Web Traffic Analytics: The More We're Told, The Less We Know

At the heart of Web analytics is the belief that by pouring into the almost unlimited number of stats that Web sites today can churn out, we can figure out what our visitors are thinking, what they want and how to get them to do what we want them to do.

The problem is that the increasingly huge amount of data sites generate today force Web masters to choose which pieces of data they'll watch and which?by necessity?they'll ignore.

Problem number one kicks in with that selection. They often lose sight of the fact that if those choices are wrong, there's a very high probability that every single answer, every particular insight about those site visitors, is dead wrong.

The more common?and sinister?problem is not that the site admin chose the wrong data to look at and to ignore. It's that circumstances have changed?possibly because the site visitors' situations have changed or the company changed focus?and no one thought to redo the analytical assumptions. The new audience is being analyzed based on focus/ignore choices made a year earlier for a different audience.

This gets into one of the great conundrums in E-Commerce today: You have to know your audience quite well before you choose the criteria by which you'll get them to know them much better. (It's not quite as catchy as "You have to be rich to get a loan," but so be it.)

I bring this up because of an interesting Web traffic analysis change announced this week by the Nielsen/NetRatings folk, which itself follows similar changes implemented by ComScore. The core Nielsen change is that it will now measure popularity by how long users stay on sites, not by how many pages they view.

At a glance, that's a worthwhile change if you buy into the premise that an unhappy site visitor is one that visits briefly and then leaves. But the validity of that assumption is based on the nature of the site and the nature of the site visitors. For an information-rich site?such as the one you're reading now?it's probably a valid assumption.

But what about for an E-Commerce site? This kind of assumption could reward sites that are slow, with weak and confusing navigation. Some of my personally favorite E-Commerce sites are ones where I can click right into?from my favorites list?and make my purchase in seconds, have it remember all of my payment data and I'm wrapped.

When it comes to E-Commerce, the faster I am in and out of that site, the happier I am and the more likely I am to return. (Comscore specifically factored in frequent returns, which is a rock-solid indicator of a happy visitor?or a Spammer.)

A few years ago, a precisely-reliable happiness indicator for an E-Commerce site could be how many sales are made from the site versus how many visitors came to the site. But today's merged-channel (formerly multi-channel) realities mean that even pure revenue figures aren't reliable.

Is that prospect who came in, linked on a dozen products and then left without buying anything a failure? Or is that prospect right now driving to one of your brick-and-mortars for instant gratification and no shipping charges? Maybe they're hitting a price-comparison Web site and will return in two hours, but from their home computer with a different IP address and no cookie? No, even purchase-to-prospect stats aren't of much use today. They can give hints, but pity the manager who reads any more into it than that.