Helping the Invisible Hand: Simple Measures to Make Online Ad Exchanges Work
At the same time, the ad exchanges have goals to offer premium impressions. This is a dilemma, and an opportunity. Automated buying and selling of ad inventory via exchanges is a marvel of modern advertising technology. The stakes are high, with traffic running at 25 billion impressions a day across the top exchanges. If we could use the speedy and efficient trading mechanisms of exchanges while letting premium publishers get a fair price, the marketplace would sing.
Yet, here we sit, with a miraculous ecosystem that can't quite do what the marketplace needs. Exchanges have no clear way for buyers to bid on inventory quality based on the criteria that mean "premium." Would you buy a cantaloupe while wearing a blindfold and gloves? Buy impressions on blind exchanges, and that 's exactly what can happen. The bidder does not know who the publisher is , and that 's not likely to change.
What indicates "premium"?
Try this test . Close your eyes and take a mental stroll through a media salesman's briefcase. Just past the aspirin bottle and boarding pass, we'll find words about the stature of the media brand: trustworthy, super-targeted, big reach. In a brochure we'll find words and graphs suggesting that the right audience will see your ad in the glowing halo of great content.
In exchanges the situation is different. Advertisers will find their own audience, thank you very much. Still, the impression carries no information about the context into which the ad is about to be served.
Fortunately, there are several things we can measure that are indications of great context: Page-dwell time, topical alignment, ad-hover time, ad-load and content-sharing propensity.
These data indicate that a page would be a good place to put an ad and correlate with brand success generally. Today, we can't use such data in the bidding process.
This basically thwarts a rational marketplace. The quality of the goods is unknown by the bidders. The invisible hand can't do its job.
What's standing in the way? First, making the right measurements available to the transaction system is a little tricky. This is because you don't know how long a consumer will dwell on a page or hover over an ad until they are finished looking at it, long after (in internet time) the bids have been committed. To bid on context quality you need the data before the ads are rendered.
To fix this, what we need is a teeny little forecast based on the history of access to each page that carries advertising. This data must be made available to the buy side of the exchanges.
As with all forecasts, the input is history. In effect, we need a normative database of page and placement quality indicators, such as "How long does a consumer typically spend on this page?" Publishers would share the URL with the exchange ... but the exchange could protect them by keeping the domain blind to the exchange buyers.
People know what quality pages feel like to the consumer. These pages are kinder and gentler, content-rich, useful and readable. They make for a comfortable, unhurried experience. These are the pages readers dwell on. These are the pages and placements advertisers would buy—if they could.
In my experience, great content, generously presented, results in higher click-through rates and vastly better ad performance.
It's not like I'm sticking my neck out here: Publishers have been saying this since Gutenberg.
If buyers could bid for impressions based on quality measures, premium publishers would have an incentive to bring their inventory to the exchanges. Prices would float up for inventory that was promising, and the transaction overheads of digital ad buying would decline.
With a little help from the right measures, the invisible hand could do its job.