ellipsis flag icon-blogicon-check icon-comments icon-email icon-error icon-facebook icon-follow-comment icon-googleicon-hamburger icon-imedia-blog icon-imediaicon-instagramicon-left-arrow icon-linked-in icon-linked icon-linkedin icon-multi-page-view icon-person icon-print icon-right-arrow icon-save icon-searchicon-share-arrow icon-single-page-view icon-tag icon-twitter icon-unfollow icon-upload icon-valid icon-video-play icon-views icon-website icon-youtubelogo-imedia-white logo-imedia logo-mediaWhite review-star thumbs_down thumbs_up

Demystifying the world of ad networks

Demystifying the world of ad networks Jay Friedman

It seems that at this point in the interactive world there could be an entire book written on ad networks. For the sake of brevity, and your sanity, we'll keep this much shorter. The most basic introduction is that there are two kinds of ad networks: Those that own their own inventory across multiple sites and sell them as a "vertical network," and those that don't own any content but operate like a Priceline or Hotwire for online ads. The latter are horizontal networks, which take the millions of unsold ad impressions available at any time of day and sell them in aggregate.

Let's start with the first kind: those owning their own content. A good example would be Martha Stewart Living's Omnimedia (MSLO). This company owns all of the Martha Stewart-branded sites plus WholeLiving.com and Pingg.com. A rep from MSLO can offer you inventory across all of these sites and create a custom program, often with custom ad content or packages within these sites.

A big benefit to using networks like these is the ability to have your ads look or behave differently than a normal IAB ad unit, if you have enough money. Plus, if you and the publisher share an audience, the partnership works well. The downside is if you're an auto manufacturer, while much of MSLO's audience may at some point be your target, very few of their users are shopping for a car at any given time.

Now, imagine you are a large website or even group of sites. Your site has 20 billion monthly impressions available to sell and your 50-person sales force sells around 10 billion of those per month at rates of $8 to $12 CPM. There are 10 billion unsold ad impressions each month that, without a secondary market or avenue for this inventory to be monetized, would be a wasted opportunity for that publisher. Publishers like these have the choice of running house ads (for which they receive no immediate monetary pay out), PSAs (same problem), or offloading them onto an ad exchange or network at a severely discounted rate.

Nearly every publisher, large or small, faces this scenario to some degree or another. The challenge for the publisher is that if they sell the other half of their inventory for less than it's selling for in the market, the market may not want to pay $8-12 anymore; advertisers will want to pay the discounted rate.

So, how does the market for unsold impressions work? The inventory is sold through secondary-market ad networks (SMAN). There are two primary types of ad networks operating within this group. The first is the arbitrage-style network and the other is a demand-side network using a demand-side platform (DSP) as its technology. Both buy inventory across hundreds (or thousands) of sites and resell it to their clients in bulk fashion.

There are currently about 50 secondary-market ad networks. These networks differentiate themselves in five basic ways: inventory quality, method in which they buy inventory, research data associated with each impression, customer service, and the ability to optimize your campaign to generate the best results.

Within each of these characteristics are multiple methods and styles. Understanding the network game can be complex, but it is an area of the business well worth knowing. Often times, inventory sold by these networks is discounted 50 percent or more compared to the regular "retail" rate charged by the publishers outside of the network. The trade-off to the buyer is they normally will not be allowed to see how individual sites perform; metrics are reported in aggregate. Individual site report could compromise the publisher and the sales staff's ability to sell inventory on their own at higher CPMs.

Here is a quick guide to understanding these five differentiators and ways you can ensure good, safe performance for your clients when working with these vendors.

Inventory quality
Originally, SMAN's were made up of thousands of long-tail (read: small and unknown) websites that didn't have their own sales staffs. Nameyobaby.com isn't necessarily a bad site, but very few media buyers would look at that site and say, "Wow, now that is premium quality inventory!" Premium is only premium based on the definition of the media buyer, but most media buyers define "great quality inventory" as inventory coming from sites their clients are familiar with. 

These sites could be People.com, ESPN.com, and others. Some argue the site doesn't matter and it's the user that counts. Others argue the site trumps anything and ad content seen within the context of familiarly-branded content is crucial. Because of this, along came "premium" ad networks. Some of these do in fact offer brand-name inventory. Many claim to be premium, but their inventory certainly wouldn't meet the definition described above.

The only real way to know what kind of inventory you're using is to use a verification service such as Adometry to see the exact sites where your ads are being placed. This way, if a network shows you a list of 100 great sites but only runs you on 10 of those -- and they are all of lesser quality -- you will know whether or not you want to work with them again.

Next page >>

Buying method
The original ad networks operated on an arbitrage basis. They would establish a relationship with publishers in a semi-secret fashion, buying their unsold inventory, aggregating it with hundreds of other publishers, and then selling this aggregation to clients in either a "run of" fashion or in category channels (i.e., entertainment, sports, news, etc). This is where it became standard to never share the site lists and metrics from the individual sites the client was actually running on.

If a brand-name client knew they were getting inventory on certain brand-name sites for a severely discounted rate, the publisher would have no leverage in selling its own product directly (or at least getting a good price for it). Many of the largest and most well-known networks operate in this fashion. So, these networks tend to buy tens or hundreds of billions of impressions and then do their best to sell that inventory to clients for whom it makes sense.

The other way networks buy their inventory is on a real-time, or at least client-by-client, basis. This is how some networks, and all DSP operators, work. The networks that do this figure that if they pre-buy their entire inventory, they may have inventory that doesn't match the needs of their clients at that time, and they wouldn't want to force any non-matching inventory on a client. So, using exchanges or simply making the purchases as each client comes on, these networks buy the right inventory at the right time. Because of this the cost may be (but not always) slightly higher than the pre-buying networks, but it's up to you as a client to determine whether or not that premium is worth the benefit.

In the early days of ad networks, the ability to run across hundreds of sites and optimize to the best performing ones was reason enough to love networks. However, as networks and data exchanges built cookie pools of various user behaviors, the sites on which advertisers were running became less important and the user who actually sees the ad becomes the focal point.

"Why does it matter if the ad serves on NameYoBaby.com if we know the user is in the market for a sport-utility vehicle?" This was and is the thought process of many networks. You decide how you would answer that question. The important point, though, is that through data exchanges and private cookie pools, almost every network now uses cookie data to target users without personally identifiable information, but for better targeting.

Other types of targeting can come from demographic, psychographic, or even clickstream data. The latter has the largest volume of data behind it (full disclosure: our networks incorporate clickstream targeting) and enables networks to identify sites on which your clients will be most likely to find their audience. Coupling these three types of targeting with user-level data targeting can be a powerful combination for any advertiser.

So it's not all about the data after all. Just like any other medium or business, you have to like working with the person servicing your business. They have to be capable and flexible and seek to understand exactly what your goals are. Some networks won't take a buy of less than $25,000. Some won't take small geo-targets. The more time you can spend "interviewing" an ad network and letting them interview you, the better the fit, guaranteeing the campaign will be done right the first time.

Back in the late '90s and early '00s, automated optimization was a huge deal. Everyone's algorithm was better than the next firm's. Then agencies and advertisers learned that while some automated optimization engines can do a decent job, most of these engines are set to move inventory slowly -- say, over 15 to 20 days -- so that no rash moves are taken with a campaign.

If you're the type of advertiser who runs consistently for at least 90 days without creative, offer, or targeting changes, an automated optimization engine will probably work just fine. However, even the networks with the best automated optimization engines know that if a campaign is short or the parameters will change frequently, human intelligence and intervention is crucial. The quality of this human intelligence goes back to the "People" section above, and is often a differentiator.

For full disclosure's sake, my company owns and operates three horizontal ad networks, so I certainly believe in them. I also know they're not right for every situation. Having personally come from a traditional media background and then ultimately purchased inventory at one point or another from 25 different networks, I also know that the trade-offs inherent when buying from networks seem hard to get past when new to online.

We've worked with dozens of clients in this area, and every single one was glad they ultimately worked with a network, whether ours or someone else's. The benefits of great data plus the ability to optimize across hundreds of websites at a low cost very often provides for an "ROI no-brainer" to clients at all spending levels.

Jay Friedman is the COO of Goodway Group, which owns and operates three ad networks: Beep! Automotive, IvyPixel, and Sway Political Network.

On Twitter? Follow iMedia at @iMediaTweet.

Jay Friedman is COO of Goodway Group, and a partner in the 3rd-generation family company founded by Milton Wolk in 1929. Friedman joined in 2006 to add a digital media component to Goodway’s offerings, beyond the existing print and promotional...

View full biography


to leave comments.

Commenter: Nichole Kirsch

2010, June 11

Lots of great information on networks, and it's nice to have it all presented within the context of a single article.
But what about back end transparency? We believe a big differentiator among networks is the ability to see or verify where your ads are being placed down to a page level, sometimes eliminating the need to pay for ad verification. In the debate about sites vs. audience, what really matters is knowing what your getting.