AdverNET '99 COVERAGE: Measuring Performance

From: donna stryk <donna.stryk_at_tenagra.com>
Date: Fri 5 Nov 1999 09:13:40 -0600

We've all heard that only 50% of our advertising actually
works, but the problem remains...which 50% is it? Online
advertising provides marketers with more accountability than
any other medium. The online world has given us numerous
ways to monitor, count, track, and analyze the success of
our interactive marketing. AdverNET '99 dedicated the
fourth session on the media buying track to the subject of
measuring campaign performance. This session proved to be
one of the most informative and interesting of the
conference.

The session began with Dave Zinman, Founder and Vice
President of Marketing at AdKnowledge explaining the process
of ad delivery on the Web. He said "the key essential
ingredient in why there are so many companies doing so many
types of measurement on the web is that web advertising is
delivered one at a time." From there, he explained how an
individual's browser requests a web page from a web server.
This page includes a tag that will in turn request the ad
from a local ad server. From this process, two types of
measurement have evolved. The first is browser centric
measurement. This method can take a sample of computer
users and place a program on their system, which tracks
exactly which ads they have seen and how they responded to
those ads. Another way to conduct browser centric
measurement is to embed a program in the ad that extracts
information from the browser and sends it back to the
server. The second type of measurement technique is server
centric measurement. This method measures the number of
requests that the server received for a particular ad as
well as where those requests came from. The panel of
speakers included representatives from companies
specializing in both types of measurement. You will gain a
better understanding of the pros and cons of each as you
read on.

Zinman continued by giving details of how a server centric
measurement system such as AdKnowledge can help clients
track the activity after the initial click. When an ad
banner is served the server delivers a cookie file, which
tracks the user's information. When a user comes to the
advertiser's site, their cookie file is read again and that
information is sent to AdKnowledge. Using these files,
AdKnowledge can track several types of activity. First,
they can track who saw the banner and clicked on it. This
is the all too familiar click through percentage. Second,
they can track how many of the customers that clicked on the
ad proceeded to purchase, subscribe, or whatever the desired
activity is for that particular campaign. Finally,
AdKnowledge will log the cookie file of everyone who saw the
ad. If someone sees the ad and later comes to the site
without actually clicking on the banner, the AdKnowledge
system will read the cookie and let you know that this
person had previous exposure to your banner. These
measurements quantify the true value of an impression on any
given site.

Another aspect of online advertising that appeals to
marketers is the timeliness. It is relatively fast to begin
an online campaign and changes can be made more easily than
in traditional media. However, in Internet time the period
it takes to get a campaign started can seem like an
eternity. According to Dave Miller, Director of Flycast
MediaNet, there is a four-week cycle regarding online banner
advertising for marketers not using third party ad servers.
This cycle includes five main steps: trafficking banners,
running the ads, site and performance reporting, analysis,
and sending out new trafficking instructions. This cycle
creates problems for web marketers, which Flycast's MediaNet
solution helps alleviate. The first problem that marketers
face is the mountain of data that can be extrapolated from
an online campaign. Unfortunately, there is never a clear
course of action attached to the data and different sites
may report differently. Flycast provides optimization
reports that help you sift through the numbers and make
intelligent decisions more easily. The second problem is
the length of the cycle itself. Using MediaNet's user
resources, marketers can reduce the turnaround time from
four weeks to one to three days. Marketers can delegate
data manipulation and other tasks that, although necessary,
they do not have to do themselves to Flycast. This frees up
their time to take on planning, negotiation, creative
development, and strategy development. Another obstacle for
online marketers is the time spent managing technical
challenges. Flycast provides technical troubleshooting to
help you resolve these problems faster. Finally, the
current system provides good reports, but not necessarily
more customers. To help make sure that each campaign
succeeds, Flycast dedicates an advisor to the account and
performs timely business analysis to help make optimization
decisions. Again, this is a server centric measurement
company and all tracking data is based on how many ads were
requested from the server and cookie file information.

After hearing from these two server centric media tracking
companies, Don Westrich, Director of Business Development
for Thinking Media spoke about his browser centric system.
He started by outlining some of the pitfalls of server side
measurement. The first problem is that not every ad that is
served is actually seen by the viewer. Some of these ads
never load completely or users may close the page before the
ad has a chance to load. Secondly, servers have to bust
cache in order to derive accurate numbers of impressions. A
third problem is the emergence of a new metric that should
be looked at, the duration of the display, which servers
cannot determine. Westrich also said, "timely and complete
reporting has been problematic" for companies measuring
through servers. Additionally, servers have no way to track
behavior within rich media banners and post click or ROI
analysis is difficult as well. Thinking Media's new Active
Track helps combat these shortfalls by adding small Java
applets to a web page, which send information to their data
center every time the ad runs. This process confirms the
actual ad views seen, the duration of each display, and the
number of cached views. Active Track also provides hourly
centralized reporting. According to Westrich, the system is
easy to implement and works with all types of banners,
including rich media. The biggest drawback of the system is
its dependence on Java script. They are able to use a
backup technology that doesn't provide the same quality of
data in instances where the browser is unable to read Java
or the user has it turned off. Another shortfall that
Westrich mentioned was that they could only work within
their own rich media.

Tim Meadows, Vice President of Marketing at NetRatings
offered information on the second type of browser centric
measurement, sampling. NetRatings has joined with Nielsen
and implemented a panel-based methodology to track web
advertising. Unlike the other methods discussed, their
system is not campaign specific. By independently tracking
who sees which ad they allow marketers to see where their
competition advertises, what creative they used, how many
impressions they received, and what click through
percentages they realized. The Neilsen/NetRatings targeting
feature is also very interesting. You can log in and enter
the demographics of your particular target audience. The
system will show you a list of the top sites that reach your
demographic. You can also see click through rates, which ad
viewers clicked on, and the demographic of the viewers that
actually clicked. Unfortunately, because this system tracks
everything the sample audience saw and reacted to on the
web, it may be more valuable as a starting point for the
campaign rather than a measurement tool. They can
effectively track what a viewer did once they clicked on the
banner, but they can only provide that information in
regards to the members of their sample. Meadows stated that
the limitation of their product is that they "have a sample
of the population...but the way statistics work, you need a
certain amount of activity to make it statistically valid."
This means that smaller campaigns will obviously NOT have as
accurate reporting as larger ones.

Finally, Jim D'Arcangelo, Senior Vice President of Marketing
and Business Development at Media Metrix discussed another
panel based measurement system. Media Metrix provides
measurement solutions based on specific campaign objectives.
D'Arcangelo said "for advertising agencies especially,
you're going to have to measure the whole cross dimension of
advertising effectiveness both online and offline." He
stated that companies must perform cross media measurement.
This means that they must track how offline advertising
affects online behavior, how online advertising affects
offline behavior, and all other combinations of the two. He
went through a couple of case studies to help drive this
home. If your campaign objective is simply to drive traffic
to the site, you could plot your page views by day and then
cross-reference that with the start and end date of all on
and offline advertising. This will show you the general
effect that different types of advertising have on the
objectives.

The session ended with an interesting question from Zinman.
He mentioned that "it kind of seems like Media Metrix and
Nielsen/NetRatings data is more used to understand your
business in context of other activity on the web. Whereas
the server side stuff is more used by a particular campaign
to understand, in detail, what is going on with that
campaign." Meadows agreed to a certain extent, but he also
explained that the server side measurements do not provide
the demographic data that the panel based methods do. So,
the question remains...Where should advertising activity be
measured? Should we track results from the server that
provides the ad or should we track results from the browser
that actually sees the ad? Obviously, there are pros and
cons to each side. I'd love to hear what some of our
members here on the list have to say about the issue!

=========================================================




Received on Fri Nov 05 1999 - 09:13:40 CST