years. In fact, this may have set off a revolution in which we will see more and more inventory moved to digital channels -- it might not be all available in real-time, but the efficiencies of this model cannot be denied. It's very possible that one day that any inventory that can be sold this way, will be sold this way... not just online, but TV, print, radio, etc.
The debates of the value of audience vs. context, technology vs. manual expertise, commoditization of inventory and its effect on our industry, etc., are already taking place in board rooms, industry conferences, and the blogosphere. As exciting as the new opportunity might be, there is resistance from the owners of the status quo, who will find their importance and market share drastically reduced by this evolution. As well, there is concern about the quality of this inventory and, as is always the case when targeting individual users, there will be privacy issues to take into account.
Regardless of the positive and negative context surrounding ad exchange buying, there still exists some mystery into how the technology actually works. Below is a top-level overview on how a publisher impression gets passed through the exchange value chain and ultimately gets served as an ad to the end user.
Note: This entire process happens in less than one-third of a second.
An online user makes it to a publisher's site via a link or direct URL typed into a browser. The page loads and swoosh...
The publisher ad server
...the site's ad server recognizes that an ad box is on the page that needs to be filled. Publishers have a variety of choices on where to buy inventory. They can have in-house sales teams that work to presell their best (premium) inventory, ad networks that agree to help sell the inventory (either on an exclusive or non-exclusive basis), and, of course, ad exchanges, where advertisers can bid, in real time, for the impression.
So the first-party ad server may put the ad impression up for bid on the exchanges directly or through...
The publisher's tools
...which can enable them to let the impression be handled by yield optimizers (such as Rubicon, Admeld, and PubMatic) that can help them maximize their site revenue. They can consult with optimization service teams to help set pricing, decide what kind of ad units should go on specific pages, make deals with ad networks and exchanges, etc. As well, these partners can offer propriety technology to facilitate and optimize ad sales.
If the publisher tool decides at this time that the best value for the impression is on an exchange, it will send the impression there.
Currently, there are only a handful of "major" ad exchanges:
- AdECN (Microsoft)
- DoubleClick Ad Exchange (Google)
- Right Media (Yahoo... currently testing real-time bidding)
These media entities have direct deals with publishers, networks, or publisher tools to sell inventory on their open platforms. On its DoubleClick Ad Exchange, for example, Google has migrated inventory from its very successful AdSense program, which enables it to sell advertising this way on literally billions of web pages. Exchanges negotiate rates with the media providers and get paid to simply handle the transaction.
Advertisers and their agencies can elect to "get a seat" on these exchanges in order to be involved in the bidding marketplace. Exchanges have self-service, back-end platforms that media buyers can log into, set up and manage campaigns, and run analytics reports to analyze and then optimize their accounts without ever talking to a sales person.
However, many advertisers elect to not work with exchanges directly, and opt to use a demand-side platform (DSP) as their trading desk of choice.
The demand-side platforms (DSPs)
Demand-side platforms have risen to help advertisers manage all of their exchange campaigns through a single interface. These DSPs utilize the ad exchange's data connection via application program interfaces (APIs), which provide the same functionality as working on the exchanges' own systems. Some agencies and holding companies have built their own DSPs, and many third-party DSP providers have popped up, such as Turn, Invite Media (Google), [x+1], MediaMath, DataXu, and Triggit.
DSPs provide more features than just letting advertisers manage all of their buys in one spot. By aggregating the entire exchange buying process into one system, advertisers can utilize the high-powered algorithms of their DSPs to optimize their bids. These algorithms can be tuned to target a specific cost-per-click, cost-per-sale, return on investment (ROI), or other business goals. They can also provide integration with various third-party data providers to allow for simple access to purchasing inventory. Another recognized benefit to running everything through a DSP is that advertisers can enable universal frequency capping across all of the exchanges they work with.
Because an ad needs to load quickly, DSPs have service level agreements (SLAs) that set how long they have to return a bid for the impression. The standard time frame for a DSP to return a bid request to an exchange is around 50 milliseconds, or one-twentieth of a second!
Bids are generally handled at the cost-per-thousand (CPM) level, and advertisers can choose to vary these bids based on the value of the users each placement is targeting. A common targeting choice is contextual, as exchanges classify their inventory by categories (i.e., sports, entertainment, news, finance, etc.) and sub-categories (entertainment - movie; entertainment - music; etc.). It's probably safe to say that an advertiser involved in promoting a new film would be willing to pay more for to users on movies sites than on financial ones.
Another common targeting option is geography -- by country, region, state, DMA, etc. And because you can mix targeting directives in a placement, you could, for example choose to create a campaign that just targets users on automotive-related sites in California and bid accordingly for that traffic.
It's important to note that advertisers are not manually bidding on every impression in real-time -- not even Superman could keep up with 150,000 ads per second. Rather, advertisers choose targeting settings and a maximum bid for what they would pay for an impression that meets those parameters. As with paid search, analysts can continuously optimize their choices by changing bids and targeting settings to ensure the best possible ROI.
The data layer
The piece of the puzzle that seems to most accentuate the value of exchange buying is the various data that can be leveraged in order to target users. By tagging user browsers via anonymous cookies, advertisers can retarget to their website's visitors on publisher sites by bidding more for those users than their competitors; reaching shopping cart abandoners with multiple ad impressions in the first few days after leaving the site has become a smart way to utilize this feature, for example.
Third-party data providers tag users and then sell advertisers access to identify those users later on an exchange. These companies (such as BlueKai, eXelate, and Experian) can provide user segments such as male or female, age bracket, income bracket, etc. How do they do this? There are various ways. There are data providers that tag users online who are registering product warrantees; they can provide exchange visibility to audiences based on the types of products they own. Other data providers make deals with travel aggregators so that advertisers can (anonymously) target users who have searched for Caribbean vacation deals or European hotels. Some data providers work with social networks that have registration data on millions of users and can help find audiences with specific niche interests.
The most valuable data seems to be that which can help identify in-market users. For example, a data provider signs a deal with a top-tier automotive review site to allow it to tag its users on key portions of the site. Days later, a major car manufacturer is able to buy impressions on other sites to target those same users who were looking for specific car-related information.
Ultimately, advertisers are going to pay more to reach the most desirable audiences to their business. Data empowers the advertiser to be able to better judge the value of each impression as it is passed to them.
The verification systems
If a bid is submitted and won by an advertiser, the ad can be served at this point. However, to provide a level of protection, verification technology is sometimes inserted into the value chain. Because the focus on the exchanges is audience buying and the human element has been removed, there has been a fear in the advertiser community that the ads may not be served in the best way for their brands -- e.g., being placed next to overtly sexual or violent content. Being served next to a competitor on the same page is also something that could potentially happen. Verification systems try to discern the context of the page before serving the ad, and it will block the ad if it determines the page doesn't fit the advertiser's specified settings.
Another function of verification systems is to validate the buy itself. As the impression comes up for bid and is won by the advertiser, the system determines if it really matches the target that advertiser was bidding on. Is the ad being served in the right geographic area? Is it ensuring compliance to the industry-specific guidelines and not breaking any rules (for example, pharmaceutical companies have strict editorial needs)? Once again, if the verification technology detects anything outside of the ordinary, it will block the ad from being served. Also in this category are contextual providers such as Peer39 and Proximic, which leverage semantic filters in real time to protect advertisers from buying the wrong impressions.
Verification technologies can differ in methodology and implementation, but, basically, if the impression passes the verification system's filters at this point, the ad is ready to be loaded into the blank ad box on the publisher site.
The third-party ad server
On some occasions, advertisers can load ads directly into the exchanges or allow their DSPs to serve them. However, most major advertisers load the exchanges (and DSPs) with creative "tags," which are ad requests that are filled by their third-party ad servers.
Third-party ad servers solve the same problem with serving ads as a DSP does with buying from exchanges. Instead of having to manage all of their display buys for each individual publisher (or exchange), the third-party ad server allows for everything to be managed in one place -- reporting, tracking, creative management -- one platform for the agency or advertiser teams to worry about.
So, the exchange sends the advertiser's creative tag (either directly or from a DSP) to the publisher's first-party ad server. The publisher's server loads the creative tag into the blank ad box. The creative tag "calls" to the advertiser's third-party ad server to then serve the ad. The third-party ad server will serve whichever creative is deemed appropriate for the situation. For example, if it detects that the end user is in Texas, there may be a specified ad to be served to that user. Or, if the third-party ad server detects that the user visited the company's website and checked out a certain product page, it may be directed to serve an ad related to that product.
Abracadabra! The ad is served. Remember, this entire process has to happen virtually instantaneously to ensure a quality experience for a website user. From the moment the page begins to load, to the publisher's tools farming it off to an exchange, to the exchanges receiving bids from their own users and their partner DSPs, to being filtered through the verification systems, and finally returned by an advertiser's third-party ad server-- it all takes place in around less than half of a second.
It's almost impossible to imagine that technology buying, such as the exchange system, will completely replace human interaction. Frankly, not all inventory can be sold this way. However, with the promise of lower costs from efficiencies, fewer wasted impressions through smart targeting, scalable reach, and controlled frequency, it seems very likely that these technologies will evolve into even more impactful tools. Competency in these platforms has already become a priority for leading advertisers, and new ideas for innovation are springing up every day. It's reasonable to assume that we have only scratched the surface on the capabilities of these systems and that they will be even more powerful and ubiquitous in a very short time.
On Twitter? Follow iMedia Connection at @iMediaTweet.