Lots of startups and VCs are betting their business on web advertising as a business model. They point to Google, Yahoo, and other big sites as examples of success and assume they will have proportional success. Not likely.
In a previous post "Is Ad Targeting The Next Big Thing" I asked why we aren't seeing higher CPMs and better targeting based on all the attention data from Web 2.0 services and click stream data. I heard from several advertising executives and would like to share their insights here.
Michael Yavonditte, formerly CEO of Quigo (acquired by AOL) answered my questions with these concise and insightful responses;
(1) It's hard to use all the available data to target. There are lots of reasons for this. Advertiser confusion is a big one. Too many choices can overwhelm advertisers and create a watered-down marketplace.
(2) Advertisers will pay more for inventory that converts better but they also need scale. Often times better targeting leads to smaller quantity.
(3) I know of some "attention data" being used/tested. The results can sometimes be surprising and counterintuitive. It's very hard to find new targeting data that can also scale across lots of inventory or sites, which is why behavioral targeting hasn't become a multi-billion dollar market yet.
(4) There is a ton of inventory but not enough high quality inventory at the moment. I think prices will eventually move up as new ad formats are invented and big branding dollars begin to match online user consumption habits.
OK, let's dig a little deeper. First let's level set on some definitions.
Page Views - We need to distinguish between advertising on Search Engine Result Pages (SERPs) like Google, and advertising on typical web content pages like Yahoo.
SERP ads - Google has been very successful with SERP advertising for two basic reasons, and lots of complex reasons. First, SERPs have the advantage of targeting an advertisement to a very specific search term entered by the user. Second, the user is in active "search and discovery mode" so they are more open to an ad offer.
Display or text ads on content pages from Yahoo, MSN, AOL etc., don't have either of the above advantages. Content pages have hundreds or thousands of words so it is hard to target. And, the user is in passive "browse mode".
How do ad servers work? Most ads, other than SERP ads, are targeted based on content on the page, in very broad categories like; news, sports, business, politics, etc. They try to dig deeper into the context of the page, but with varying degrees of success.
Daniel Jaye, a former exec at Tacoda, (now part of AOL) explained to me some of the complexities of ad servers. Here is a snippet from our discussion;
Most ad servers do not allow the ability to perform complex targeting based on the wide variety of data available, but the issues are complex.
1) Ad Targeting technology (most ad servers have the following insufficiencies):
a. Multi-valued attributes (ie a browser (person) may be a member of several segments, how do you choose/prioritize?, traditional Boolean logic doesn’t suffice)
b. Complex Boolean targeting (ORs, ANDs, etc in specific order of evaluation)
2) Inventory Management
a. Multiple criteria: Ad servers have to prioritize delivery based on multiple criteria today: (roadblocks/takeovers by content or by targeting criteria, schedule commitment, performance objectives, etc.). The more data we add, the more difficult it is to predict and allocate available inventory, manage pricing/scarcity etc.
With regard to real results, the performance improvement varies. For some categories (auto intender, consumer electronics shopper, etc.) the performance gains are clear and relatively easily obtainable. For the broader set of offers, automated optimization has many challenges and has NOT demonstrated a clear advantage over basic techniques like frequency capping and contextual targeting.
There were several industry terms in Mr. Jaye's response that need further definition;
- Frequency Capping - means restricting or (capping) the number of times (frequency) a web user sees a specific ad within a 24 hour period, across a whole network of web sites. This is a simple concept, but incredibly difficult to execute millions of times a day, in a split second, for every individual user, over hundreds of web sites.
- Contextual Targeting - means targeting ads based on the words (context) on a page. Again, simple in concept but incredibly hard to do at web scale in less than one second.
- Road Blocks / Takeovers - when one advertiser wants to dominate (Takeover) all available pages and (Block) other advertisers. Sometimes referred to as a "carpet bombing advertising campaign".
So, the ad serving technology needs to improve significantly. Results from ad targeting campaigns need to be proven. And, the ad serving technology must be easy for a junior ad buyer to understand and use.
Sometimes tech folk forget that at the other end of the digital tentacles reaching across the ad-powered web, sooner or later, is a person pulling the levers. Joe Blow media buyer - think pimply-faced state college graduate 2-years out of school making $18K/year and living in Queens - simply can’t keep up with the pace of innovation on the web. They learn a few of the biggest ad targeting systems, which incidentally are the only ones with sufficient scale for them to complete their buys and move on, and crank 90% of their ad budgets through them. Once in a while one of the better ad network sales guys buys them some really good sushi and makes a decent case for some experiment, and voila, that technology gets thrown a $10K “test buy” bone.
The equation the ad buyer solves in his head:
- What is the probability that doing this will make me a hero with my boss and/or client?
- What is the probability that said boss and/or client is going to smack me for wasting time on this $10K diversion instead of doing everything I can to “optimize” the $500,000.00 buy we already have in process on Yahoo?
- Is the potential upside of 1. materially greater than the potential downside of 2.?
Everyone agrees that ad serving technology will improve, and that "attention data" will be used to better target ads. They also agree that advertisers will be willing to pay higher CPM rates once the effectiveness is proven. As usual, the technology will advance faster than the customer's ability to use it and desire to pay for it. The big question is when will all these factors converge to launch another multi-billion dollar market? My guess is two to three years.
What do you think? Leave a comment and join the discussion.