As a metric, unique visitors are uniquely inaccurate

Brass Tacks Design

For years, newspapers have touted their millions of page views and thousands of unique visitors in an attempt to convince advertisers that newspaper sites could be an effective advertising medium. But many of these uniques are mere phantoms and a third of the pages counted as page views are hardly pages at all. More on that later.

But first, ever wonder why newspapers never tell advertisers how often their unique visitors actually visit? Because the national average is two times per month, if you calculate using the numbers published by NAA.

That's right – two times per month. Where did I get data for this calculation? From Nielsen/Netratings – "The global standard for Internet audience measurement and analysis."

While I believe that newspaper sites get more than two visits per visitor per month, that's not what the numbers say.

But before I go any further, let me make this caveat crystal clear: I am not saying all the numbers are wrong. I am saying some of the numbers are wrong – specifically the unique visitors. So don't email me to say that page view stats are correct. I agree – they are correct. But newspapers and many other sites are publicly using their errroneous unique visitor stats as a means of selling advertising. And that is wrong.

Journalists hate math, but they love the truth. So let's endure some arithmetic to get to the bottom of this.

Let's start with The New York Times. According to Nielsen, the Times had 388,090,753 page views in November 2006. If you divide this number by the 29.33 pages viewed during each visit, you'll find that NYTimes.com was visited 13,231,870 times in November. Divide the number of visits by the number of unique visitors to determine the how often each visitor came to the site: 13,231,870 visits ÷ 13,232,524 unique visitors = .99 visits per visitor per month.

How can that be? Each unique visitor to the Times online only visits once a month? That's impossible!

While I don't believe NYTimes.com gets 29.33 views per visit, that's what it says in the Nielsen report that any advertiser can read.

And I believe NYTimes.com gets more than one visit per visitor per month, but that's what you gotta believe IF you believe the numbers – and I don't.

Numbers don't lie. (Actually, they do lie when the data collection method is flawed – as I'll explain shortly.)

Let's apply this same formula to the St. Petersburg Times, using the same data from Nielsen. Here's the formula in two parts:

Total page views per month ÷ Page views per visit = Visits per month
Visits per month ÷ Total unique visitors = Visits per month per visitor

Here's the formula in one statement:

Total page views per month ÷ page views per visit ÷ unique visitors = Visits per month per visitor

Now let's plug in St. Pete's numbers:

15,543,600 page views per month ÷ 12.34 page views per visit ÷ 1,259,558 unique visitors = 1 visit per month per visitor. I believe St. Pete's site gets more than one visit per visitor per month, but that's not what the numbers say.

Download these stats and do the math on the top 100 websites. The results will vary, but not by much.

Of course, newspaper sites aren't the only ones perpetrating this misconception. Virtually all sites sling the same hash – newspapers are just easier to catch in this sham.

But wait, it gets worse. According to Greg Harmon of Belden Associates, thirty to forty percent of page views come from the classified advertising portion of newspaper websites. So the news pages aren't providing as much exposure for those banners and buttons as newspapers are leading advertisers to believe.

Here's the deal: Despite the hundreds of awards they've received, newspaper websites are an abysmal failure. To wit: Newspaper sites have lost eight percent of local online advertising in the past two years, according to Gordon Borrell of Borrell Associates, Inc.

But back to the stats. Let me explain why all these numbers are hooey – and the people pushing them are hooey-er. The problem begins and ends with the way sites measure "unique visitors."

Newspapers use two inherently flawed methods to determine unique visitors. The first method relies on registration. This method is flawed because online sites don't require registration for their homepages – making it almost impossible to tell how many uniques visit the homepage, then exit the site. Without registration on the homepage, we can't tell whether five people visit the homepage once a day, or one person visits fives times a day.

The second, and more deeply flawed method used more pervasively, relies on cookies and IP addresses. This method assigns a unique number to each computer and assumes each visitor uses this computer exclusively, rather than one computer at home and another at work. If you visit a website from two different computers, you are counted as two unique visitors. This phenomenon alone could account for almost a doubling of unique visitors.

This method also assumes that cookies are never cleared, either manually by users or automatically when security software is installed. It also assumes you'll use one computer, and only one computer, for your entire lifetime, unless the site "zeros out" all its uniques with some frequency.

Each time cookies are cleared, a new unique visitor is "born" and the rolls of unique visitors swell beyond all reason. For instance, the St. Pete Times claims 1,259,558 unique visitors and The Tampa Tribune claims 637,247 uniques for a total of 1,896,805. The population of the Tampa–St. Petersburg–Clearwater MSA is 2,395,997. What's the likelihood that 80% of the men, women and children in Tampa Bay visit one of these two sites?

Webmasters will tell you that "out-of-town visitors" inflate the ranks of unique visitors. Even if that were true, how much do local advertisers care about out-of-town uniques? Not at all.

So what's to be done about all this? Lots:

Stop reporting hits, page views – and especially uniques – which I think everyone agrees are impossible to determine from server logs alone. Counting the number of times an online ad appears on a screen is no more meaningful than counting the number of times that an ad appears on newsprint – except in print, someone paid for the privilege to see it, and those people really do count.

Until we know how many people visit two or three times a week instead of two or three times a month, we have little to encourage businesses to advertise with us.

There's only one metric that matters – online revenue – which as a proportion of total revenue remains in single digits at every newspaper in the U.S.A. – except those that have taken a big hit in print advertising. If print is the leaky lifeboat to online profitability, then we better start bailing and paddling as fast as we can, because at this rate, it will take a long time to float our online boats.

Newspapers need accurate data.
Belden Associates can determine a newspaper site's unique visitors based on log data, intercept studies and telephone interviews. It's a much smaller number than newspapers report today, but it's an accurate reflection of audience. Borrell Associates, Inc. can show you how many local online dollars are slipping through your fingers and where to focus your sales effort.

Newspapers need to redesign their sites from the ground up to become effective environments for advertising.
Brass Tacks Design has provided the most dramatic and forward-looking solutions for classified and display advertising online.

Online ads must be designed for the online environment.
Greg Swanson of ITZ Publishing has seen the future and can make it available to you today for 40 bucks an ad – a bargain at twice the price if you ask me. Then he can teach your staff effective sales techniques.

Newspapers need to learn how to sell online advertising, with an emphasis on self-service, lead development and cost-per-action.
CPM is DOA. Read API's Newspaper Next to learn more.

And consider this from David Evans at Forbes.com:

Traditional media sell advertisers a pig in a poke. Advertisers don't know whether a reader actually looks at their ad much less buys anything as a result. And they can't really target their ads beyond picking a type of newspaper and section to focus on in the hopes of reaching a particular demographic group.

Google and its ilk only charge advertisers when a viewer clicks on the very page containing their ad and perhaps, in the future, only when the viewer actually buys something. Plus, they can use all that information collected from past searches and other information they've gleaned about viewers to target ads with an increasing degree of accuracy. The technology is a different as a Schwinn one-speed bike is from a Porsche 911 turbo.

Class dismissed. Now get to work and save the newspaper business.