The Best Ad Server Wins

At investment conferences and presentations to hedge fund, mutual funds, and private equity investors, I’m asked the same question (in different ways): which of the large search engine ad networks will win: Google, Yahoo, or MSN? Perhaps the questioner includes banner networks, such as ValueClick, Burst Media, and AD.com, to try to assess the effect the top search engines will have on the Internet ad networks outside of search.

In a world where our beloved (or feared) search engines are transforming themselves into auction-based advertising network marketplaces, online marketers also want to predict which players will lead the field in the next several years. We need to ensure we know the ins and outs of each network and the best practices in campaign structure strategy and execution.

The answer boils down to just a few factors:

• Critical mass of serious advertisers. For a network to continue attracting publishers, it needs lots of advertisers. But those advertisers must be willing to pay a reasonably high rate that’s close to what the advertising is actually worth. It would be even better, some would argue, if that network had advertisers willing to pay even more than the impression or click was worth to them in some cases.

  As I’ve said before, there are only two kinds of advertisers at the top of the paid search results: brilliant marketers and total lunatics. In a large pool of advertisers and within an auction marketplace, there will often be some advertisers who behave irrationally and emotionally, bidding more than any rational analysis would indicate is reasonable. Publishers like auction media networks with large advertiser bases because the likelihood of irrational bidding increases. You could still have a functioning network with the largest, most efficient rational advertisers, but adding more advertisers adds a bit of value, particularly in SEM (define), where there’s a long tail of very niche searches, including those with specific geographic requirements.

• Critical mass of quality ad impressions. Advertisers don’t even bother signing up for (or filling out insertion orders for) a network unless it has a certain critical mass already. Many would argue the best way for a network to have a critical mass of ad impressions would be to own the publishing properties on which the impressions appear. In the case of a search engine, Google owns its traffic, acquired Blogger to further enhance traffic directly under its control, and cut a $1 billion check to AOL to lock in a great partner.

  Part of what attracts a publisher to a network is the large advertiser base, and part of what attracts an advertiser to a publisher is a large network. This makes starting a network difficult unless you have a significant chunk of traffic owned or controlled by the network (or a ton of money with which to guarantee payouts to a publisher). Marketers, particularly rational ones, want a balance between scale and targeting. The larger the network, the more likely the advertiser can get both in one place.

• Ability to detect and eliminate click, impression, or action fraud. To gain and maintain the truest of advertisers, all networks must demonstrate an ability and a commitment to detecting and eliminating fraud among publishers and across advertisers. Larger networks with existing revenues can typically devote more resources to fraud detection, prevention, and policing as their costs are amortized over a larger revenue base.
• Ability to accept medium-quality publishers without value dilution. For auction media networks, every publisher wants to join the network where they’re subsidized by the high-quality publishers already in the network. However, the publisher can’t allow new publishers into a network because the network’s overall value drops. Google has tried to address this issue with Smart Pricing, where the publisher receives a lower CPC (define) — and, therefore, effective CPM (define) — and the advertiser is charged a lower rate as well, preserving ROI (define). The system’s far from perfect, but the concept is sound.
• Ease of use for advertisers (absence of friction). Advertisers, agencies, and company owners don’t have a lot of time to learn how to use an administrative front end to a new network. A network must be easy enough to use to justify additional time and human resources. Of course, all the major PPC (define), CPM, and CPA (define) auction networks currently provide some kind of API (define) connectivity to third-party agencies or tools, making all of them easy to use for the segment that has access to API-driven technology.
• Targeting efficiency, yield maximizer, and relevance engine. Targeting efficiency while maximizing yield is the Holy Grail for an ad network. Once a network has the advertiser, ad impression inventory, and a fairly frictionless system (through an API or by being easy to use), the real difference boils down to targeting efficiency and a relevance engine that also factors in yield (effective CPM). A yield maximizer must include improved targeting and a relevance engine because consumers tune out advertising that’s not relevant, and those same consumers often welcome relevant advertising.

  Marketers are often willing to pay more to reach the segment of their target market for which the advertising is particularly relevant. The PPC search marketplace is built on this concept. As the PPC search networks evolve beyond pure search-driven ad servers and start using better contextual and behavioral targeting methods, all three constituents are happier. The marketer reaches a better-targeted set of consumers; the publisher makes more money by running the more targeted ads; and consumers see better-targeted advertising. The best ad-server code is one that doesn’t even care whether the marketer is being invoiced on a CPM, CPC, or CPA basis. The smart ad-server program will convert everything to a predicted CPM. Higher predicted CPMs are possible as the targeting and relevance increase.

In the near future, the mega-networks with staying power will be those that can pay publishers more because they have both a large advertiser base and the ability to deliver relevant advertising to consumers. Currently, a lot of ad buyers, both on- and offline, don’t worry much about relevance beyond a broad set of targeting parameters. BMW will advertise on a golf site, for example. Broadcast-style advertising won’t disappear because marketers need the reach. However, the ability to match the advertiser’s targeting preferences (as expressed by some combination of keywords, behaviors, demographics, dayparts, or psychographics) with consumers, while making the publisher the most money, is a killer combination.

We can expect to see major improvements in non-search relevance and targeting in networks over the next year. The consumer will win as ad relevance increases, but the real driver will be money. The networks understand that in the end, all other things being approximately equal, the best ad server wins.

Meet Kevin at Search Engine Strategies in San Jose, August 7-10, 2006, at the San Jose McEnery Convention Center.

Want more search information? ClickZ SEM Archives contain all our search columns, organized by topic.