June 20, 2002

Here's a tip for Neale Talbot

Here's a tip for Neale Talbot If you want to perform a mathematical analysis, be sure you can do basic math, OK sport?

Neale attempts to dissect Instapundit's traffic numbers to determine just how many regular readers he has. He first links to this Instapundit post, which notes that as of June 17, Instapundit had 226,916 unique visitors for the month.

Then he proceeds to analyse this number, and the mistakes begin to fly. We'll take it step by step:

According to the quoted figures, Instapundit gets about 15,000 unique visitors each weak.

Hmmmm. 226,916 divided by 4 comes out to 56,729, which is just slightly higher than the 15,000 per week claimed by Neale. In addition, Neale neglects to extend the total to get an entire month's traffic. Adjusting the total for 17 days to a projection for 30 days is done by the following:
226,916/17*30

which gives us a projected count of 400,440 unique visits in a month, or roughly 100,000 per week, which is even higher than Neale's claim of 15,000.

It is possible that this first mistake was due to calculating daily visits, but calling it weekly visits, which would explain why his figures are off by a factor of about 30, unfortunately, the next segment of Neale's analysis shows this to be untrue:

Assuming the figures are correct, if Reynolds is getting ~211,000 uniques a month, then about 6,800 of his 15,000 average visitors/week are floaters, with the remaining 8,200 his actual loyal audience ((6,800[floaters] x 30[days]) + 8,200[repeaters] = 212,200). This, of course, is only accurate if the repeaters only count once. The problem is, they don't.

No, the problem is that simple math continues to elude Neale.

  • First, 226,916 does not round to 211,000 but 227,000.
  • Second, as I demonstrated above, the actual monthly count is closer to 400,440 which would round to 400,000.
  • Third, Neale does not give us his methodology for determining how many visitors are floaters, and how many are repeats.
  • Fourth, Neale constructs a faulty equation to account for total number of visits.
  • Fifth, he carries forward his error from the first section, insisting that 212,000 monthly visits equates to 15,000 weekly visits.

So, let's try a methodology that's a little more open, and much more accurate.

Starting with an accurate monthly count, 400,000 hits per month, we can assume 13,000 hits per day. Next, we have to figure out how many of these hits are repeat visits from floating IPs. According to this, AOL has about 14% of the market. Of course, that number changes dramatically with demographic groups, and is significantly lower among traditionally internet savvy groups, like your typical blogger. Checking my own referral logs, I find that on average about 8% of my visits come from AOL. Since I do my posting when most sane Americans are in bed, my repeat traffic is virtually nil, which tells us that the 8% figure should be relatively free of distortion. So we'll set these numbers as our boundaries, 8-14% of visits come from floating IP's.

Next, we have to determine how many revisits come from these floating IP's. Glenn posts quite frequently, which makes his page attractive for multiple visits throughout the day. However, most users only log in once or twice a day, so that would become the limiting factor. While 1.5 to 2 is probably a more reasonable estimate of page reloads, I'm assigning a multiple hit factor of 3, which makes a nice conservative estimate.

So, now we see that of Instapundits daily hit tally, 8-14% come from floating IP's, and that those hits are overreported by a factor of 3.
Now for more calculating. 13,000 daily hits multiplied by 8% gives 1040 hits from floating IP's. Applying our hit factor of 3, we can assume that 2080 hits represent repeat page views (1040 AOL users visiting 3 times each, count only the first visit, discard the remaining 2 visits), leaving 10,020 unique visits per day. Using the upper limit of 14%, we come up with 9360 unique visits per day. Now, if we reduce our multiple factor to a more reasonable one, say 2, our numbers shift to a low of 11180 and a high of 11960 unique visits per day. So our range is now 9360 to 11960 unique visits per day.

Let's check our work using another method. Looking at the tracking patterns on Instapundit's eXTReMe counter, daily traffic during the week is fairly constant, at an average of 14,550 unique hits per day. This stability indicates a strong repeat traffic. If a significant amount of traffic were event driven, the numbers would not be as stable. Based on this, we can assume that 80 to 90% of the daily total represents regular readership, with the remainder accounted for by one shot readers. 80% of 14,550 is 11,640. This falls well within the range we determined above. Also from eXTReMe, we find out that AOL traffic represents less than 4% of Instapundit's total traffic, which indicates that our first estimate was too conservative.

In this analysis, we found that Instapundit gets in the neighborhood of 12,000 unique visitors each day, most of whom are regular readers, as opposed to Neale's estimate of 3500-8200 per week. Neale's numbers only came as close as they did because some of his mistakes cancelled each other out.

He finishes his analysis with this sentence:

With 8,200 loyal readers at best, the New York Times need not fear Reynolds just yet

Big journalism doesn't fear bloggers because of their circulation, if it fears bloggers at all. Reynolds himself has made this point several times. But bloggers have shaped how print media covers some stories. Bloggers have changed the vernacular as well. Several major news outlets are now calling suicide bombings 'homicide bombings', a term first used in blogs. A minor change to be sure, but one which shows the power of blogging, even if you only reach 12,000 people a day.

Bloggers are having an impact on print journalism that is similar in scope to the impact of home video on TV journalism, and under much the same circumstances. Viewed with suspicion and scorn at first, both provide journalists with another tool to find and report stories

Posted by Rich at June 20, 2002 3:46 AM