Today's Ad Exchanges Don't Work

Considering Digital - Joe DoranJoe Doran is a board member at Legolas Media. He was previously founding CEO of Media6Degrees and GM for Microsoft Digital Advertising Solutions.

A few years ago, online publishers were told that salvation was just around the corner if only they’d sell their inventory on ad exchanges. Buyers would bid—in real-time no less—on available inventory & audiences that the publisher hadn’t sold.  Selling the unsold inventory the expensive salesforce didn’t move and at near premium pricing due to a competitive liquid market. Enabling the publisher to maximize yield and rolling in cash.  Unfortunately, ad exchanges haven’t always worked as advertised.

Even when put to good use, inventory sold through exchanges doesn’t represent the bulk of a publisher’s revenue. While some in the exchange business believe this is because they haven’t yet convinced publishers that ad exchanges produce value on several key fronts…

First, ad exchanges continue to face a crisis of transparency and control. Yes, transparency is getting better, and increasingly publishers have the ability to control which advertisers buy their inventory (on some exchanges). But there’s still a strong perception among publishers that ad exchanges are beyond their control. It is a terrifying proposition for a publisher to have their audience & data to be sold (or worse yet – stolen) without their input. It means they have no real control over their business and their future. As long as that perception continues, fearful publishers won’t anchor the bulk of their business around ad exchanges, and I don’t blame them.

Second, the lack of liquidity in exchanges  presents problems for publishers who must balance guaranteed deals (with pre-determined price and volume) against the highly volatile market for their inventory from an exchange. Without a methodology for balancing these two concerns—and eliminating potential conflicts—ad exchanges will remain ancillary to a publisher’s revenue model. After all, predictability is the lifeblood of a media company, and so far exchanges haven’t allowed publishers to bank on their CPMs.

Third, a key drawback to exchanges is that they really don’t support rich media ads. Part of the blame for this can be put on the liquid nature of trade, and part can be put on number of platforms needed to enable an exchange. But putting the issue of blame aside, the reality is this: exchanges have become known for dealing mostly in what I call the lowest common denominator ad formats (think about the last mortgage ad you saw, and you’ll have an idea of what I’m talking about). While those cheap formats serve a purpose, they don’t generate much revenue for the publisher, and they don’t impart a premium experience to the user, either. The result is that neither buyers nor sellers turn to ad exchanges to tackle large-scale, rich media buys.

Finally, in theory, an ad exchange should be able to discover the real price of the inventory. This is what markets are supposed to be good at, after all. Unfortunately, much of what goes on in an exchange is driven by complex algorithms that factor in two concerns: the value of the eyeballs behind the impression as predetermined by past behavior and the overall ROI status of the activity. While there’s tremendous value to that kind of decision-making, it tends to leave out key factors such as the quality of the content (in terms of its historical performance), the social affinity of the user and the content, and the uniqueness of the audience.

When robust technology platforms and markets are working to their potential, publishers should be able to bring forward other attributes of their inventory and have it priced accordingly. Eventually, that process will empower sellers and buyers to develop a more diversified set of methodologies for valuation.

If a solution can’t address all of these challenges directly and simultaneously, then it simply isn’t a solution that will move ad exchanges forward into the future and allow publishers to capitalize on the opportunities that are now within reach.   And ad exchanges and auction based display inventory will be left to the bottom of the inventory allocation barrel and the scraps from the big media buyer’s budget.

15 Comments

  1. In regard to your last point, don't forget that exchanges have a major inhibitor to effective real-time bids (and thus true pricing) - even large RTB buyers are forced to throttle their buys. It's not that these are complex algorithms so much as throttling to meet their server capacities.
    To have efficient pricing you need relatively full information, but both buyers and sellers only receive a subset of information. This isn't just a publisher issue, but rather buyers are potentially also suffering because they aren't able to test across a broad array of impressions (both due to their own budget constraints and due to the limitations on the current platforms out there).
    I'm willing to go out on a limb and suggest that both exchanges and yield optimizers are making the market more efficient, but that until major technical hurdles can better be solved the opportunity is limited.

    Reply
    • James - Thanks for the comments. I agree with you that "Real Time Bidding" doesn't make a real time auction. Lots of technical hurdles that people are trying to figure out to ensure you can bid appropriately but it isnt there yet.

      Exchanges and Yield Optimizers are good for the ecosystem and have helped us move forward. I just believe that they aren't going to grow much or unlock big dollars until it fixes some core business as well as technical issues.

      Reply
  2. I don't agree with the crux of this argument. Yes the exchange marketplace needs to mature but there is no doubt that for remnant / DR focussed inventory it represents the future (what is the alternative, back to the network model?)

    Higher value brand, rich media, video et al media buying can and will continue to be sold in a more traditional manner. Tech will automate to a point here but the qualitative value of such buys will always necessitate some human involvement.

    Reply
    • Bear - I agree with you, for remnant/DR focused inventory exchanges represent the future. But the promise of exchanges was that the bulk of the display inventory would clear through an auction - just like Search. Where buyers & sellers can get scale & efficiency against a very labor intensive media buying process. We are not on that path as an industry.

      Exchanges are valuable. They serve a great purpose. I am a big proponent and early builder, buyer, seller of exchanges! I am just saying the market growth is going to tap out without changes and it will limit us to remnant/DR placement. Most publishers don't think of their inventory as remnant - even ones with a ton of it. And to lure big TV budgets online, we need to do more and work harder to make it easy to spend billions of media dollars and have them work for our advertisers.

      And it is well understood in the industry that eCPMs in remnant have been declining for years. Arguments around the flood of inventory or that arbitrage plays are taking too much share of the pie or that the inventory was overpriced to begin with. The arguments don't matter as publishers are only getting about ~50% of the eCPMs for this inventory class from where is was just 2 years ago - and that is with increased investment in people and tech. Those economics force publishers to do other things with their inventory...like not put it in exchanges. This might actually help the exchange in terms of creating scarcity.

      Reply
      • Thanks for replying Joe. Your points about publishers not thinking of their inventory as remnant and the low value of DR inventory are well taken and big challenges for the space.

        Most publishers will be unable to sellout their inventory with brand rates. My hope is that exchanges and RTB allow them to get a fair price for the remainder, something that hasn't been true in the recent past.

  3. Andy M

    Joe,

    Great post. Couldn't agree more. One of the things that we have found is that in order for exchanges to provide meaningful value to publishers, they must have all of the things you talk about. Transparency. Liquidity. Right-pricing. Support of cutting edge, video creative.

    Right now, they do (very close to) none of the above. They are designed to address, like you said, the "lowest common denominator ad formats" and in this new, highly engaging medium, highly engaging, unique ad units are needed.

    Furthermore, the biggest drivers of "performance" are based on leadback and basic retargeting. How are marketers supposed to acquire new customers when their agencies are so focused on targeting the people who have already been to their website?

    Bottom line - ad exchanges are like the stock markets. They are only good if they have the key ingredients. Smart people trading. Intelligent arbitrage and trading techniques. Relevant ways of valuing data and media (CTR is not relevant).

    The building blocks are there, but we are still a ways from realizing a true, productive and efficient, "exchange."

    I may be showing our cards a bit here, but I am comfortable in remembering that back in the day, not all ad networks were created equal.

    And, this is true today -- that not all exchanges (and those that utilize them) are created equal......

    Rock on Joe. Good post.

    Andy Monfried

    Reply
  4. Jordan Mitchell

    Good post, Joe.

    One other point -- the supply of online advertising inventory is greater than the demand for inventory, which means that an auction-based pricing mechanism (particularly 2nd price auctions) is certain to yield the lowest price. This makes auction-based ad exchanges more advantageous to buyers than sellers. Compounding this problem is the fact that every auction-based exchange enables buyers to bring their own data, and essentially bid for higher-value users/impressions at commodity prices. In financial exchanges this is called “insider trading”.

    Combine these factors with the recent technological advancements on the buy-side, and it means that exchanges are simply arbitrage marketplaces -- and arbitrage is not good for publishers, it's only good for arbitrageurs.

    I blogged about this in a little more detail at http://kickstand.typepad.com/metamuse/2010/03/restoring-balance-to-the-online-advertising-market.html if it's of interest to anyone ...

    Reply
    • Jordan,
      I read your post - it is very good. The point I would make about excess supply is that it isn't all created equal. The explosive growth of supply has come at the bottom of the publisher quality (from the POV of a publisher or advertiser looking at engagement) most of it in transactional environments (social media, e-mail, blogs, ect). This is not the same as a rich content adjacency. The fact that top tier content publishers are seeing a decline in eCPMs suggests that something else is wrong - not just too much supply.

      The auction in display is very complex. You have bidders that will compete on different attributes - some on user data, some on site / category data, some on ad unit/type. This means that to get a efficient auction you need at least 3 to 5 X the bid density you would in a simple keyword based auction system in search. That is hard to do. It favors the buy side when the market is not efficient (as well as the arbs).

      Great points Jordan - Thanks for the comments.

      Reply
  5. LookingConfident

    .
    Jordan Mitchel has posted, that: "the supply of online advertising inventory is greater than the demand for inventory, which means that an auction-based pricing mechanism (particularly 2nd price auctions) is certain to yield the lowest price."

    But is this so (?), is my question in reply to both Jordan and those in this exciting industry.

    Isn't it (surely?) the case that all inventory remains dormant until such time as a 'user' happens to access any particular (specific) page? And ONLY at that time, is an impression called for (to be served) and that as such, any amount of inventory (the billions of pages added daily) has no bearing whatsoever or, should never ever be taken into account (when talking the subject of inventory supply/demand for inventory), for UNLESS a user happens to access an individual (any particular) page or, a number of pages for that matter, within that (growing by the minute, available web inventory) TOTAL of content pages, on the web.

    Correct me if I am wrong, but isn't that the basic (or, fundamental) function of an RTB auction marketplace, after-all?

    LC

    Reply
  6. Joe:

    Not only is transparency a huge issue but many times the ads running on the pages contain malware which is proven to lead to identity theft. Ultimately, advertisers and consumers are at a much higher risk.

    Best,

    Michael

    Reply
  7. LC,
    Thanks for the comment. You are right that until the page loads, there is no inventory. Just like airplanes sitting on the runway isn't revenue until it takes off. The trick is in managing the efficient scheduling of advertisers to take those ad placements (or seats on an airplane). This is yield (or inventory) management.

    Publisher have history of past traffic and predict how much inventory may be generated in the future. They presell those ad units through their direct sales force - on a time based sponsorship basis or an impression basis. Any unsold inventory they allocate in a couple ways - to house inventory, to advertisers as bonus or make goods, to resellers (ad networks or contextual networks) and if they still have anything left they drop to exchanges. Sometimes they lump the resellers and exchanges togetether.

    The promise of RTB is to let it all demand sources compete and let the highest price win against all this inventory. But in practice, RTB exchanges are only allocated a sliver of inventory and not the best quality.

    Reply
  8. Patrick

    This is an extremely interesting discussion. I think we all agree that not all inventory is created equal, but you're also ignoring a huge benefit that exchanges and data targeting can bring to advertisers- audience. Content has always been a proxy for audience. Most publishers do not have the scale to add sophisticated audience targeting. This is a clear benefit that publishers cannot provide and is one reason for shifting budgets to exchanges/DSPs.

    Furthermore, publishers participate in exchanges at will. They choose to auction their inventory, so there is really no reason to complain that it's not being valued highly enough. If that's the case, they can always stop participating and sell 100% direct.

    The only forces at work here are those of the market. Advertisers are looking for particular metrics and will seek out the path of least resistance to reach them.

    And while we lament the state of web publishing, let's also remember that publishers arbitrage too- they just do it with SEM and other cheap traffic to ensure delivery on high-paying campaigns, undermining their advertisers' goals.

    I think this is an extremely valuable discussion to have, and ultimately I think it will be resolved when there is seamless integration with a DSP and a large scale ad server (I'm thinking Invite-DART). Then you can more finely stratify your inventory and allow access to real-time buyers at premium price points.

    Thoughts?

    Reply
  9. DSP Jockey

    Some of these points are valid - the rich-media nut is barely cracked; video is a distant promise on the horizon; transparency is still an issue for any brand-focused client (Can we give you an example site list?). But the bottom-line concept of the ad exchanges is something that the industry has to take note of.

    The notion of squeezing even more efficiency out of digital media is enticing for any cost-conscious advertiser. Knowing that there is available reach to facilitate data strategies is evolving every month.

    And I would much rather implement across ad exchanges than turn my media dollars over to dirty-black-box networks that dump my behavioral campaign into a mixed-bag potpourri of RON and meaningless long-tail impressions.

    Give this machine time to develop and we will all be better off, along with our clients.

    Reply

Add a comment

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>