How to Get Smart About Non-Guaranteed Inventory

Vadim Telyatnikov is senior vice president of OpenX Lift.

For many publishers, selling non-guaranteed inventory is a lot like making sausage: Pig goes in, links come out; let’s not think too much about the in-between, thank you. But publishers who have a better understanding of the process could vastly increase their yield, if only they’d have the courage to look closely at a complicated and daunting technology. Here are a few ways publishers can get a whole lot smarter about their programmatic selling.

Calculate your true CPMs. The highest CPM offered doesn’t always yield the most revenue. Discrepancies, or the number of impressions lost during the sales process, have a huge impact on your effective CPM and your bottom line. Let’s say that an ad network is willing to pay a $10 CPM for your inventory and you send it 10,000 impressions. Can you assume you’ll receive a check for $100?

Discrepancies always occur, but some networks have a higher discrepancy rate than others. If the $10-paying network has a 20 percent discrepancy rate, the most you can get for your inventory is $80. But if you sent that inventory to an ad network that offered a $9 CPM but has a discrepancy rate of just 10 percent, you would receive $81. Now multiply that extra dollar by the tens of millions of impressions you serve each month, and the impact on your bottom line is obvious.

Fill rates change the dynamics even more. The ad network offering the best CPM may specialize in retargeting campaigns or campaigns that focus on a particular industry sector, which means it will only fill a small portion of the impressions you send them and default on the rest. That leaves you scrambling to find another buyer (and incurring another set of discrepancies).

To calculate the true CPMs your ad networks pay, consider the price they offer in conjunction with their discrepancy and fill rates.

Know your buyers. When ad networks send back impressions, or default, they put your revenue at risk. You must find another buyer before the user clicks away from the page. Thus, minimizing defaults makes good economic sense.

The best way to lower your default rate is to understand which types of impressions your ad networks fill and which they reject. Identify the network most likely to buy each user prior to an ad call, and then send the impression to that partner first. Likewise, avoid sending impressions to ad networks that are likely to pass.

This approach boosts revenue because it determines the best demand channel based on the unique combination of user, page, session depth and more.

Break it down. Many publishers do a good job selling their inventory by site, section and user attributes, like lifestyle. But advertisers pay top dollar for many other attributes, including frequency, or the number of times a user sees an advertisement, and session depth, the number of pages clicked during a visit.

For example, many advertisers believe the ads users see on the first few pages they visit are more valuable than the ads they see on the tenth page. These buyers understand that the first page a user sees may not be your homepage, so they value session depth more than URL.

Other ways to segment are by geographic region and by section of your site. When you traffic these segments separately, you’ll find out what kind of inventory networks value most.

Use ad exchanges as a learning tool. Knowledge is power, and in our industry, that power comes from robust inventory-performance reporting. With access to the right reporting, you can analyze bids, see the unique combinations of user and site attributes each advertiser bids on most frequently, and the CPMs they’re willing to pay. You can see the value advertisers assign to each section of your site and set CPMs appropriately. Most importantly, you can use this information to create custom ad programs for your direct sales team, generate leads and segment customers.

Make your buyers work harder. The best way to increase revenue is to sell your inventory one impression at a time via “super auctions” hosted within your ad server, where all of your demand partners — ad networks, ad exchanges, agency trading desks, DSPs — compete simultaneously for your impressions the moment an ad call occurs. While this is difficult to execute manually, OpenX Lift automates the process.

OpenX Lift uses predictive algorithms that crunch historical data and calculate the CPMs networks are likely to offer for your impressions, based on each impression’s unique combination of user, page and session attributes. In a sense, these algorithms simulate a “bid” you can then use as a price floor in a real-time auction. As a result, networks compete with RTB buyers, and every impression is sold at the highest possible price.

Watch the video below to find out how OpenX Lift works.
Digiday Top Stories