Not surprisingly, different sources of data will give us different tiers of data sophistication. The most common source for campaign performance data is the sites themselves, the ones that sell the media packages. They need to track performance to make sure they give the right number of impressions or other measures to their paying advertisers.
This can create a conflict of interest, however. Some agencies like to keep tabs on the sites by employing alternative counting sources, which include third-party banner servers, client-side reporting mechanisms, and information tracked from the advertising client’s Web site.
This is a potentially confusing topic, and the terminology involved certainly doesn’t help. Hopefully, going through the four categories, one at a time, will dispel some of the murkiness.
Almost all Web sites will provide postbuy reports that show their servers’ figures for how many impressions of the advertiser’s ad were shown. In most cases, this is the number against which the insertion order will be resolved, unless the client or agency has enough clout to demand the site use the client’s or agency’s server.
Some sites will give additional information, hoping to provide a competitive advantage over the other sites. Sometimes buyers will see breakouts of performance based on user segmentations. A site might break out the data by the demographics it has or, more likely, simply by the section in which the ad appeared. This data can be useful to help determine the most effective placements for that site but is most often not comparable to similar data from other sites.
At best, this type of data will give you Tier I and Tier II data. A minority of sites will offer (at extra cost) to “instrument” your ad such that they can track a subsequent action on the advertiser’s site.
Some agencies make a regular practice of creating their own alternative to this. They set up separate Web landing pages for each and every media buy or creative execution they create. That way, every site gets a different referring URL, allowing the agency to track buyers on the client site to individual buys or creatives.
Third-Party Banner Servers
Employing banner servers is either a brilliant masterstroke or a confounding obstacle, depending on your perspective. The people responsible for trafficking ads out to the right sites at the right times and in the right sizes know that their jobs are sometimes possible only with the aid of such service. Those people responsible for creating quick-turnaround performance reports swear by them.
Media researchers, on the other hand, know that the vagaries of banner-serving technology dirty their numbers just enough to make some forms of analysis useless. Banner servers make advertising relatively quick and efficient, but if a campaign objective relies on a complex analysis of numbers later, they might thwart the very purpose of a campaign.
Since sites use different definitions and different metrics to count impressions and other media data, we need some sort of consistent reporting mechanism. Early on, it became tempting to use banner servers for this purpose, and now we’re pretty much stuck with them.
Early efforts to force sites to adopt a common definition of performance and to be able to automatically get that data into an agency database – thus avoiding banner-server data pollution – met with abject failure.
It used to be that many agencies had a tough time getting a lot of sites on one campaign to allow the use of a single third-party banner server, largely because each of the sites had a vested interest in promoting the use of another banner server. You could use a third-party server, as long as you used the site’s third-party server. The market has opened up since those days, however, and most sites now make a practice of accepting the major players in the market.
Buyers usually employ banner servers in one of three ways.
- Sometimes, though rarely nowadays, a client company will mandate that a particular banner server be used. Sometimes this allows large companies with multiple agencies to norm all their data in one place, giving greater potential leverage. When clients do this, agencies frequently find it problematic when the client’s banner-serving system proves incompatible – either process-wise or data-wise – with the agency’s existing serving system.
- More often, an agency develops a contract with one of the off-the-shelf banner servers, passing the costs on to the client. This is most typical in agencies that conduct a lot of online advertising.
- For those companies or agencies that do not conduct a great deal of online advertising, most banner-serving companies offer campaign-by-campaign pricing. Some agencies choose to have the banners served only for some campaigns or for some clients.
It should be noted that some agencies have a problem allowing some of the banner-serving companies to become infrastructure partners. Many of these technological infrastructure providers have been bought up by Web sites and Web site networks, creating a conflict of interest. A few years ago, agencies would have dropped those media vendors as potential media sources, but today agencies believe that this straddling of both sides of the fence is acceptable. It would be the moral equivalent of an accounting-package service provider in traditional advertising — like Donovan — starting to sell its own media to agencies through its own accounting system. The frequency of the practice has bred an attitude of acceptance in the industry, at least so far.
Agencies can employ banner servers to get both Tier I and Tier II data, as long as they work with the banner-serving company to instrument the Web pages on which they hope to measure customer actions.
When users come to a site with an advertisement that is instrumented with a client-side applet, their browsers load the graphic and begin to run the applet. This little application then observes the users, determining if they are actually able to see the ad (as opposed to having already scrolled past its location by the time it loaded). The applet can then follow the users through to the client site, providing that the client’s site is also instrumented with the applet technology.
Because the applet can get more refined information, like whether the ad was occluded off the page when “viewed” and how long a user’s cursor lingered over the advertisement, this allows for a much finer analysis in the end.
These tools provide data directly to an agency or client database, showing how individual customers react to a particular piece of creative in a particular media placement. Since the technology relies on cookies, it can also match up with an existing client database of tagged individuals on the Web.
Client Web Site Information
Finally, the advertising client’s site provides information into the mix. In the best of worlds, marketers would be able to help determine exactly how performance information is collected on Web sites. This would allow them to set up systems that would link to banner servers or other types of mechanisms to allow for information linking from ad view to purchase.
Alas, in the real world, marketing departments and information technology (IT) departments are on different floors, if not in different buildings. And managers of these groups very infrequently find themselves in the position of forcing the other to do their bidding.
With many of my own former clients, the IT people felt threatened about marketing folks getting involved with these undoubtedly technical topics. A defensiveness tended to arise when young marketing people came into the offices of IT managers, talking about how they needed to instrument Web pages and use pixel tracking via banner servers.
Ironically, the resistance to cooperate was greatest at some of the largest technology companies, like Microsoft and Compaq. It seemed that the higher the status of the IT group within a company, the longer it took to get them to come to grips with the data collection required for smart marketing.
When preparing a campaign, the planner has to be realistic about which data sources will be available and in what time frame. Taking a census early on in the planning process will allow the planner to figure out what types of analysis will be possible and desirable. And, in some cases, sites and technology partners can be added or subtracted to the final list of partners to accommodate.
Next week, in the fourth part of the trilogy (thank you, Douglas Adams), we will talk about which sets of performance numbers need to be used for which purposes.