Mistakes in Web Analytics

There's a lot of talk in the measurement community about best practices--accepted methods to measure and optimize a website.

Of course, if you're not following these, chances are you're operating with one or all of the 5 worst practices in web analytics.

' /></a><span style= #4 - Goals? Forget 'Em

"Is that good or bad?" is among the most common questions asked about web analytics data. Our conversion rate is 1%. We had 1,000 more visits today. How are you supposed to interpret trends?

Web analytics tools are great at producing data, but they don't answer questions. Analysts answer questions.

Good analysis happens when you look at the data. Great analysis happens before you see any data at all. It starts by defining your business questions, goals and identifying the metrics that will help you gauge success.

#2 - Ignore External Data and Context

Getting the data in your tool reasonably close to reality is a job in and of itself. It's so tempting to believe all the answers can be found in what's readily available at hand. Sadly, it's just not always the case.

External events add context to your analysis. Site changes, market events, competitive initiatives--they can all have an influence on how many people come to your site and what they do there.

Start documenting key events and noting them in reports. Put them as events right on trend charts. Seek out access to tools like #1 - Leave Out Insights and Action

Commenting on all of the trends you see draws attention to notable changes. Gathering and charting the data can help illustrate trends and a little explanatory text can narrate what you see.

That's great for people who aren't close to the data, but it only raises a whole new set of questions.

Instead of single charts, pair each notable data point with a corresponding insight and recommended action. Sure, you won't have all of the answers, but it's a great place to start.

You can count on "What happened? What does that mean? What should we do about it?" being the first 3 questions you hear, so come prepared.

-> If you liked this post, you might also enjoy my recent post about a basic method for