|
Introduction
In order to understand the various ways interactive ads are delivered today and how these
different delivery methods affect measurement, CASIE asked Richard (Dick) Bennett of
Audit Bureau of Circulations (ABC) to author a paper on this topic. Dick is the Senior Vice
President of Audit Services for ABC and ABC Interactive. This paper will describe a number
of different approaches of serving online advertisement banners, as well as different
technologies which have been developed to deal with the measurement problems caused by
caching. While a few definitions of key words have been included in this paper, a full
glossary of terms (CASIE Glossary of Internet Advertising Terms and Interactive Media
Measurement) is available through the Association of National Advertisers, Inc.
Based on research fielded by ANA and CASIE, the buyers (advertisers and agencies) want
accurate and reliable measurement of ads displayed. How that is accomplished is up to the
trackers, measurers, and Web publishers, and neither CASIE, ANA, nor AAAA are
supporting any specific methodology.
Ad Serving and Tracking
This paper will briefly describe the five most prominent methods of ad serving and tracking
that are in use today. They are static ads, insertions, dynamic ads, cache-measured ads, and
browser-measured ads. Before we get into the details of these specific ad delivery systems, a
gross simplification of the process is as follows:
• Step One: User requests a page from content site
• Step Two: Content site requests an ad to insert in page
• Step Three: Ad is inserted in the page
• Step Four: Ad is downloaded to the user's browser
• Step Five: Ad is displayed on the user's screen
• Step Six: Ad is viewed by the user
All the steps above can be measured except for the last one. Click-rates/click-throughs/clickto-
views are certainly indications that the ad was viewed, but there are many ads viewed with
no measurable action.
The five techniques which will be described next equate to the steps above as follows:
• Step Two: Requests = Cache-Measured Ads
• Step Three: Insertions
• Step Four: Download = Static Ads or Dynamic Ads
• Step Five: Displayed = Browser-Measured Ads
Special Note: The preceding descriptions do not attempt to define whether a given ad is being
requested by a "real" visitor interacting with the Web/ad server, or whether the ad is being
"pushed" to the browser in some automatic fashion. This concern could be addressed by
designating the suggested terms as active or passive.
Static Ads
This approach is characterized by the placement of the ad file name within an image tag
inside the HTML page. The user's browser then calls for the image (ad file), after loading the
HTML page, which will in turn be recorded as a "hit" in the server's log file. Thus the Web
server will have recorded a "hit" for the request for the page as well as a "hit" for the request
for the ad. Additionally, the server will also be able to record the successful (or unsuccessful)
downloading of the ad file to the browser. This is done via status codes that rely on the
underlying protocols of the Internet, TCP/IP, to determine disposition of requests. This
technique is used by sites with no ad rotation/server system, or an in-house developed
application. Sites which employ this technique place and rotate ads in
batches. For example, they might run a program once per day to tag each page within the site
with the appropriate image (ad file) name intended to be displayed on that page. Ad activity
recorded for servers using this method of rotation is performed by the web server when the ad
file is served. Measurement via this technique would qualify under the IAB's existing
definition of an Ad Request.
Caching can have a significant effect on measurement using this technique. Assuming that the
site employs little or no cache-busting techniques, this measurement approach understates
activity due to caching because any request (for a Web element) which passes through a proxy
server that has previously cached that element will most likely not be seen or recorded by the
Web server.
Insertions
With this technique, an ad server software component (decision maker) is integrated with the
Web server software, so that when a page is served to a user's browser, a decision is made
regarding which ad to serve the requesting browser, and the HTML code for the chosen ad is
"inserted" into the HTML page as it is served. Tracking of ads served in this manner is done
at the point where the ad server makes the decision about the ad being served. This method is
used by many ad management software products (Accipiter, NetGravity, RealMedia, etc.)
currently available and is a normal configuration. This method of measurement is qualified
under the IAB definition of Ad Request. However, a significant concern about this technique
is that the actual serving (delivery) of ads is not recorded and therefore not measured.
Measurement of ads using this methodology is subject to several inaccuracies. First, browsers
configured to have graphic files turned off are prevented from making the actual request for
the ad, and therefore, the user was never exposed to the ad, even though the software recorded
and measured the ad request.
Secondly, when a Web site's pages are served from cache, the ad server software has no
opportunity to perform the ad insertion, resulting in the display of ads that can not be rotated
or recorded by the ad server software. This results in the undercounting of actual ad activity.
It also results in unintended ad rotation.
Dynamic Ads
Sites employing this technique embed a tag into an HTML page which makes a generic
request for an ad from an ad server. The ad server then determines which ad to display and
delivers the ad to the browser. Activity is recorded when the ad is
served/delivered/downloaded to the browser. This methodology is used by some ad server
service companies as well as some ad management server software.
This technique is very similar to the approach of using static ads addressed earlier. It differs
because the page is served without reference to a specific "ad file," but rather to the "ad
server." Thus the number of ads served and measured using this technique should
approximate the number of pages served by the server, assuming all pages have ads. When
used in conjunction with cache-busting techniques, this method can closely approximate the
ads that get displayed on browsers. Although certain "smart" proxy servers have been known
to cache ads served in this technique, there are differing opinions as to the number and effect
of these "smart" proxies.
While ads served this way still meet the IAB measurement guideline, the result can be
significantly different measurements than those derived with previously mentioned
methodologies (static ads and insertions). Differences relate to the varying degrees of caching
of ads as well as the fact that deliveries for actual ads are being measured versus insertions for
ad files on pages served. Obviously, the more cache-defeating mechanisms employed by the
site, the higher the measurement. The dynamic ad serving methodology also accurately
handles the recording and measuring of "gif" file activity where the browser has graphics
turned off.
Cache Measured
This method is a modification of the dynamic method. Instead of the ad server returning the
ad to the browser, the server delivers a directive, telling the browser the ad to display, and
also where that ad is located. The browser then must go out and retrieve the ad. The ad
request is recorded at the point that the server makes the ad decision and issues the directive
to the browser.
This method allows for the approximate measurement of ads stored and provided from cache
by the proxy server or the browser, as well as those ads actually provided from the ad server.
It does not measure deliveries, but rather requests for the given ad, although somewhat
differently than when the insertion method is used.
This measurement methodology provides statistics which are obviously not comparable to
most of the above methods unless they are employed with cache-defeating mechanisms.
When set up correctly, the cache measured technique accurately deals with browsers
configured with graphics turned off.
Browser Measured
This technique allows for the recording of the ad activity via software running on the browser.
It employs software which runs on the browser, like Java, which handles the request for an ad
from the server, the playing of the ad, recording of the ad played, as well as the transmission
of the recorded activity back to some central repository.
This activity actually is capable of measuring Ad Views/Displays, that is, ads that were
actually displayed on the user's screen.
How to Control Cache
Additionally, differences between measurements of a similar category can be explained by the
degree of caching, cache-busting or cache measurement employed.
"Cache-busting" may be defined as the forcing of a web element (page or ad) across the
network to every browser that requests it. Cache-busting can be performed through a number
of techniques. Some of the more commonly used cache-busting techniques include [the
following]:
1. The site can set the expiration date of the ad to sometime in the past, i.e., January 1,
1990, telling the cache that the ad is out of date, so that the proxy server needs to get a
new ad upon the next request.
2. The site can give unique names to each ad that it serves. Since each ad served has a
different name, the proxy has problems identifying requests for the same ad, which
allows the request for the ad to be passed back to the server.
3. The site can set various other "header" parameters (HTTP file attributes) to request
that proxies not cache the ad. Those headers include the pragma, no_cache, and cachecontrol.
While use of a combination, or all of the above techniques will greatly reduce the amount of
caching of a site's traffic, a number of "smart" proxies now exist, which have been trained to
ignore these maneuvers and will cache in spite of them.
Cache measurement, on the other hand, refers to the ability to measure activity that was
served from cache. Currently there are two known methods for cache measurement. One
method of cache measurement is described above. Another would be to obtain the proxy
server logs from all proxies which cached a given web element.
Disclosure of cache busting or cache measurement techniques employed by a site would
enhance a user's evaluation of the statistics presented in measurement and auditing reports.
If you have comments or questions, please direct them to Dick Bennett via email
<bennettrp@accessabc.com>
© 2001 ABC Interactive.
http://www.abcinteractiveaudits.com/news/white_paper_2332.html
Dick Bennett
SVP Audit Svcs/CTO
ABC Interactive
1701 Golf Rd
Tower III, Suite 300
Rolling Meadows, IL 60008 | 
Disclaimer
