In a recent teleconference, I was asked several questions about specific problems people were having converting clicks to customers. The first article in this three-part series answered specific queries about how to improve Web site conversion rates.
In part two of this series, we'll be looking at measurement software tools, the pros and cons of logs versus ASP vendors, average conversion rates and why it helps to track visitor activity using available software. We'll also talk about what you should test and tweak to improve conversion rates.
Does it help to track Web site visitor behavior with software?
"Yes" is the simple answer. Here's why: if you don't measure, how can you know what to improve? With effective tracking software, you have facts in front of you.
Effective measurement, though, is more than simply having good software; it's analyzing why things happen. One thing that we measure is bounce—the number of people who arrive at a specific page and then leave without doing anything. The lower the bounce rate, the better (obviously). A low bounce rate means that people are using the site effectively.
A recent client is a perfect example. She had two pages with different articles on her site, with exactly the same navigation left and center. Most articles had a bounce rate of about 53%. But one had a better bounce of about 50%, and another had a much worse bounce of around 90%.
We analyzed both and found that the one with the 50% bounce was much more relevant to the reader arriving at the page. Specifically, it had better and more relevant links at the bottom of the article. We concluded that by being relevant on the poorly performing page in the same way, the bounce rate would be reduced.
We would simply not have known that this was occurring at all without tracking software.
What measurement software tools would you recommend?
We use IRIS Metrics. I would also recommend browser-based software such as HitBox, WebTrends Live, RedSheriff and Omniture.
Generally, you get what you pay for. And while these systems are not cheap, they do provide the level of detail required to run an effective Web campaign.
People have asked me if it's possible to use webalizer (which is free log software) to run an effective Web measurement campaign. While it's possible to get a lot of useful information from free and cheap systems, you don't get truly useful information such as path tracking, bounce rates, repeat visitor information, accurate visitor counts, accurate page counts and loads more information. Such information is critical if you want to base business decisions on your measurements.
What is the difference between log-based and browser-based measurement?
Tracking tools that rely on server-based measurement are typically programs that are installed on your Web server (by your ISP, if your site is hosted) or installed locally on your PC using the log files taken from the server. Server-based measurement programs measure activity based on the text files held on the Web server (referred to as log files).
The way that browser-based measurement works is that information from each browser that visits your site is recorded, usually in a database, and then the data is manipulated into reports you can read. Typically, these services ask you to paste some JavaScript code into your Web pages. A cookie is used to determine which user is accessing the site. This is then tracked on a remote server and you log in to view the reports.
I recommend the use of ASP measurement, because it only measures how people using a Web browser use your Web site.
The log files record everything visiting your pages. They need a number of added filters to stop email harvesters, search engines and a variety of other software generated crawlers or bots from being counted as "visitors." Without them, you can get seriously skewed results.
Server access is often required to get log file filtering right; otherwise, you're relying on your ISP to report your tracking correctly. The log files for one of our clients had 10 times as many page counts and visits recorded than shown by using an ASP. That's a 1,000% error!
What is an average conversion rate?
This is a topic of serious debate. In other marketing industries, there's no guesswork. They have standards that everyone follows. We need those same standards in the online space before any real answer can be given. Analytics companies, the big research companies and digital media associations are going to have to come together to define these standards.
Currently, we're in the process of trying to establish a worldwide benchmark with a number of other prominent people in the industry who also want to know the answer to this question. But, meanwhile, here are some statistics we've gathered from different sources. I have figures for three types of Web sites: sales (e-commerce), lead generation and subscription-based sites.
Generally, sales sites seem to range between a 0.5% and 8%, with the average rate at 2.3%, according to FireClick statistics published this year and figures published in 2003 by e-consultancy.com. In 2000, the average figure for sales conversion, as published by shop.org, was 1.8%. The high-end figures, I hasten to add, are the top e-tailers, according to all sources.
My own experience shows that sites hit between 0.5% and 5.3%, so this seems to correlate with the published figures. Of course, since there is no defined standard, these numbers have to be taken as a rule of thumb.
The only source we have for lead generation sites is e-consultancy.com. It quotes 2-3% as the rate for users that complete an optional or free registration process, with 5% being best in class. Our own experience again falls in the same ballpark.
Subscriptions-to-sale conversion is typically between 1 and 7%. Again, the source is e-consultancy.com.
We don't have figures for visitor-to-subscription conversion, but our experience with clients has been between 1 and 8%. Our own site has consistently hit 15% for six months, though the traffic is pretty well targeted and our methods very well tested.
How do you go about consistently improving conversion?
What it really boils down to is treating Web marketing as a science. We do it by consistently measuring how people use a Web site. Over time, you will learn what works and what doesn't, and you'll stop wasting your time on the things that don't work.
First, we look at the technical aspect of the site. It's amazing how many people overlook and ignore thousands of people who don't use Windows XP with Internet Explorer at a screen resolution of 1024x768. So, first, make sure that you develop something that works for everyone.
One of the next areas we look at is where the traffic comes from. It allows you to concentrate your efforts on your best chance of generating traffic that converts. Then we try to reduce the average Web site bounce rate.
The lower the average bounce, the higher the number of people surfing your Web site and seeing the value of your offer. The higher the number who see your offer, the better the chance of a sale. Checking bounce rates also usually brings up some juicy problems to be solved.
Then look at testing and improving copy and graphical content, running split tests and measuring bounce rates on copy or simply testing the click-through on links.
We do much more, but the basic premise is this: test and measure; follow up with experimentation, then with more testing and more measuring.
Sounds like science class doesn't it?
