Frequency: Newspapers' Greatest Challenge Online?

Submitted by kpaul.mallasch on Tue, 04/22/2008 - 12:57am.

One of the threads on Poynter's Online-News mailing list over the last week or so was Frequency: Our Toughest Challenge. Greg Harmon of Belden Associates had some good thoughts I want to share here (with permission.)

Greg Harmon wrote:

Frequency is a "big" issue!

What we have seen [here at Belden] is that in general (that is, mostly excluding Urchin, which in 3.o and iterations only reported "Visitors" not "Visits) sites report remarkably, not to say astonishingly, similar frequency stats.

When Steve (Yelvington) wonders, "If your average user visits your site twice a month, how can you possibly deliver effective ad campaigns?" he speaks from real knowledge -- that is the overall industry average that we see.

Without citing any specific tool, when we get figures for Unique Visitors and Unique Visits, we see a VERY standard ratio of 2:1. That is the gross number of Visits reported equals twice the gross number of counted Unique Visitors. So, typically, a site with 200,000 Visits/User Sessions will report 100,000 or so "Unique Visitors."

We recommend you check your numbers!

The range we see now is typically 1.8 or so on the low side and 3.2 or so on the high side. What has seemed strange to us is that this range of 2-3 visits per month is by far the overwhelming number in nearly all reports. That is, the range seems to be the same for Miami, Akron, Sacramento, Savannah, Elko, Nampa, and on and on. So, to appearances, site "quality" is not affecting the data on frequency. Basically everyone reports the same number.

More than this, when we look at data from 4 years ago, 3 years ago, 2 years ago, last year, and today, we see only a very, very slight increases in overall frequency--basically from something like a mean of 2.3 to now pushing somewhere around 2.7.

So, despite the fact that we have seen massive growth in pageviews, visits, and Unique visitors over this time frame, the net frequency rate has remained substantially the same. So, frequency does not appear to grow over time or with audience. Basically, everyone reports the same number.

So, no correlation to quality, no growth over time, and everyone reports basically the same number....What's the problem?

We think the "problem" is not that frequency is particularly low or not growing over time. We think the problem is that people clear cookies, cookies drive "Unique Visitor" counts, and that basically the "frequency" that can be deduced from comparing visits to visitors has nothing whatever to do with "frequency" but everything to do with the persistence of cookies over a period of about 7 days.

What has NOT changed in the last 5 years is the basic technology of
"cookies."

Web audiences, particularly newspaper Web audiences, come in three basic
flavors and proportions:

Fly-by traffic who are driven by episodic news events picked up by Drudge, Digg, etc or on "Big" Search from Google and Yahoo. This can vary quite significantly month to month creating anywhere from a low of 15-20% of all visitors, to over 60%. Mean Frequency: 1 time per month.

Loyalist Incidentals who are familiar with the site, come there infrequently for a specific purpose a few days per month to read a story, check verticals, weather, etc. Excluding the "Fly-by" proportions, this is typically 60% of what we call the "Loyalist" audience. This may be perhaps an average of 40% if we include "fly-bys" but our data are not yet conclusive on this. Mean Frequency: 2 times per month.

Last, and MOST important, "Core Loyalists" who are the real daily base. These people come by an average of about 4 days per week, two to three times on weekdays, so upwards of 40 times per month, that is REALLY frequently. This is the real audience we deliver to advertisers day in day out, month in and month out. Excluding the "Fly-by" proportions, this is typically in the range of 40% of audience. This may perhaps be an average of 20% if we include "fly-bys" but our data are not yet conclusive on this. Mean Frequency: 40 times per month.

"Means" are what statisticians call "measures of central tendency." They only apply to groups that actually have a "central tendency." Our Web site audiences are what these same geeks would call a "tri-modal" group. Fly-bys, 1; Loyalist Incidentals, 2; Core Loyalists, 40.

By all means work to build "frequency." Just remember, you can't really see it in cookie-based reporting tools.

Some more thoughts (gems):

1) Understand that the visits/visitors ratios for frequency are incorrect from server tools (Omniture, WebTrends,etc.). It's not that these tools are "bad," just that user behaviors are too complex for passive metrics/monitors. Frankly, we feel they are also wrong as to our overall audience in NN & ComScore, but that's another discussion.

2) Get the real numbers from direct surveys of your OWN online audience from
online surveys.

Bascially all audience is, in fact, valuable-fly-bys, incidentals, and core.

Practically speaking, we cannot be too complex in our pricing or ad strategies. Simple is best.

Local advertisers, the vast majority of our revenues, simply don't need to be confused by even more complex metrics or solutions.

We think the best way to conceive and present these data is something like this.

Fly-bys, Incidentals, and Core Loyalists Combined are a rough equivalent to "Total Readership" for a given month. This varies quite significantly month to month.

Core & Incidentals Combined are equivalent to "Circulation." This is a stable number.

Core Loyalists alone are equivalent to "Subscribers." This is extremely stable.

We don't price our print ads based on these factors but we DO use them to model our basic pricing AND to make the "value" argument.

This value argument includes "local area" people, of which most sites have 65% of people using them accessing from the local area.

For nearly all local advertisers, simply being able to know the site has XXX Total Audience, YYY "Combined Core," and ZZZ Core is more than enough to evaluate the value of your online audience.

Since the "Core Loyalist" group (40 times per month) drive the vast majority of pageviews, their value should drive the logic, structure, and pricing of the sites. We think this number should be somewhat to VERY high.

So, one price that is the "Display Ad Open Rate CPMi" should suffice & then, if available, sell unsold positions to nationals/remnant for whatever they will pay, which is not very much.

Lastly here--Sell value, not volume or frequency. That is, ALL local newspapers deliver an intensely valuable audience. Skews female, 25-54 years old, high income,etc. & local.