DoubleClick's VP of optimization solutions explains how allowing ads to compete in new ways increases existing inventory's revenue, with no added effort.

Our culture thrives on competition. It is a driving force that compels people to work harder and find ways to excel. Whether it's the thrill of a two-horse race, the excitement of March Madness or the Super Bowl, we have a penchant for competition because we enjoy seeing the bar dramatically raised. Competition in its purest form drives its participants to find their peak performance, ultimately driving the best possible results.

The challenge of prediction: who will win the Super Bowl?
For online advertising, competition is about enabling your ads to compete so the best ad is served. But selecting which ad to serve requires an accurate prediction of which ad will perform best.

Predicting performance for an ad is like predicting the stock market or a sporting event. Look at the recent Super Bowl. Would anyone have predicted that the Bears would score a touchdown in the first play of the game? At that moment you may have predicted that the Bears would win, but you would have been wrong. Who would have known that it would rain in sunny Miami, causing six turnovers in the first half of the game? So what affects the outcome of a sporting event? In this case, it's how well the players performed that day as well as all of the external factors, none of which can be easily predicted.

While it's hard to predict which team will win when you pit the two best teams against each other, what if you matched the top ranking team against the lowest? What about an NFL team against a college team? Prediction becomes easier if you create segments with obviously different attributes.

Let's take it a step further. What if you broke up the teams, looking only at individual players? If you created a fantasy football team comprising the very best players from across the NFL, would you be more confident in predicting they would win? Of course you would. When you think outside the box, you can create optimal segments allowing prediction to be more accurate.

For those who play fantasy football, it's a lot of fun to spend hours analyzing players and their stats, but do you want to spend that kind of time analyzing ads? While some ad-ops professionals might enjoy spending late nights and weekends doing this, I think most people would rather click a button and have an automated tool optimize for them.

Get your inventory in a competitive mood
At the end of the day, optimizing online ads is about matching those ads with the users most likely to respond. In previous articles, I spoke about the value of optimization with regard to your premium advertisers. By improving performance for your top paying advertisers, you raise the value of your inventory, generate more loyalty and attract more spend. And you accomplish all this with less effort, while improving both your top line and bottom line. These are great benefits with which to start, but getting the full benefit of optimization requires optimizing all of your inventory.

By allowing ads to compete in new ways, you can gain more revenue from existing inventory without any additional effort. For premium ads and advertisers, your primary concern is to meet the delivery goals you've set, with an optimization goal of driving activities and clicks.

For remnant or preemptable inventory, the equation changes. You aren't necessarily looking to meet delivery goals, but rather to serve the ads that are going to make you the most money. Managing yield here is not about schedule; it's about dollars and cents.

In a year when publishers are looking to double revenue, and ad spending is expected to exceed $16 billion (up from the bubble peak of $8 billion in 2000), yield management is a key to realizing that growth. Finding innovative ways to maximize the revenue from existing inventory is critical.

For CPM ads within your remnant space, competition is about sorting the ads in such a way that the highest CPM ad serves ahead of a lower CPM ad. If you work with multiple ad networks to monetize your remnant inventory, you'd want the $1.50 ad to serve ahead of the $1.45 ad. That may sound trivial, but when multiplied by a dozen ad networks with millions of impressions, the dollars add up quickly.

For CPC and CPA ads, maximizing revenue is not as simple as sorting the ads. They must be matched with the users most likely to respond in order to reduce the number of impressions burned through to achieve a profitable response rate.

Allowing such ads to compete with CPM ads is not easy. For example, how do you weigh a $.50 CPC ad against a $1 CPM ad? It all depends on how many responses that ad receives. If you can predict the response rate of the ad, then you can translate that CPC or CPA rate into an effective CPM (eCPM) and allow all the ads to compete on an equal scale. 

Make your ad serving smarter
Predicting the response rate for an ad is as difficult as predicting the stock market or the outcomes of a sporting event. So how do you improve the accuracy of prediction?

Just as with the above Super Bowl analogy, whereby segmenting the teams and players improves the predictability of the outcome, segmentation can improve prediction for ad behavior. If you segment your audience and your content, and track the response rate per ad for each segment, you can more accurately predict an ad's eCPM. 

Crunching the wealth of data available to you from online advertising to manually optimize would be next to impossible. You have to segment the audience, track behavior for each ad per segment, estimate the eCPM, and then set your ad server to choose the highest value ad to serve first across hundreds of CPM, CPC and CPA ads. That would require a lot more than a few late nights and some long weekends. Automated optimization solutions built into your ad server can do this automatically with tremendous effectiveness and ease of use.

Tap into new sales channels
When ads compete, it opens the door for previously un-pursued sales channels. Direct marketers we've spoken with would love to buy inventory on a CPC and CPA basis, but they've been turned away in the past by publishers unwilling to assume the risk of booking such ads into their system. The risk, according to publishers, was that those ads would eat up millions of impressions without any clicks or activities. With advanced optimization and intelligent competition, CPC and CPA ads become not only safe, but desirable. Happily, you and the advertiser not only share the same objective, but now have the tools to achieve that objective. 

You can create a competition tier within your network where a broad set of remnant inventory can compete to achieve the best results. That inventory can now include: 1) lower-value, directly sold preemptable CPM ads, 2) inventory sold to ad networks and 3) CPC and CPA inventory directly sold to marketers. Throw your entire remnant inventory into the proverbial melting pot and see what happens.

When ads compete, everyone wins
Competition raises the bar for all players involved. By allowing your ads to compete in new ways with each other, you can get the highest value ads to match with the most appropriate audience, and serve in the optimal order. Consumers see the ads that are more relevant to them, advertisers get the performance they want and publishers drive more revenue out of their existing inventory. When ads compete in the most effective ways, everyone wins.

Brad Bender is vice president of Optimization Solutions at DoubleClick. Read full bio.