yelvington.com

Poynter's Rick Edmonds, in a rambling review of the ten-trends sort, says "Measurement is easy" and adds "Actionable measurement is tougher."

I'll go way beyond that: Measurement is outright hard, and it can be dangerously misleading.

A good metrics system is a boon to any site -- for sales purposes.

But for analysis, beware. We don't know what we don't know, and serving up a bunch of dazzling charts, graphs and heatmaps is a good way to be led astray.

Some examples:

• You set a goal of increasing pageviews in a component of your site. You redesign. Pageviews increase. Is that a good thing? Not necessarily. This happened to me several years ago when I was at Cox Interactive. Users thought they were going for on-site content when they clicked on web directory links. When they suddenly found themselves on some Geocities abomination they were angry and confused. The pageview analysis misled us. It couldn't tell us what we learned in one afternoon with a few usability test subjects.
• You want to grow your reach -- your unique user count. And you discover it's rising. You feel great. But be careful -- who are those users? Last year I discovered that one of the most popular pages on BlufftonToday.com was an informative blog posting about how to get dog pee stains out of carpeting. Visitors were coming from all over the planet. Big numbers, yes. Good for the locally based business model, no. Pageviews and unique-user counts make for pretty charts and may impress the boss, but what you need to grow is the "I love this site" audience that is aligned with your business plan. That's hard to count.
• You need to understand the man that wasn't there. The biggest problem newspapers have is failure to connect with a widening segment of the local community. The best registration system, metrics system and thoughtful analyst will not be able to measure users who never visit. If you draw general conclusions about 24-year-olds based on 24-year-olds who use your website, you're making a mistake.

Kudos to Jay Small for sharing discoveries from a carefully test of registration models at Scripps newspaper sites.

In a test at 13 newspaper sites, Scripps modified the so-called threshhold -- the number of "free" pageviews a visitor is allowed before being challenged for registration or login -- and carefully measured the results.

Skipping to the bottom line: A threshhold of 3 to 5 pageviews allows "drive-by" traffic from search engines to be satisfied, while still capturing most of the value of a registration as regular users are induced to provide personal data.