The last days of the ad exchange

Darren Herman wants a conflict-free ad exchange. So do I. We assume there could be one because we look at the much bigger and more efficient financial markets and see they are run on exchanges. But, in reality, the financial world has moved on from exchanges, and so will the ad world.

An article of faith: markets inevitably evolve from inefficient middlemen/arbitrageurs to efficient, transparent and fair crossing platforms. From Bazaar to Exchange. And then they live happily ever after.

But wait, consider this: of the buy and sell orders UBS handles in NASDAQ-listed stocks, it sends less than 5% to an exchange. The rest it internalizes (the infamous “dark pool“). That is, UBS matches buyers and sellers of stocks amongst their own brokerage customers.

Brokerages avoid exchanges. Instead, these days they tend to either be market-makers, internalizing as much of their trading as possible, or sell their order flow to third-party market makers, like Knight Trading*.

Internalization allows brokerage firms to minimize exchange fees, keep the bid-ask spread for themselves, and avoid giving information to competitors. Internalization does not provide transparency and fairness, like an exchange does, and that is part of its draw for the market-makers. Putting your orders into the exchange, where everyone can see them, is like a poker game where everyone else can see your cards. Exchanges level the playing field, and if you’re smarter than average, you dislike level playing fields**.

Exchanges were the best way to minimize transaction costs when communication was cheap but computing power was expensive. That time is past.


All the major ad exchanges are now owned by some of the biggest online media companies. AdX: Google. RMX: Yahoo! AdECN: Microsoft. They are no longer open markets, they are internal markets. As Yahoo! turns off Invite Media and everyone else contemplates the same, the media companies start to look like silos. Google’s acquisition of Invite is the final clue. Invite is their e-Trade, the customer UI that allows easy access to their inventory. Whatever audience you’re buying, Google can find it in their content network. So can Yahoo!, to an extent, and Microsoft and AOL and FAN and Akamai. If they don’t happen to have an audience in house, they will buy order flow by subsidizing publishers to come into their content networks. Each of these companies can internalize all orders that come to them. They will not interoperate*** and they do not need to interoperate.

By internalizing, these media companies get to keep the transaction costs and keep their market activity quiet. They also get benefits that the brokerage houses aren’t allowed–because the brokerage houses are regulated–like pushing their own inventory even if there is a better deal for their customer somewhere else.

How this will play out: the major media companies will each buy or build DSP-like capabilities to allow data-driven access to their inventory. They will build out their network of publishers so they can fulfill any audience request internally (internally here meaning either their own inventory or that of their enfiefed publishers.) At that point media buyers will be faced with an array of relatively undifferentiated media companies to buy from, each offering to best place the media buyer’s ads in its own audience.

Sound familiar? This was exactly the situation of the media buyers four years ago, vis a vis the ad networks. We have taken our two steps forward and are now taking one back, to a closed world where media sellers protect their margins by obfuscating what they are selling. Not with the complete opacity of the ad networks, but through the inability of buyers to learn because they are kept apart from the data and segmentation that guides their buy.

Media buyers need to figure out how they can hold their own against the big media company market makers. They need to build, buy or closely partner with a DSP, one that is direct connected to all the large media companies and pub brokers, and lets the media buyers have their own proprietary data, algorithms and results.

Tomorrow’s online ad buying world will look a lot like yesterday’s, only with more technology.


Prediction without predictions is just prattle, so here are some:

  1. There are not enough DSPs to go around. I count maybe ten indies that have technology up and running. A few more in the works that I know of. There are at least twice as many companies that will need to build or buy one. I don’t expect more than two or three of the current indies to still be independent in 18 months.
  2. Yahoo! will realize it needs to buy order flow. It will bring in some premium ad networks to provide audience balance in RMX by buying or partnering.
  3. Direct connection between DSPs and pub brokers/publishers will proliferate, producing much fail among the technologically naive.
  4. Third-party ad exchanges will stop calling themselves that and start calling themselves what they are, pub brokers.

We’ve had a lot of innovation in the last five years. It’s starting to worry the entrenched players. Their reaction is to move us from innovation to integration. There is some good in this: allowing audience buying, dynamic optimization and RTB at the major media companies is a big step forward. But there is a lot of bad also. There is still a lot of innovation that needs to happen, especially on the optimization side. And the publishers that aren’t big enough to be market-makers themselves are going to be even worse off than they are now.

An independent ad exchange–really a private crossing network with a clearinghouse function–needs to exist. It will allow innovation to continue outside of the spotlight. But it needs to rise up organically from the industry, because there’s no money to be made, no exit. The New York Stock Exchange was formed as a cooperative by a group of brokers who needed interconnection and interoperability. They didn’t support it because they thought the entity itself would be valuable, they did it to make their own businesses more valuable. What we need now is our very own Buttonwood Agreement with the same aims and similar methods.

* A brief survey of the market microstructure issues around internalization is here. But read it for its discussion of transparency.
** The leading proponents of the recission of NYSE Rule 390 in 2000 were the big brokerage houses. Cynical voices said this was because they were also the largest investors in the ECNs. But it seems obvious from a remove that the investments in the ECNs arose from the same cause as the desire to get rid of 390: the desire to trade in private.
*** This is where the financial services analogy starts to fall apart. The big market-makers interoperate not just through the exchanges but through ECNs and private crossing networks. The reality of financial markets makes this necessary. There is not, at this stage, and won’t be for some time, the same need for interoperability between, say, Google and Yahoo!


  1. This post is spot-on and I think your predictions will prove to be accurate. We are moving into a world of smarter, closed ad networks, not agnostic, universal ad exchanges. Bad news for media buyers and their clients, but probably good news for entrepreneurs.

  2. “And the publishers that aren’t big enough to be market-makers themselves are going to be even worse off than they are now.”

    Any predictions for the publisher side of this equation? Who will provide their technology?

  3. Tom –

    I think publishers have a few choices.

    If they are considered “premium”–primarily meaning big, as far as I can tell–they will be asked to supply inventory to the exchanges. They may even be paid extra to do so, for the same reason market-makers pay for order flow.

    For non-“premium” publishers, the choices will be:
    1. AdSense. Good luck with that.
    2. An ad network, hopefully one that is big enough to be a market-maker itself because the smaller ones will be in the same position as the non-premium publishers.
    3. A pub broker, in which group I include both the yield optimizers, like AdMeld, and the pub agents, like PubGears.

    And, because there will be less transparency, the pubs will also need to partner with companies like MetaMarkets and Yieldbot so they can build some learning in-house.

  4. Great mind ticking post:

    Unless you are Google – then you will continue to trade with competitors – it is like the empty air plane seat, impression call happens only once and cannot be sold again!
    Ad networks are the market makers in terms of aggregating demand and supply, i agree that they should be of considerable size – scale.
    Exchanges will continue to act as balances between demand ad supply for market makers. I agree that networks will build their internal environment and most of their business will be what we call “managed to managed” using external exchanges for additional media on the supply side and better monetization on the demand side.
    I agree with the general statement that our industry is becoming much more technological dependent.
    And that DSP’s with or without self-service capability are the new generation ad-networks.

  5. Terrific post Jerry. At OpenX we strongly agree with your thesis, summarized in the statement: “An independent ad exchange–really a private crossing network with a clearinghouse function–needs to exist.

    This is why we built OpenX Market. The first truly independent second generation exchange built from the ground up. We’ve been running for more than a year now (we launched in April 2009) and OpenX Market already reaches nearly 400 million unique users per month (384 according to Quantcast).

    A few fundamentals: OpenX Market runs a full real-time second-price auction for every impression. For buyers we support both rules-based bidding and real-time bidding for buyers, as well as complete set of retargeting tools. We power a complete clearing house on the back end to make life simple for publishers and advertisers. We provide a range of quality tools for publishers and through our Price Protection feature we enable publishers to set exact floor (reserve) prices for their inventory so they can participate in the OpenX Market without taking any economic risk.

    Crucially, we are open to *all* classes of buyers and sellers. We’re delighted if a publisher already uses OpenX’s ad server, but OpenX Market integrates with any ad server. Similarly on the buy side, we are are open to intermediaries (ad networks, DSPs, agencies) but also direct advertisers. We strongly believe having as many, diverse participants as possible creates a thriving marketplace.

    Furthermore, we are totally transparent. We charge a 20% transaction fee for winning bids. It’s the same, worldwide, for everyone. In the other exchanges you mention, it is extremely hard to find out what the economics are, largely because many of them are individually negotiated with participants.

    The approach is doing terrifically well and we’ve started to roll it out internationally through our global alliance model of regional partners. Our first announced partner is France Telecom/Orange who is operating the Orange Ad Market in Europe, which runs as the local European version of the OpenX Market. All demand and supply flows into one single global market to maximize competition. And of course we support the local languages and currencies, as well as cross-currency bidding, meaning a US buyer can bid in $ for French supply in Euros, and we handle the conversion through our clearing-house.

    I thought I’d share a small sample of press coverage from the past year:

    We think providing an independent exchange platform is hugely important for the industry. We’re glad you and other influential players see it the same way and hope we can play a key role providing this service.

    Talk soon.

    Tim Cadogan

Comments are closed.