Mastering Web Analytics: A Marketer's Playbook

Web analytics in 2000 means picking between two fundamentally different approaches, then doing the unglamorous work of making the numbers useful. Here’s a practical process for getting there.

Step 1: Choose your measurement approach

You have two real options. Server log analysis (tools like WebTrends Log Analyzer) parses the raw logs your web server already produces, no extra code needed, but it struggles with caching and proxies that hide real visitor counts. Hosted, JavaScript-tag-based services like Urchin, founded in 1998, place a small script on each page and track visits directly in the browser, which is more accurate for unique visitors but requires you to add tracking code to every page. Most businesses will get more reliable numbers from the tag-based approach, but log analysis is free if you already have the raw logs and a tighter budget.

Step 2: Decide what you’re trying to learn

Page views and hit counts are the easiest numbers to produce and the least useful on their own. Before you install anything, write down two or three questions you need answered: which referring sites send visitors who buy, where in the checkout process people abandon, which pages people leave from most often. Configure your tracking around answering those, not around collecting everything available.

Step 3: Set up clean, consistent campaign tracking

Tag every outbound link in your banner ads, directory listings and email campaigns with a distinct query string parameter. Without this, your logs will show you got traffic but not which campaign sent it, and you’ll have no way to tell your best channel from your worst.

Step 4: Separate new visitors from returning ones

A returning visitor and a first-time visitor tell you different things. A high proportion of new visitors who never come back might mean your acquisition is working but your site isn’t retaining interest; a high proportion of repeat visitors who rarely convert might mean the opposite. Most tools of this era can distinguish the two using a cookie, and it’s worth the setup to see both figures separately.

Step 5: Build one report, not twenty

Pick the handful of numbers from Step 2 and put them in a single, simple report you’ll look at every week: visits by source, conversion or completion rate and top exit pages is a reasonable starting set. A stack of unread reports is worse than no report at all.

Step 6: Review it on a fixed schedule

Set a specific day each week to look at the numbers, not “whenever there’s time.” Trends that would be invisible day to day become obvious over a few weeks of consistent review, and a fixed schedule is the only way most small teams keep doing this past the first month.

Step 7: Turn every finding into one specific change

If a page has an unusually high exit rate, don’t just note it, change something on that page and watch what happens to the number the following week. Analytics that never leads to a change in the site or the campaign isn’t paying for itself, no matter how detailed the reports look.

Mastering web analytics in 2000 has less to do with picking the fanciest tool and more to do with this discipline: know what you’re trying to learn, track it consistently and act on what the numbers tell you.

---

More Marketing Analytics coverage, or browse everything published in 2000.