NONE: ONLINE-ADS>> WEB-AD'98 COVERAGE: 2/12/98 - HOW YOU'RE GETTING
ONLINE-ADS>> WEB-AD'98 COVERAGE: 2/12/98 - HOW YOU'RE GETTING
richard_at_tenagra.com
Fri, 13 Feb 1998 07:18:08 -0600 (CST)
SCREWED, #5
Sender: owner-online-ads_at_o-a.com
Precedence: bulk
Web Advertising '98 Email Coverage
HOW YOU'RE GETTING SCREWED (or The Truth About Measurement), rpt #5
February 12, 1998
This is the fifth in a series of 10 reports from Richard Hoy, who
is covering the Web Advertising '98 conference in New York this week.
You will receive these reports in addition to your normal Online Ads
posts/digests.
This coverage is archived at:
http://www.o-a.com/webad98/webad98-archive.html
======================================================================
======================================================================
This conference coverage is generously underwritten by:
--CLICKTRADE--
http://www.clicktrade.com/webad98
Link Partner Programs to reward other Web sites for linking to yours.
Joining ClickTrade is FREE. As a member, you can be an advertiser,
link partner, or both. Advertisers reward link partners on a
pay-per-click basis for linking to their site. Link partners earn
revenue for sending traffic to advertiser sites.
**********************************************************************
Advertiser Benefits: Link Partner Benefits:
-------------------- ----------------------
-Drive traffic to your site -Make money
-Save time; outsouce the process -Select link type(text, graphic, java)
-Flexibility of payout rates -Control where to place links
======================================================================
======================================================================
HOW YOU'RE GETTING SCREWED (or The Truth About Measurement)
Brad Aronson, i-Frontier ( brad_at_i-frontier.com )
Ever get post-campaign reports and the numbers just look weird? It may
not all be in your head.
"When we use media site reports, we've found the reports vary wildly
depending on the definitions the sites are using," explained Brad
Aronson, president of i-Frontier.
Brad detailed several examples of artifacts that contribute to this
wild fluctuation. He cited a rather interesting case that happened to
BPA, a major auditing service, involving spiders - the automated
programs search engines use to catalog site. BPA was doing a
three-month audit on a site and discovered a huge surge in traffic in
month three. The reason was that someone just happened to be
repeatedly testing a new spider on this particular site. Brad pointed
out that BPA only noticed the spider because of the irregularity in
traffic. But what if the spider was hitting the site all during BPA
audit? They might have reported inflated numbers for the traffic.
Brad cite another case that involved frames and page views. The
Internet Advertising Bureau (IAB), the trade organization trying to
standardized measurement, defines a pageview as all the elements that
constitute a page when you look at it. So by the IAB's definition a
page comprised of three frames is one pageview. Brad has advertised on
sites that have told him they count each frame as a pageview. This
practice inflates your numbers and looks like you have greater reach
than you really do.
One of the biggest culprits messing up accurate measurement is caching
by proxy servers. Brad made a great point about how caching effects
might really skew a media buyers evaluation of buys on two different
sites. Last year AOL released data regarding its caching practices to
its content partners. One of those partners is Warner Bros., who
reported to Ad Age that the Warner Bros. Web site traffic was actually
30% higher than previously believed. This was because this previously
hidden 30% was occurring on AOL's servers, and was therefore
unmeasurable on the Warner Bros. site. Now if you are evaluating buys
that include the Warner Bros. Web site, can you fairly compare that
site to others?
Caching isn't the only problem plaguing the evaluation of a site's
reach. The way measurement companies determine the top visited Web
sites are very different, but seemingly equal in validity. Media
Metrix, for example, uses software place on the machines of a sample
group of users. Hot100, on the other hand, aggregates data from
several major ISPs around the country.
Then there is the problem of how sites define an ad impressions.
"What is an ad impression. Or an ad view. Or an ad request," said
Brad. "Everyone's calling different things. However, from what I'm
seeing, when people do reporting for clients they just put it all
under a heading call 'impressions'."
Here is how some major organizations define an impression:
what they call it definition
----------------- ----------
I/PRO Ad View the number of times
an ad is requested
IAB Ad Request the number of times a
page with an ad is requested
AdKnowledge Impression the number of times 100% of
ad graphic is downloaded
"Depending on who is giving you your data, you can't really compare
what is going on," said Brad.
It gets even more murky if you are dependent on ad server software to
tell you what is going on. If a page is delivered to a user and that
user has the graphics turned off (so your banner was never
seen/loaded), the Accipiter ad server package would count that as an
impression. If you're using Netgravity, the count varies depending on
what software version you a using. If you are using the version of
Netgravity for a stand-alone site, then it would count the
"graphics-off" scenario as an impression. If you are using the version
of Netgravity for ad networks, then that same scenario would not be
counted as an impression.
Another interesting tidbit Brad shared was artifacts in click-throughs
sites may report. If people click on an ad then bail because the site
is too slow or their ISP is overloaded, that is still reported as a
click-through. Or if a user clicks more than once because he or she
didn't think it took. Brad said a top 10 publishing company told him
that click-throughs are consistently off by 5-7% percent from these
sorts of artifacts. And a top tracking company told him it was more
like 10% of the time. He showed a couple of examples from campaigns he
ran where the site reported 100 click-throughs, but the client's logs
showed only 50.
Brad described how you can use inconsistency in reporting to your
advantage. For example, a major publisher that i-Frontier works with
recently change their counting methodology from every time the banner
loads to every time the page loads. This means impressions are being
counted even if a person has the graphics turned off. Brad used this
as leverage to get a lower CPM.
Brad suggested that in order to measure your site performance
effectively, you should do the following:
1.) determine what your goals are (sales, leads, brand building,
advertising?)
2.) determine the criteria that is most appropriate to measure
3.) agree on a reasonable benchmark
4.) choose the right tools
5.) analyze the data and act on that analysis
Some practical ways to use log data include:
--altering navigation to improve the flow of traffic to specific parts
of a site
--looking at where your traffic is coming from, which might suggest
places to advertise and give you insight into the quality of your
audience
--deciding on site features/enhancements based on the type/version of
browser most of your audience is using
--locating the best placement for ad banners or buttons
--scheduling updates before peak traffic times so lot of people see
the new content
--scheduling maintenance at off-peak times so you disrupt the traffic
as little as possible
Brad wrapped up his talk by touching on the issue of cookies and gave
some numbers, reported by Ad Age, on the cost of implementing cookies
for specific functions.
Median cost of using cookies for:
delivering a different home page to new users versus repeat users
national median cost: $2,250
storing registration information to access password-protected content
national median cost: $13,000
recommending content/products
national median cost: $42,500
This ends the fifth report from Web Advertising 98. Stay tuned for
more detailed session synopses.
======================================================================
======================================================================
This conference coverage is generously underwritten by:
--CLICKTRADE--
http://www.clicktrade.com/webad98
Link Partner Programs to reward other Web sites for linking to yours.
Joining ClickTrade is FREE. As a member, you can be an advertiser,
link partner, or both. Advertisers reward link partners on a
pay-per-click basis for linking to their site. Link partners earn
revenue for sending traffic to advertiser sites.
**********************************************************************
Advertiser Benefits: Link Partner Benefits:
-------------------- ----------------------
-Drive traffic to your site -Make money
-Save time; outsouce the process -Select link type(text, graphic, java)
-Flexibility of payout rates -Control where to place links
======================================================================
======================================================================
SUBSCRIBE INSTRUCTIONS
To subscribe to the Online Advertising Discussion List,
Fill out the form at: http://www.o-a.com/
UNSUBSCRIBE INSTRUCTIONS
You may leave the Online Advertising Discussion List at
any time by sending an email message with the word UNSUBSCRIBE
(just the word) in the body or content of your message to:
online-ads-digest-request_at_o-a.com
if you get the digest version, and
online-ads-request_at_o-a.com
if you get the non-digest version.
Online Advertising Discussion List
http://www.o-a.com/
a publication of The Tenagra Corporation
http://www.tenagra.com/
