CheckM8's Oren Netzer talks about how online publishers can take a page from the airlines to maximize revenues from ad inventory.

Many industries use yield management as a technique for extracting the maximum amount of revenue from a fixed quantity of perishable goods and/or services. From the online publisher’s perspective, proper yield management means maximizing the revenue from selling its ad inventory. Most publishers sell their ad inventory at different pricing levels, based on the extent of the audience segmentation (run-of-site or targeted) and the size and types of ad units. Since ad inventory that is not sold is lost, publishers attempt to sell their entire inventory to maximize revenue. On the other hand, since a publisher’s inventory is limited, selling a lower CPM campaign that takes up inventory that could have been sold at a higher CPM rate is less profitable. The challenge of yield management is to balance selling your ad inventory at the highest possible rate against minimizing the amount of your unsold inventory.

I was planning a best practices article on yield management, so I sent out emails to ten leading online publishers to ask them how they dealt with yield management and how they measured it.

I didn’t hear back from any of them. When I placed phone calls to some of them I found out why -- they were all still trying to figure this out themselves. So, I decided to suggest a method for yield management, based on my experience as a technology vendor and lessons learned from the airline industry, which is primarily responsible for the birth of yield management.

Prior to deregulation in 1978, all airlines offered the same government-approved fares. When People’s Express introduced the industry’s first low-fares nationwide, major carriers were forced to react or go out of business. That’s when American Airlines introduced the first true yield management system that allowed it to dynamically adjust fares based on historical and current booking patterns. Since then, yield management has been adopted by many industries that have similar inventory characteristics.

The first step is to determine a definition of yield in online advertising. I’d like to suggest that since most of the online inventory is sold on a CPM basis, and the user site navigation model is based on viewing pages (rather than viewing ads unfortunately), the most logical definition for yield is probably revenue per page view, although alternate definitions could be revenue per unique user or revenue per ad.

Why is Yield Management important?

Let’s look at the main factors that make yield management important:

• Relatively fixed capacity. If publishers could increase or decrease page views or ad impressions at will, there would be no reason to manage capacity.
• Ability to segment markets. The publisher can segment its audience into different groups. Segmentation can be done by demographics, behavior (behavioral targeting), geography, content (sections on site) or keywords. The key to this segmentation is the fact that advertisers are willing to pay variable prices to reach different audience groups.
• Perishable Inventory. Inventory that is not sold is lost revenue. The publisher should strive to minimize unsold inventory.
• Low marginal sales cost. Publishers have very high fixed costs. The variable cost of running an additional ad impression is negligible.

Let’s take a look at what’s affecting yield or revenue per page view:

Audience segmentation

The airline industry segments the customers on each route into various groups and sets a different price for each group (or “class” as the airlines call it). Two people sitting next to each other on the same flight could have paid completely different prices for their tickets. Classes are differentiated by the terms of the ticket. These terms include cancellation policies, purchase time and return dates. A businessman -- who buys a ticket a few days before the flight and wants flexible dates -- will pay more than a leisure traveler who bought his ticket well in advance. The airlines have conducted a thorough analysis based on historical data to decide how many seats should be allocated to each class. The allocation of seats per class and the ticket price per class are the keys to maximizing airlines yield (revenue per passenger mile).

In online advertising, inventory is segmented into classes based on targeting a specific audience. Targeting a particular section of the site or a particular demographic would cost advertisers more than running a run-of-site campaign. That’s a good differentiation of classes since advertisers are willing to pay more for targeted audience.

The next step is to limit the amount of inventory that is allocated to each class. For example, publishers may allocate up to 10 percent of their inventory to be sold on a run-of-site basis, while the remaining inventory should be sold targeted. The targeted inventory could further be allocated to the sales organization by different audience types.

This allocation should be done based on combination of historical figures and a best guess that is based on the experience of the sales organization. To allow this, it is imperative that publishers track historical sales by ad units and segmentation criteria, as well as sell-through rates and reservation to booking conversions rates. Making sure that booking and availability data available to sales people in real-time is key to driving their sales decisions.

Some publishers redirect their unsold inventory to ad networks to be sold on a performance basis to take full advantage of unsold inventory. There is an argument against this that in the long term this can decrease the value of the inventory and the brand.

Overbooking

One statistic in the airline industry is that 50 percent of initial reservations get cancelled or are no shows. To deal with no shows, airlines typically overbook flights by 15 percent. This is the historical figure that allows airlines to minimize empty seats on flights. Sometimes, airlines wind up overbooking (too many passengers show up for the flights) and the airline is required to reschedule some passengers to later flights and compensate them.

Online advertising uses a similar model in which inventory is overbooked. Overbooking is used to account for both cancellations of campaigns and insertion orders that don’t get booked. Additionally, overbooking tries to compensate for inventory forecasting inaccuracies. Publishers must establish an “overbooking percentage” in their reservations system based on their historical sell-through rates: This percentage will specify the amount of inventory that may be overbooked by the sales organization.

Audience overlap

A big issue for airlines is that a lot of their routes overlap since they use a hub-and-spoke model. In the hub-and-spoke model, passengers flying from New York and passengers flying from Miami might end up on the same connecting flight to Los Angeles. This adds a degree of difficulty to yield management.

Unfortunately for us, we have a similar complexity called “audience overlap” in selling online ad inventory. The same user that visits our site can belong to different audience groups. When I visit a site, I am part of the “male 18 to 34” demographic, but I also belong to the “New York City Metro” geographical targeting criteria.

Dealing with this issue requires the publisher’s ad server technology to have discrete knowledge of the audience types on the site. The ad server must know the exact amount of users and page views that correspond with each of the targeting criteria, as well as their visitation frequency over any period of time. If you are using external targeting technologies (for example, behavioral targeting), you must make sure they are well integrated with the ad server data.

The ad server should then be able to prioritize ads to users based on knowing the “booked to available” ratio of each targeting criteria. This means that when I go into the site the ad server would prefer to serve a “New York City Metro” targeted ad to me rather than a “male 18 to 34” because its “booked to available” ratio is higher.

Ad units per page

Deciding on the number and types of ad units per page is another significant factor for yield management (revenue per page). You are better off selling two $10 CPM ads on a page ($20 yield per page), then selling three $6 CPM ads on the page ($18 yield per page).

Lately, many online publishers have been redesigning their sites to decrease the number of ads per page to reduce clutter, and they have also been adopting larger ad sizes (728x90, 160x600, 300x250).

This corresponds with the airlines maintaining a balance for number of seats on the plane. The smaller the legroom, the more seats you can squeeze into the flight, but the lower your airline quality and service is.

Conclusion

Yield management is a complex task that will take the industry years to master. One of the first steps publishers should take is to make sure they track and save the necessary historical data. This means tracking sell-through rates, tracking demand patterns for targeted buys and for ad products as well as tracking effective CPM of various ad units and audience segments. It is also essential to make sure ad serving technologies can use this data to prioritize campaigns and allocate ads to users.

Oren Netzer is the director of business development for CheckM8, an innovator in online advertising. By offering new approaches and technologies for ad serving and rich media, CheckM8 increases efficiency and expands the growth of Web publishers and interactive agencies worldwide. CheckM8 technology is used by New York Times Digital, Tribune Interactive, ABC National TV Sales, Terra-Lycos, MSN Europe, E.W. Scripps, iFilm and many others.