In most of the world (the U.S. excluded), ad agencies get rebates from media companies based on their "pooled buying" power. In other words, they get cash returned, above and beyond their normal commissions, based on all the money they spend for all their clients.

Clients aren't routinely informed about exactly how much of these rebates is generated by their particular spending. Even when clients press for answers, they often allow agencies to keep the extra money so clients can continue to pay lower fees.

In the U. S. the pooled buying concept never really gained a foothold. But now agencies have the opportunity of using similar pooled buying and selling techniques to gain additional revenue in the digital-media realm. In this case, the money doesn't come in the form of media rebates but from buying low and selling high, a technique called arbitrage. Agencies are able to sell high because they add audience and behavior data to the original digital-media buy.

When I last wrote about pooled buying rebates, I wondered how a client knows how much of the rebate came from its own spending and how much the client is entitled to.

Now the question is: When agencies purchase media, do they use "pooled data" to enhance the value of the media buys as well as pooled media money from all their clients?

In the case of pooled-buying rebates, clients didn't much care as long as they thought they had driven a hard enough bargain with their agencies. "Most clients set nice, aggressive terms and don't worry about how the agency delivers them," the head of a British advertisers group told the Guardian. "In short, the advertisers know the buyers need profits and don't necessarily mind how they generate them so long as they don't feel obviously taken advantage of."

But data might turn out to be more valuable than money. More data leads to more accurate targeting and insight about the audience. So arbitraging pooled digital-media buys with pooled data not only can make the agency more money but provide clients with richer data.

So what's better? Proprietary data from one client or broader and more comprehensive data from all the agency's clients?

In the arbitrage process, the agency bids on online inventory through ad exchanges and their own trading platforms. The agency might buy the media, add in data, and then sell it at a profit to individual clients depending on their requirements.

But if the agency uses its various clients' data to add to the value of its media buy, shouldn't the client benefit somehow? And does the agency use its own money to fund its arbitraged digital media buys or the pooled media money from its various clients?

Murky waters
It's getting very murky out there in adland. Group M's global digital chief, Rob Norman, called his agency's approach "transparent but not disclosed." He said his clients know his group makes money, but they don't know the cost of the original media buy -- no more than they would if they bought inventory from a third party.

The ad exchanges themselves may be part of the problem. "With the technology of exchanges and the way the industry has developed, there are no controls of inventory," Christian Carrillo, VP of innovation at DataXu, a digital advertising technology provider, told the Financial Times.

The ad exchanges allow marketers to buy digital ads based on who will view them rather than the websites where they will appear. As a result, the FT reported, it's becoming increasingly hard to keep track of the websites where the ads actually appear, leaving an opening for "fraudsters" to profit by registering phony traffic on shady websites.

So when the agencies buy inventory on the ad exchanges, do they notify clients where their ads ran or are the clients only interested that the ads reach an interested audience? Is anybody keeping track of what websites the ads appeared on -- and for what clients -- especially since they could be charged for phony clicks?

What I don't get is what's the incentive for clients to allow their agencies to arbitrage their media buys, allowing the agency to mark up the cost of the media by using (at least to some extent) their own data?

Unless, as with pooled buying rebates outside the country, they're allowing the agency to make money only if fees are held down. (Agencies are quick to point out they're keeping fees plenty low. Jack Klues, VivaKi CEO, told a 4A's panel that agencies are in a "dangerously risky spiral" with the current cost model. The agency says it doesn't practice arbitrage.)

But the big question in my mind is: How does the agency earn its profit by making money on one side and giving it back on the other?

At the end of the day, arbitrage is not without its own complications. As Fred Gabriel, editor of Ad Age sibling Investment News, told me: "Arbitrage is all about risk. The arbitrageur is taking a risk and expects to be compensated for it. In the case of digital-trading desks, agencies are buying media at one price and betting that clients will be willing to pay more for it down the road. The big risk, of course, is that they won't and agencies could find themselves on the hook for the price of the original inventory."