Doing a number on Web site traffic

In 2004, new kid on the block MySpace was watching its user traffic rise fast. But executives were having a tough time luring new advertisers to the social network because it didn't show up in the lists of top Web sites compiled by Internet audience trackers ComScore and Nielsen Online.

"We had a hard case telling people how big we were and how many unique visitors we had. They didn't believe us. We didn't show up on people's radars," said Jason Feffer, former vice president of operations at MySpace who is now president and chief executive at opinion-forums site SodaHead.com. "It was frustrating that advertisers wanting to advertise to that demographic would go to Friendster when we were 10 times bigger than them."

That changed fairly quickly for MySpace. But the problem persists for many other Web sites competing for advertising dollars in an environment where the difference between being ranked fourth or fifth in a category can directly impact the bottom line. Being undercounted by the audience measurement firms costs publishers advertising deals and threatens their ad-based businesses.

For instance, MLB.com, the site for Major League Baseball, has repeatedly complained that audience figures reported for the site by Nielsen Online are much lower than those shown by MLB.com's internal data. League subsidiary MLB Advanced Media has asked Nielsen Online to stop reporting on its traffic because of the discrepancy.

How big of a difference is there? Nielsen Online reported MLB.com with about 6.2 million unique visitors in December while MLB.com recorded 19.4 million unique visitors during that month. For 2007, Nielsen Online's figures were one quarter those recorded by MLB.com, according to MLB.com.

"That's like a pilot of a plane landing in the wrong continent," says Bob Bowman, chief executive of MLB Advanced Media. "There are probably some advertisers or agencies that actually believe the numbers and are steering their ad buys elsewhere."

Nielsen Online is responding to MLB.com's concerns, said Mainak Mazumdar, vice president of measurement science and research service MegaPanel at Nielsen Online. "The best way to address this is to understand what the methodology is and the audit will show that."

An old but tenacious conflict
This isn't a new problem, but, it is hoped, it is one that will be addressed this year. The Interactive Advertising Bureau, which represents more than 300 Web publishers, including CNET Networks, asked ComScore and Nielsen Online to submit to audits by the Media Rating Council, a nonprofit group that approves audience ratings systems. In a letter sent last April to the heads of the two companies, IAB Chief Executive Randall Rothenberg complained that they were "still relying on panels, a media-measurement technique invented for the radio industry exactly seven decades ago, to quantify the Internet."

"To persist in using panels that undercount or ignore the diverse populations that are the future of consumer marketing is to deny marketers the insights they need to build their businesses," Rothenberg wrote. "And it certainly appears to us as if they are being undercounted or disregarded, for our members' server logs continue to diverge starkly from your companies' sample-based assessments, by 2x to 3x magnitudes, in some cases far beyond any legitimate margin of sampling error."

Nielsen Online and ComScore say the audits are on schedule to be completed by the end of the year, said David Doty, a marketing and public relations senior vice president at IAB. Meanwhile, the IAB is also working on guidelines for publishers for measuring their own traffic. Those guidelines should come out in the second quarter, Doty said.

The IAB has become a "referee to an increasingly ugly street brawl," said Tim Hanlon, executive vice president at Denuo, the media futures arm of ad firm Publicis Group.

Third-party audience measurement provides a measurement independent of Web publishers who have financial incentive to inflate their numbers to ad buyers.

The numbers will never be exactly the same because the methods of measurement are different and they measure different things. ComScore and Nielsen Online use data gleaned from samples of Internet users who have monitoring software installed on their computers, while publishers are looking at their actual server log data. Panels give a glimpse into who is visiting a site; server data shows exact visits to the site.

To some, the debate comes down to which is more accurate--a sample panel whose stats are projected out to represent a wider population, as with television and radio panels, or server data that shows the number of visits to a site, the number of page views, and the time spent?

[inachu]

MLB info

Every pc at work I have seen that had malware on it only had cookies from MLB or websites from India or internal website.

Since our server was clean that leaves india webservers or MLB and I blame both and put a ban on both. Since then my network is nice and clean and trojan free.

[jchasin]

A View From the Other Side

Hi. I'm Josh Chasin, Chief Research Officer for comScore. A couple of things.

1. We always hear that site centric data actually reports a census of visits. The thing is, not all so-called "visits" are associated with visitors. Some quick examples: non-human traffic; RSS calls; tabbed browsing (if a user opens a browser and 5 tabs open simultaneously, each site logs a server call and thus a visit; us panel companies only log a visit when that darned human using the computer visits the site, by making the tab active.)

2. The biggest disparities between site and panel data are in the area of Uniques. There are many reasons why the two data sources differ, but it is instructive to understand the impact of cookie deletion, which we documented in a White Paper (go to our home page; you can't miss it.) Essentially, the disparity between site centric Unique counts and panel-derived Unique estimates can be pretty much entirely explained by the inflation of site centic numbers caused by multi-counting the same visitor if he/she shows up again having deleted the site's cookie.

To be clear: the differences in Uniques are a function of publisher data classifying a return visitor as a new one.

(Indeed, average daily Unique data from panels and site logs is far closer, because the impact of cookie deletion grows over the course of a month and is almost negligible for a day.)

3. On exclusion of cell-only households-- I agree. This is an egregious problem. But it only affects Nielsen, not comScore; in fact 19% of our panel comes from cell only households. I hope MLB finds our data to be more to their liking than that of our competitor. (And may I just say, Let's Go Mets!)

Cheers.

--josh--

[joebuff75]

Quantcast delivers!

I've become a big fan of QuantCast in recent months because it's
a free service, it uses a panel and has its own tracking codes for
a more accurate analysis. The stats are pretty close to my own
server logs and almost exactly the same as Google Analytics -
give or take a percent or two.

From my understanding a visit is counted in minutes and
pageviews. When somebody opens several browser tabs from
the same site, then the number of pageviews per visit might be
inflated but the visit length remains unaffected, and therefore
the overall visit number is accurate.

Before the dot-com bubble, I was able to afford getting traffic
stats through large companies (PCData) but it cost $30000+ a
year. QuantCast gives me the same information and more for
free. Back then, it seemed that you would only be included in
"official" traffic rankings when you were also a paying member -
- otherwise small networks were excluded from those ratings as
the audit companies had no way to link various sites to the same
publisher.

QuantCast allows you to group all your sites together to form a
network account and get aggregated traffic stats. This is very
important for small publishing houses with many sites. Using
this approach, my network of sites has a traffic rank of around
7600 of all web sites (1.1 million uniques/month:
http://www.quantcast.com/p-6btf0Pyh9xonk).

There is still room for improvement on QuantCast, though:

- calculating the ranking on a worldwide basis instead of
ranking according to U.S. traffic.

- a filter to exclude the sites in the adult business (to get a
ranking of advertiser friendly G/PG-rated sites)

- creation of sub-categories and filters, e.g. ranking by
male/female visitors (my audience is 75% female) which would
allow advertisers to easily drill down in the rankings to find
suitable sites.

I guess those features will be delivered over time... advertisers
should complement the data they get from comScore, Nielsen
etc. with other sources like QuantCast, Hitwise, Compete, Alexa
to get an overall picture of a site's traffic. Another possibility is
to check the Google Adwords traffic prediction tools for sites
that accept Adwords ads...

The more options there are, the better the understanding of
traffic patterns becomes...

Joe.

[modthocn]

But the problem persists for many other Web sites competing for advertising dollars in an environment where the difference between being ranked fourth or fifth in a category can directly impact the bottom line.

Being undercounted by the audience measurement firms costs publishers advertising deals and threatens their ad-based businesses. free psp games