When Algorithms Collide

Around the Net: When Algorithms Collide

Potential conflicts of optimization solutions

The online advertising marketplace has seen its fair share of arms race-like tit-for-tat technology battles over the years. Whether it was between spammers and spam filters or pop-up ads and pop-up blockers, our largely self-regulated industry has technologies battling for supremacy all the time. While most would
see the examples above as a black-and-white battle of good versus evil -- spam and pop-up ads being the poster children for black-hat marketing tactics -- the battle currently being waged between sellers of online display advertising and buyers of that inventory is fundamentally different. Each party has a right to claim the banner of fair play and neither really views the other as evil -- just not necessary aligned with their best interests.

To get a better sense of the battlefield, let's examine the goals each side is attempting to achieve. Advertisers are looking to generate high-value conversions, CPA or conversion rates that deliver positive ROI, and a reach and frequency that is optimal for their product and brand, all with the greatest operational efficiency. Publishers, on the other hand, are looking to maximize the margins they make on producing or sourcing their ad inventory, while trying to generate the minimum advertiser performance necessary to get their insertion order renewal.

In an effort to meet these goals, advertisers and publishers pursue different optimization algorithms. Advertisers tune their buying algorithms to take advantage of their deep understanding of customer profiles and cross-publisher behaviors to target their best, most-likely-to-convert prospects. They can bring tons of data to the party from their own CRM databases and purchase third-party data to augment their own customer profiles.

Most advertisers are able to differentiate the value of a conversion and carry this analysis forward to evaluate both the revenue generated by a particular publisher buy, but also the profitability of the buy. Advertisers are looking to reflect these differences in more flexible, dynamic CPM pricing that can be updated and revised frequently to reflect the prospects and customers they are reaching on a publisher site. They also often use a single set of tags and a single ad serving system to aggregate and unduplicate conversions to drive a consistent view of their campaign's performance across their entire buy.

Margin of Error
It is a very different view from the publisher's perspective. Publishers have a deep understanding of their viewership profiles and traffic patterns and how this relates to their inventory availability. They have historical data on their cpm rates and sell-through as well. And while they have a good sense of the reach and frequency of their site, they lack the necessary feedback from most buyers on how this fits in to the campaign as a whole. The scale of the buy they are able to package is also limited to their traffic. Publishers are usually not given much information, if any, on the value of the conversions they are driving for advertisers. And while most publishers prefer CPM pricing for the obvious reason of revenue forecasting, most also recognize advertisers have some performance goal they are attempting to hit and will therefore request that their publisher tracking tags be deployed on the advertiser's site.

These different goals and capabilities often lead to a discussion about allowable publisher margins. It's no surprise, therefore, that publishers are investing in technologies that help them generate the highest bids
against avails, maximizing ad inventory yield and margin. But how much margin should publishers be generating? While the margin floor is the percentage that will allow them to stay in business and cover their costs, the upper end of the allowable margin is harder to define.

The advent of ad exchanges and open marketplaces for buying and selling ad inventory has the power to resolve the arbitrary nature, perceived or real, of the publisher margin question. The bidded media model does not, in and of itself, resolve the apparent conflict of advertisers' and publishers' goals, but it does provide common ground on which to find a workable solution, creating a marketplace where publisher and advertiser algorithms can compete and set fair market prices for inventory. To fully leverage open marketplaces, both sides, as well as the marketplaces themselves, need to focus on expanded data sharing, targeting enablement, and more dynamic media pricing.

In an effort to work better together, advertisers need to be willing to share more meaningful data with publishers about overall campaign reach, frequency, customer profitability and audience targeting variables. At the same time, publishers need to make their publisher-side data on context, placement and behavior transparent to the advertisers so that advertisers can fully value inventory performance.

On Target
To put the insights gained from data into action, further industry investment and focus need to be placed on enabling advanced targeting. The increase in the number of open marketplaces is driving a need for standardized targeting methods. This problem can be more fully resolved by marketplaces accelerating their development of real-time bidding environments, and allowing advertisers and their agencies to bring their own data-targeting platforms to bear. For the segments of media buys not placed in open marketplaces, publishers would increase their revenue and help advertisers by directly enabling more advanced audience targeting based on what advertisers learn across their campaigns.

Once the processes and infrastructure for data-driven media targeting are further enhanced, more dynamic pricing is the mechanism by which the mutual economic opportunity can be achieved. Advertisers need to evolve their display media plans to be more in line with search campaigns by making audience profiles analogous to search terms, while publishers will do well to reevaluate fixed CPM pricing or artificial price floors for more dynamic CPMs. At the same time, the marketplaces themselves need to ensure that the auctions are resolved in an unbiased way for both sides. Fully leveraging the open marketplaces will drive not just the specific goals for each side, but also drive the common goal of scale, leading to larger total digital channel opportunities for the advertiser and more inventory sell-through for the publisher.

3 comments about "When Algorithms Collide".
Check to receive email when comments are posted.
  1. Scott Brinker from ion interactive, inc., November 30, 2009 at 8:27 a.m.

    Excellent article. It's time for the online marketing industry to push for data standards that make the marketing funnel, writ large, more interoperable. Folks like the IAB should worry less about banner ad formats, and more about the consistency of the data flowing through the end-to-end process.

  2. Tom Troja from Social Sympony, November 30, 2009 at 10:42 a.m.

    As usual, Ted and Leon are on target... ditto to their POV. Like to add to the conversation.... about the creative. At the end is a person looking at a brands display creative. Need to think about looping back that reporting info with A/B testing and other creative optimization plays inside the data. We will get this data standards figured out sooner than later.... the emotional goo of who likes what why... what imagery and words.... is the real puzzle and payoff for brands and advertisers who are paying the bills.

  3. Greg Hall from Yebol, November 30, 2009 at 9:55 p.m.

    Good work. It's very insightful.

Next story loading loading..