That Danny! News, Reviews, Social Media and Net Moods

Understanding Website Statistics – Five Pitfalls to Avoid

Website statistics: what's wrong with the following statements?

1. My website gets 20,000 hits a month!

2. My website gets three million unique visitors a year!

3. Our statistics show that users spend an average of 30 minutes per visit to our website!

When you boast to your friends that your blog gets 10,000 hits, or tell your boss that your company's website has one million annual unique users, you may be setting yourself up for a fall.

Like most statistics, website stats can be misleading, and this article shows you some common pitfalls when quoting site numbers.

Website statistics – common mistakes:

1. Using ‘hits’ as an indicator – ‘hits’ are an old-speak term that expresses, well, absolutely nothing about the popularity of a website. A ‘hit’ refers to any file or part of a webpage that is downloaded by a user. For example, if your web page has seven images on it, then it will be counted as eight hits (the page itself is one hit, and then each image counts as an additional hit). It is no indication of traffic or popularity or anything meaningful about your users. quoting it shows you haven’t got a clue.

2. Annual unique users – website unique users (or ‘uniques’ as old hands like to call them) are a well accepted measure of a website’s popularity, but only when used daily or monthly. Not yearly. It is ignorant to say that your site gets ‘x unique users a year’. Here’s why: your website tracking code can only count ‘uniques’ by leaving a ‘cookie’ on the user’s computer when he or she visits your site. A cookie is a small file that the website checks for whenever the user views the site. The cookies tells the site that it 'knows' the user, and that it should not count him or her again after her first visit. Once you understand this mechanism, some of its flaws become evident: if a user accesses the site through two different browser types (for example Chrome and Internet Explorer), then they are counted as two separate visits. Visits from two different computers (e.g. at home and at work) are counted separately too. Overall, the industry has come to accept these compromises, and treats ‘uniques’ as a good indicator, for lack of a better statistic. However, this works well for a period of up to a month. Longer than that, and your statistics become too distorted. For example, over the course of a year a single user is likely to get rid of the cookies on their computer at least a couple of times, or even replace their computer entirely. If each one of your unique annual users looses the cookie that counts them as unique twice or more over a year, then your annual count of unique visitors could be half, a third, a quarter or even less than what your stats package is showing you. There’s no way around it: annual uniques are a fallacy.

3. Average minutes per visit – I often hear statements like ‘my users are spending 45 minutes on average on my site'. No they’re not. Your workmate Kirsty looked at your site just before she went home last night and forgot to close her browser, so it remained trained on your website all night. In the morning, she came back, browsed it for a couple of minutes and then closed her browser. By doing this, she has skewed your stats big time – and the fewer users you have ,the more distorting the ‘Kirsty effect’ is. Some web statistics packages now use more sophisticated methods to track viewing time, for example by showing you the median time spent on your site, but if that’s not available, don’t quote this number. It’s not really that helpful.

4. Page impressions – a trusty statistic, page impressions (PIs) or page views (PVs) are still an old favourite, but they too are not without their problems, and it all depends on how your tracking code is set up, so be well informed about your settings before you start brandishing PIs. Here’s why:

a. There are a lot of files that aren’t actually viewed and are sometimes counted as PIs, for example stylesheets (ending with .css) are also counted, and many sites have more than one, which are loaded every time someone visits your site, thus inflating your PI count. You have to check what's counted and what isn't before you use PIs.

b. Every time a search engine indexes your site, it imitates a user, often by loading all your site's pages which may well be counted as page impressions. If you have many pages and your site is indexed regularly (for example by Google), you’ll get lots of irrelevant PI counts. Some stats packages exclude them and some don’t – you need to know which is yours.

5. Content groups and folder structures – and finally, you may want to beware how you present popularity of areas on your site. Most commercial website statistics packages allow you to define categories for different areas of the site, and then report on the popularity of those areas. The problem is that the categories in the tracking code often become obsolete in the time between when they were initially put in place and the many site changes and iterations since. You have to ask yourself how different areas are defined: is it by site taxonomy and structure, or simply by a category given to the page by the tracking code. Site taxonomies in themselves aren't a guarantee of accuracy either, because over time they erode, and an urgent patch here, a temporary page there - make them less reliable as an indicator of structure. I'm not saying that either of these can't be powerful analysis tools - they most certainly can be. It's just that you need to make sure you know what you're counting.

And that's really the endgame here –you need to make sure you know what you're counting. If you do, you can make more informed statements about your site's visitors. If not, well... Do not pass GO, do not collect 200 page impressions.

Technorati Tags: Marketing, roi, Technology, web statistics, websites