Part 2: Times Readership Numbers

What happened to the audience during the two years of TimesSelect?

Google reports that there were 4.8 million references to "New York Times" in the blogs over the two years. If we assume that these posts generally referenced  one of the 350,000 published articles over the two years (at least, to the same degree that a reference of "Frank Rich" referenced a particular column– a BlogPulse trend graph tends to confirm a weekly spike for a weekly column), then we conclude that the averages references per article is 14. One can infer that the columns, at least to the opinionators in the blogosphere, are ten times as popular. Still, the 177,169 references only represented 3% of the blog buzz to the Times. Suppose we double this to account for the TimesSelect columnists in the Other sections, and then apply our 20% loss, we then conclude that the perceived audience drop for as a whole was 1-2%.

That's a very rough estimate. There are people paid to make better estimates, and fortunately we have their data.

Readership Growth

Zubin Jelveh is not one of them. That is, he is paid to be a journalist; not necessarily to make better estimates. Portfolio, a newly launched magazine from Conde Nast, hired Jelvah to produce a column  Chartistry, a "one-of-a-kind melding of infographics and economics insight." The column lasted three months then Jelvah transitioned to a blog called Odd Numbers. Last week,  Jelveh looked at the traffic data from comScore and compared the page views to a composite of competitors. He figured that the Times did not grow as fast as the industry average, pegged that directly the TimesSelect service (leaving unexplained how it was that the had equivalent traffic to the composite from September 2006 to March 2007), and then calculated that missing growth translated to 1.3 billion missing page views over the last two years (given an average of 400 million page views/month, this puts the readership loss at 13%).

But Jelveh limited his source data to the comScore Media Metrix. The Nielsen/NetRatings numbers tell a different story.

Every month Nielsen/NetRatings puts out ratings on the top 20 news sites, and they would not be easily accessible online had Jonathan Dube not regularly posted the data on his blog at, which is partnered with the website of the Online News Association, of which he is Vice President (and of which I am a member). I entered the last 5 years worth of data for a the Times and a few of its competitors, along with N/NR's "Total" news readership. I defined a "news year" from October to September, since this roughly coincides with the TimesSelect period, and averaged the data over that year. (The data for 2007 is incomplete, as Dube is missing January, hasn't posted the August yet, and September's has yet to be released.) I then averaged the data over each news year. The column marked 2003-4,  should be read "The percentage growth from the twelve month period ending Sept 2003 to the twelve month period ending 2004.")

All numbers are in thousands. (see the source data) [update 10/16/07: added data from Sept. 07]

 Oct 2002-Oct 2003-Oct 2004-Oct 2005- Oct 2006-  2003-42004-52005-62006-72005-7 
New York Times 8,6229,15510,55912,00613,402 6%15%14%12%27%
Washington Post 5,9875,6427,1587,9278,341  -6% 27% 11% 5% 17%
Gannett 8,10310,76811,52812,77812,594  33% 7% 11%-1% 9%
CNN 19,26222,22122,563 23,161 27,550  15%2%3%19% 22%
Yahoo! News  16,45420,430 23,210 26,969 31,816  24% 14% 16% 18% 37%
Total News Readers 65,616 74,070 78,632 85,318 93,978  15% 6% 9%10% 20%

What surprised me was how jittery the data is, even averaged over a year. The Post's readership during the first six months of the Iraq war was nowhere to be seen in the comparable months during 2004. Even October 2004 showed less than October 2002.  The Post recovered in November with the election; the coming year saw the revelation of Mark Felt as Watergate source "Deep Throat," and the extensive reporting on the Jack Abramoff scandal, which won the Post a Pulitzer Prize. By the end of September 2005, the Post had an audience size 2/3 of the Times's.

Yet, in the two years starting October 2005, the Times defied the accepted wisdom and increased readership 27%.  This is larger than the growth at the Post, the Gannett, CNN, and the NetRatings tally of all news visits. It is also larger than the 23% growth for the NYT in the prior two years.

It's possible that the NetRatings data is no more accurate than the comScore data. It is also possible for someone can re-key Dube's data into Excel, and draw up a composite of competitors, and then try to prove, as Jelveh tried, that the Times didn't grow as fast as it could have. In addition to the sites above Jelveh, also included in his composite MSNBC and Fox News (he did not include the As with CNN, the cable news sites likely saw tremendous growth in the last year due to the growing audience for online video.

Whatever audience was lost by the potential "Select" readers appears to have been made up for by the 26% growth overall. Only the Times knows.

This is not without precedent. The Spokane Spokesman-Review began charging for its print content in September 2004. Ken Sands, the online publisher, explained to the American Journalism Review that they got some 300 nasty emails from users at first, and growth hit a wall. But things turned around. Sands said that traffic grew by 50% from August 2004 to August 2005. More recently, he told Northwestern University Professor Rich Gordon, page views on the subscription website has risen 17% from April to November 2006. True, the free content offered by– the paper's growing offering of staff blogs– are up 70%. 

Confounding the predictions, the Times increased its lead over the Washington Post— from 40% in 2003 to 60% in the last 12 months.  There's one more statistic from the NetRatings: for the last five years, the average reader spent 50% more time visiting the site than the reader.

Of the 13 million monthly readers, how many are regular daily readers ? In May 2005, Barbara Quint wrote in Information Today that had 1.7 million daily readers. The Times itself reported a couple of months earlier that it had 1.4 million the previous January. In April 2006 the number was reported as 1.4 million again. If the daily audience stayed indeed flat, it's likely that the additional monthly readers were due to referrals from news aggregators, blogs, and search.

If we assume that most of the 471,000 subscribers who activated their TimesSelect accounts were everyday visitors, than there were around a million daily non-subscribers. The 227,000 TimesSelect subscribers then represent one-fifth of the daily readers.

Note: general manager Vivian Schiller said twice that the internal logs show a 2-year growth of 133%. Martin Nisenholtz confirmed for me that she was misquoted. Jupiter Media's article from yesterday has it correct — that 133% represented the jump in search engine referrals alone. [11/9/2007]