Wednesday, September 16, 2009

Inflated traffic stats cloud pay-wall plans

Newspapers trying to assess the financial impact of potential paid-content schemes are starting with a wildly inflated sense of the size of their online audience that could come back to bite them in a big way.

In “nearly every market” included in a study of 118 newspapers of every size in every part of the country, Greg Harmon of Belden Interactive found that publishers on average report the number of unique visitors to their websites is 1.3 times larger than the population of their respective communities – and fully 10 times greater than their print circulation.

Those numbers are not just moderately overstated. “They are magnificently incorrect,” said Harmon, who presented his findings to an industry conference earlier this week sponsored by the American Press Institute.

Worse, no one knows what the true numbers are. And that’s a major problem, because:

Understanding the true size of news-site traffic is crucial to developing a model that will accurately predict the revenues – and the inevitable audience loss – that would result from any paper’s decision to charge for the content it now provides for free.

To illustrate the enormity of the miscalculation that could be produced by basing a paid-content model on the wrong audiene number, let’s consider the case of Newspaper X. The hypothetical paper serves a community of 1 million residents, has a circulation of 100,000 copies and claims 1.3 million unique monthly visitors at its website.

If the publisher of Newspaper X starts out with the belief that she has 1.3 million unique visitors to her site and generously assumes that 15% of them will pay $4.50 per month to read the news, then she will project that charging for content will produce $10.5 million in annual subscription revenue.

But what is the likelihood that NewspaperX.Com really has 1.3 times more readers than the number of people who live in its city? Common sense, as well as Greg Harmon, will tell you the chance is implausibly slim.

To develop a more conservative projection, the publisher of Newspaper X could trim the number of unique visitors to equal the 1 million people who live in the market. Based on the assumptions described for the first case, the paper’s annual online subscription revenue would fall to $8.1 million, or 23% less than would be produced by starting with 1.3 million unique visitors.

But let’s get serious. Could every man, woman and child in town really be logging on to the newspaper’s site at least once a month? Not likely.

What if the actual number of unique visitors to the newspaper site happened to equal the number of its print subscribers? If the true traffic to the site were only 100,000 unique visitors per month, then online subscription revenue would be just $810,000 per year, or 76% of the sum that would be produced by 1.3 million visitors.

Don’t like those outcomes? Pick any unique-visitor number you want. No one can argue with you, because publishers haven’t the slightest idea of what the right number might be.

The online traffic numbers cited by most newspapers today are generated by server -logging systems that are almost universally acknowledged to overstate the number of individual visitors.

But publishers to date haven’t worried about cleaning up their data, because the unrealistically high figures generated by their systems support the story they like to tell about the broad reach delivered by their combined print and online products.

Advertisers typically haven’t been fussed by the inaccuracy, because they buy print ads by the inches and online ads, for the most part, according to a certain number of page views. As long as the publisher states she delivered the promised number of inches and page views, the advertiser is happy.

Now that publishers are trying to decide whether and how to charge for access to their websites, however, it is absolutely essential for them to know true size of their audience.

This information matters not only for the purposes of projecting potential revenue but also – and perhaps more importantly – in guessing how many visitors would stick around if they suddenly were required to pay for the content they are accustomed to getting for free.

As Harmon has discovered in studying the website traffic at dozens of newspapers for several years, online news consumers are not created equal. The audience, he says, falls into three distinctly different categories:

:: 25% of the unique visitors consist of loyal readers who visit a newspaper site an average of 20 times per month and sometimes multiple times a day.

:: 21% of the audience comprises incidental readers who visit between one and three times a month.

:: 54% of the visitors are what Harmon calls “fly-bys” – people who may come once a month in response to a link on another website that steers them to the newspaper.

Given their low engagement with news sites, incidental and fly-by readers don’t appear to be prime candidates for paying for content. Further, it seems reasonable to surmise that most would spurn a site demanding payment, possibly never to return.

If you are a publisher trying to intelligently evaluate the business potential of charging for online content, you are left with a spreadsheet full of holes:

:: You don’t really know how many unique visitors you have.

:: You have to guess at the percentage of loyal visitors who will be amenable to paying for content.

:: You have to guess the price loyal visitors might pay.

:: You have to estimate not only how much web traffic you will lose but also how far your ad revenues will tumble in response to the almost certain decline in page views.

Put it all together and you can see why half of the publishers in the United States can’t decide what to do about paid content. Given the frightening uncertainty they face, it’s a wonder more of them aren’t straddling the fence.


  1. Very interesting. I forward it to some web-analytics colleagues to see their opinion. I will need more time to think about.

    thank you!

  2. We've found that the best predictor of online subscriptions for the 40+ newspaper websites we have is print circulation, not website traffic. The numbers are lower than I'm sure most publishers want but we see online subscriptions equal to 3% to 5% of print circulation.

  3. I think the newspapers forgot what they were and what they had that sold their product, REPORTING, just the facts -- fair, accurate, down the middle reporting of FACTS. Now that everything in the news is political, all people want is local ads, and those are online for all free for the taking.

    Do they actually think their cheer leading for the global warming hoax is helping them?

    Toast, and they did it to themselves, by flushing their own credibility down the toilet. Online or otherwise --- No difference.

  4. Why is it common sense that newspapers cannot have online audiences 1.3 x their local population?
    Local people are not the only readers - there are many many reasons why a local papers stories online may get national attention, especially for big events or funny stories.

    It doesn't make sense to attack the figures on that basis.

    I do agree that uniques are a poor measure of potential online subscribers though.

  5. I've started bitching about newspaper site traffic numbers in 2004 and mostly its fallen on deaf ears.

    The NAA pushed the "visited in 30 days" number which isn't a great measurement for figuring out if you're developing a loyal audience, a commercially valuable audience.

    I've studied stats from a number newspaper.coms and the best educated guess for local online reach on a daily basis is 2 to 4 percent of the adult DMA. Think about that: Many newspaper.coms are reaching only two percent of their market on a daily basis. How can that be a marketable number?

    I've spoken with Greg Harmon about this before, and his more studied efforts put the daily visit number in that ballpark.

    Also, the following paragraph is absolute bunk:

    "The online traffic numbers cited by most newspapers today are generated by server -logging systems that are almost universally acknowledged to overstate the number of individual visitors."

    I've yet to come across a newspaper site that didn't use a non-log analytics program, most commonly Ominiture, but also HitBox, Google Analytics or the stats program Town News provides to its 1,300 affiliates.

  6. I'm surprised that online subscriptions only account for 5% of total sales. Everything in this country is focused on convenience and most people have a work or home computer that they use on regular basis to keep up with major news issues. I remember the days when people looked forward to getting up on a Sunday morning and reading the newspaper. A sad turn of events.

  7. Perry Gaskill6:17 PM

    I ran some back-of-the-envelope calculations a few months back based on the NAA/Nielsen numbers and the rule-of-thumb indicators, in the aggregate, seemed to come back like this:

    Take market population and divide by four or 5 (20-25 percent) to get unique visitors.

    Average visits per UV per month will be 8.5.

    Average page views per visitor per month are 49.8.

    Time spent per UV is 44 minutes per month.

    What's unfortunate is that the Belden numbers are both way out of alignment with NAA/Nielsen and seem to be logically inconsistent internally based on readership patterns. The problem is that it would be useful to know, for example, that 25 percent of readers generate 86 percent of page views, but you can't really trust that metric if you don't believe a 1.3 population multiplier is accurate. I also tend to doubt if the 51 percent UV number is accurate in the aggregate; it may be true for the NYT or WaPo but very much not true for small and middle markets.

    For what it's worth, it's surprising to me that there hasn't been more done in the area of IP filtering or re-directs based on visitor locale. Although it may sound cool to have a global audience, can anyone explain on the ad side why a local hardware store owner would realistically expect to sell a hammer to a kid in Bulgaria?

  8. Mike Donatello6:49 PM

    No offense, but I am repeatedly dumbfounded by statements like, "You have to guess the price loyal visitors might pay."

    As either Greg (Harmon or Swanson) might attest, deriving an appropriate price isn't some black art. All it requires is a relatively modest investment in research and a willingness to adhere to some marketing fundamentals. But -- OOPS -- there's the industry's dirty wordy again: "investment."

    Anyone who believes that you can't do accurate pricing research needs an education, plain and simple.

  9. Anonymous6:50 PM

    From my own experience as a editor/publisher of an online newspaper, I found that fully one-third of my unique visits were from outside the immediate area. For one thing, once local people who moved away found out that could track local news, they became frequent visitors; another source of out of state readers were parents of the students at the dozen prep schools in my immediate area (I made sure to provide plenty of coverage of prep sports). One of hte real benefits of online distribution is that it VASTLY expands the potential audience, especially if publishers make an effort to go after "foreign" readers.

  10. Actually, you can extrapolate from known facts.

    Look at the sites that require payments now. I pay for WSJ, for example, but access it on my smartphone, home computer, and when I travel. They set a cookie and I have to sign on, but they also look at the server log and see that to the log I look like three or four people.

    Also, my wife signs in on the same computers, and several of her own... all under the same subscription, of course. How many subs do the publishers think each household is going to buy? Or even can keep straight as phones and laptops are passed around?

    Look at it another way: Boston (where, the Globe's site, is a leader in unique visitors) has close to 800,000 people, 250,000 households. But only a third take the Globe now. The rest of the paper's 300K+ circ comes from the surrounding metro area. All of that area, plus the next 20-30 mile ring where the Globe has forfeited readership because physical delivery is too expensive, are certainly worth promoting for online.

    Yeah, the calculations may be impossible for this generation of newspaper owners, but they employ circ folks smart enough to figure it out.

    BTW, global warming is real and largely caused by human activity.

  11. In the Netherlands there is an indepent organisation ( that monitors the traffic to big Dutch websites. It is a panel of around 10,000 people that are being monitored. It also undobules the statistics for people that visit a website both from their office and from their home. Simple idea, don't they have anything similar in the US?

  12. Linda6:43 AM

    Paid content is one side of this fence -- local advertising revenue is the other.

    Perry hit the nail on the head albeit via Bulgaria. Unless a paper tracks by zip code, most haven't a clue as to WHO in that mix is in their local market. (And please don't bring up registration data -- what I've seen there is worthless.)

    Most readership research, and ours included for 16 news sites, shows that the local online audience only adds about 3-6% reach in a local market.

    And yet, that is where the vast majority of revenue comes for 99% of the newspapers in this country.

    Our problem is that we continue to love big, mass numbers -- and we don't much care where they come from. It's like a drug. We haven't come to understand that the real value of the net is about WHO you have in your audience not how many. And not only do we not know the WHO, even if we did, we're not yet comfortable charging the true value of our WHO. So we return to what has worked in the past -- focus on how MANY instead.

    Think about it this way. There are many 30,000 circ newspapers in this country. And they've run very successful businesses based on reaching that 30,000. Now, all of a sudden, they can't figure out how to fun a successful business in that same market reaching 1.3 million? No surprise why. That 30,000 doesn't believe their 1.3 million audience number so while the customer may understand they are buying impressions, they have a clue as to WHO they are impressing. And try asking a sales rep to explain! It would be comical if it weren't such a damn shame.

    Scalejacking is a great description for what we are doing -- Seth Godin calls us on it in this blog attached.

  13. Anonymous7:22 AM

    Any single metric by itself will confess its flaws, if tortured enough (even "readership" figures)...

    Here's a recommendation: Look at your page views, ad impressions and (and year over year growth)... Estimate how much you might lose over time with your proposed version of a pay gate (there are many flavors--all with different effects on traffic & SEO)...

    Balance that against what you might gain on the sub side (and projected growth of subs over time)... Model it out over 3 years minimally.

    Assume that consumers will pay for some of your online content (until its disrupted by a good-enough, free, version).

    The "unique" argument might be a red herring, but it's a good reminder that focusing on a single metric is a good way to mess up the math.

  14. Anonymous11:23 AM

    If newspapers were to realize less revenue through a pay model [because they would be forced to recognize how inflated their numbers are]then it also follows that they need to fess up with their present advertisers. They are being fleeced. Let's see: much lower visitor numbers than advertised and the average person spends about 20 seconds on a newspaper's web site. And we're supposed to sell

  15. THE PRICING SHOULD BE EASY. Any newspaper wants big numbers of readers, and won't get themif they charge a lot. Start low, start slow, and promote outside of the paper, using billboards, cable tv, radio etc. it is always easier to raise the price of subscription once peopleare highly satisfied.
    The real money here is charging 4 online ads that work.

  16. Anonymous11:19 AM

    I've found that most newspapers don't understand the numbers that their analytics provide.

    Many of the "visitor" numbers are actually visitor SESSIONS. A session is typically counted as series of hits or pageviews within a 30 minute period.

    So, if I come back to the site twice a day, every day for 30 days that's 60 "visitors", not unique people.

  17. To Anonymous,

    You are confusing "Visits" and "Visitors". While I don't argue that there may be some analytics software out there that do not make this distinction, the most common softwares utilized by newspaper web sites do.

    Please see this excerpt from the help section of Google Analytics:

    "Analytics measures both visits and visitors in your account. Visits represent the number of individual sessions initiated by all the visitors to your site. If a user is inactive on your site for 30 minutes or more, any future activity will be attributed to a new session. Users that leave your site and return within 30 minutes will be counted as part of the original session.

    The initial session by a user during any given date range is considered to be an additional visit and an additional visitor. Any future sessions from the same user during the selected time period are counted as additional visits, but not as additional visitors."

    Traffic to a newspaper web site outside the local area still is valuable, but can depend on what items are attracting those outside visitors. Consider the importance of obituaries, legal notices, or pictures of a grandson, among others.

    Most of us are also using ad delivery software that can restrict a local mom-and-pop's ads to specific zip codes. A good sales team will be selling to National media buyers - how many newspapers are lacking that sale? Many local small businesses do sell outside the area - check to see if your audience fits their needs. Leftover page views can be served to a variety of ad networks. Have a lot of page views to a particular remote geo-location? See if another newspaper local to that area would be willing to sell the inventory for a revenue split.

    How about the value of the advertising to remote viewers by local hospitality, or the local visitors bureau, or the local woodworker that has dozens of birdhouses and picture frames that can be shipped anywhere. Or how about the realtor that specializes in relocations? There are many more.

    Just because that 1.3 million page views you mention isn't all local doesn't mean the delta is worthless. I just don't buy this argument that non-local traffic has no value. The inventory is here, and it has value, but to make anything of it we have to think of ourselves as more than a product for the physical bodies inside a local perimeter.

  18. Inflated traffic stats? Have we forgotten that this is the WORLD-WIDE-WEB? Traffic being 10x the market population is no surprise, sideways audience to a news site can be quite sizeable.

    @Stephen Larson referenced a common and even proven method for guestimating online subs using the print circulation number. The common figure is .5 to 5% of print circulation are targetable ranges.

    Belden/Itz seems to know something about analytics and audience measurement that the big boys (Google, WebTrends, Omniture, Quantcast, Neilsen, ComScore etc.) don't. Maybe they should call and pitch their methodology?

    Survey-based behavioral analytics are a research dinosaur. Stop looking to them for answers.


Note: Only a member of this blog may post a comment.