This is Part I of a two part article.
Part
I: The Why's and What's of Auditing
Part
II: The Audit Process
Why
Audit The Numbers?
If
you're ever feeling really and truly mean, here's a nasty trick to play
on your marketing and ad sales colleagues. Wait, with a friend, until
some of these folks get on the elevator, and follow them in. Here's what
you say to your companion in crime: "I just heard that there was a real
glitch in the traffic numbers and the last six months worth were all twice
as high as they should have been." If you don't understand why this is
mean, read on.
Many
content sites are in the business of selling just one thing - numbers
- to a very demanding audience - advertising executives. These websites
range from portals like Yahoo! through specialists like ESPN or ABCNews
all the way to search engines like Links2Go and thousands of others. The
economic job of such sites is to turn marketing dollars into reports that
show how many surfers visit the site or view the ads. While this is not
the only way that web sites earn revenue, it represents the majority of
revenues for a great many sites.
There
is an easier way to create those reports than to build the website and
spend all those marketing dollars: just make them up. Unfortunately, advertisers
got wise to that long ago, when information was disseminated through newspapers
and magazines - before television and radio, even. If you've ever enjoyed
a television show that got cancelled because of "ratings," you know how
important audience measurement can be.
Over
the years, the publishing industry and the entertainment industry developed
standard ways to determine these ratings. These standards include both
definitions of what to count and procedures for doing the counting. Take
something as simple as a magazine. Obviously, each subscription counts,
as does each newsstand sale. There are, naturally, various ways to cook
these numbers. For example, given the relatively low cost of printing,
it might be worth it for publishers to give out gift subscriptions by
the truckload to improve their numbers. So auditors look for evidence
that subscriptions were paid for.
But
there are other complications besides fraud. Publishers argue that magazines
often get passed around in an office, or between friends; they'd like
to count the extra readers when that happens. So auditors have developed
procedures for measuring that kind of readership fairly. On the other
hand, advertisers want to be sure that ads are seen by the right people,
those who are likely to respond by making purchases. The Beef Council
doesn't advertise in Vegetarian Times for obvious reasons.
But it's less obvious whether they should advertise in Newsweek,
or TV Guide, or X-men comics, or the Ann Arbor
Times.
Once
advertisers caught on to the Web in the latter part of the 1990's some
were ecstatic. "At last! A measurable form of advertising!" they thought.
And to some extent they were right. But measurement is neither easy nor
foolproof even in this brave new medium, and the responsibility for those
measurements to be accurate falls largely on the IT department. The responsibility
is immense, because it reflects directly on the corporate bottom line.
Remember that little elevator prank we suggested? If the numbers reported
by the website turned out to be grossly incorrect, the site might have
to give back money to the advertisers. Sometimes lots of money! It's rare
that this happens, and rarer that the amounts are very large, but it does
happen. And the blame, as you might guess, falls squarely on the technologist.
So
in the next sections we'll explain the different kinds of numbers, how
they are collected, what the situation is regarding audits and what to
do now to ensure that your numbers pass muster when the auditor comes
to call.
Traffic,
Ads, Clicks and The Right Sort of People
Most websites get paid for ads based on the traffic they bring to see
them. The basis might be the count of page impressions, the number of
pages on which a particular ad is served, the rate at which people click
on the ad, or the degree to which people who click on the ad do something,
such as buy product or register or request more information. It could
also be related to the kinds of people who see or do those things
Advertisers
are for the most part interested in three kinds of numbers. First, there
are traffic numbers. These are the counts of the number of pages served
from the site, the numbers of unique visitors, and the numbers of visits.
(See Counting
Website Traffic - The Skinny On Hits, Impressions, Visitors and Clickthroughs).
While
it may be enough to report simple totals on a weekly or monthly basis,
in some cases a finer level of detail may be necessary. It all depends
on your site's ad model. The ad model is the collection of different packages
that your salespeople have for selling ads. Let's illustrate with an example.
The web site associated with the President of the United States does not,
of course, accept advertising but it can be used to illustrate some different
kinds of ad models.
Which
kinds of ads might an advertiser be able to buy at whitehouse.gov? There
would probably be a banner on the front page. Some of the sections, such
as White House History or Gateway to
Government could possibly be sponsored, with a button or text ad
naming the sponsor. They might also accept banner ads that run just within
the section. This is because people who visit White House
History might be seen as being of interest to certain kinds of
advertisers - antique dealers or travel agencies for example - while visitors
to Gateway To Government might be seen as more likely
to patronize advertisers for foreign policy journals.
There
might be other considerations as well. A surfer who visits White House
History at 3 A.M. might be a citizen of a foreign country, and the advertisers
who'd want to reach such a person could easily be different from those
looking to reach the 3 P.M. visitor. Perhaps surfers from certain foreign
countries shouldn't be shown ads for advanced armaments - or those from
other countries should definitely be shown such ads.
This
website, like many others, also has a search engine. Advertisers often
want to buy ads that show up on search results pages, based on the search
term. The logic is obvious: a surfer who searched on, say, "flags" is
more likely to be attracted by ads for "flags" or "flagpoles" than is
one who searched for "Lincoln bedroom."
An
ad model consists of the options that are provided to advertisers for
purchasing ads. Most options are defined by a collection of pages, times,
and surfer characteristics. Another kind of option in the ad model is
the search term, again modified by times and user characteristics. Options
also specify the basis of charging. An advertiser might own all of the
banner impressions on a page, or pay by the number of times an ad is served,
or pay a fixed price for a number of impressions. It can get more complicated,
but we've covered most of the cases.
Traffic
numbers are important tools in the salesperson's arsenal. Before buying
an ad, advertisers want to know how many people will be eligible to see
the ad. If the advertiser is interested in buying a banner on the home
page, the obvious question is "how many people see the home page." If
the advertiser is being sold an ad that will be presented only to surfers
from the South Pacific who visit the White House Tour section between
the hours of midnight and 5 A.M., you'd better know how many South Pacific
islanders people actually surf those pages at that time. More specifically,
it is the salespeople who have to provide the information. If IT is involved
in the initial development and review of the ad model you'll have the
chance to make sure that the right numbers are collected in a form that's
easy to generate reports from. If you aren't involved in design then be
prepared for late afternoon calls for traffic measurements that aren't
normally made - and may require special programming to deliver.
Once
an advertiser decides to advertise, the interest turns from the potential
to the real. How many people actually saw the ad? How many people actually
clicked on it? Where the numbers about potential viewers probably come
from the webserver's traffic logs, sometimes supplemented by your user
database, the numbers about actual ad impressions come from the logs of
the adserver (see How
to Serve an Ad: Better to Buy Than to Build). The number of clickthroughs
can be determined by the advertisers. However, you will need to count
these yourself as well, to protect against advertisers claiming rebates
because the clickthrough numbers were lower than expected.
With
all these complex numbers, and thousands of dollars on the line from every
advertiser, why should an advertiser trust you to deliver accurate numbers?
After all, you were willing, just a paragraph ago, to not trust them to
report clickthrough numbers accurately. In the world of advertising you
may not trust, but you always verify. That's where the audit comes it.
With
all these complex numbers, and thousands of dollars on the line from every
advertiser, why should an advertiser trust you to deliver accurate numbers?
After all, you were willing, just a paragraph ago, to not trust them to
report clickthrough numbers accurately. In the world of advertising you
may not trust, but you always verify. That's where the audit comes it.