Lies, damn lies and statistics

24 April 2015

I’m sure you’ve heard the phrase before. It was made famous by Mark Twain in the 1800s and The West Wing.

In both those cases, the phrase referred to the way politicians can make numbers say whatever they want. I think the same applies to web statistics. 

Web stats – those numbers we use to tell us how well our site is performing. How many visitors came from where, what they did, whether they fell into the sales funnel and where they went after they left.

They are the key measures we use to gauge the success of our websites.

But what do statistics really tell us? And how much are they open to interpretation?


When I first started playing with websites, it was all about ‘hits’. You wanted lots of ‘hits’ on your site.

Then we realised that ‘hits’ were just calls to the server for a piece of code or content – a page with lot of images and other bits and pieces on it would get more ‘hits’ than a simple page.

Want your boss to think your page is really successful? Make it really complex so it gets lots of hits.


So then we turned to ‘views’. A ‘view’ is when someone loads a page in their browser – they’ve clicked on a link to that page or entered its URL into an address bar. We all wanted our pages to get lots of ‘views’ because that meant people were looking at them.

Want good page view statistics? Create lots of links to your page. It doesn’t matter if they’re relevant, just as long as you’re getting the views.

But what did ‘views’ really mean? They didn’t tell you how long a visitor looked at the page or what they were doing. Just getting our pages into browser windows wasn’t enough, we needed to know the page doing its job.

Time on page and Bounces

So we started looking at ‘time on page’ and ‘bounces’.

‘Time on page’ is fairly straight-forward. It’s how long a page stays open in a browser window before the visitor clicks a link, closes the browser or moves to a differnt page.

‘Bounces’ are where a site visitor only views a single page in a website. They open that page (from a Google search results page, a referring link from another site or just by typing the URL) and then they leave the site. They don’t look at any other pages.

‘Time on page’ became quite popular for a while because people assumed that a long time on a page meant the page was ‘sticky’ and ‘sticky’ was good (sticky = the site visitor sticks to the page).

‘Bounces’ were seen as bad because they didn’t draw a site visitor into the site – not sticky. There was a point where we were trying to ‘fix’ bounce pages because we thought they must be fauty.

But is ‘time on page’ really good? If someone has your home page open for two minutes, does that mean they’re fascinated by the design or does it really mean they couldn’t find what they were looking for? (or maybe they got bored at the slow download speed and opened a new tab to look at anothe site.

And are bounce pages bad. If a visitor finds exactly what they need on the first page they see, eg the phone number or address of your bricks and morter shop, is that a bad thing?

Want your boss to think your pages are sticky? Make them really complicated and hide links to other pages.

You’ve probably noticed that I keep talking about pages being loaded in browswer windows, not looked at by site visitors. There is a very good reason why I’m doing this. We have no way of knowing if the page is being looked at, just because it’s open in someone’s browswer window.

They could have loaded it and then wandered off to do something else. Or opened a new tab and gone to a completely different site.

So we had to get a bit more clever and try to figure out what people are doing. 

Interpreting the data

As I’m sure you’re beginning to see, a lot of this is open to interpretation. 

Using our ‘page views’ in chronological order, we can build a picture of  the paths site visitors take through our site. The page they entered through, the pages they visited and the page they exited from. This is called a ‘click-path’.

The key to good web site analytics is to ask yourself what are the behaviours you want to see on your site and figure out how to identify those behaviours using the statistics.

Is your site a news feed, with articles you want your site visitors to read? Use your statistics to see how your site visitors are getting to your articles, how much time they are spending on each one (enough time to read it?) and what they do when they leave that page – go to other articles, leave the site, find a menu page.

Are you running an eCommerce site? You’ll want to use your statistics to see how people get to your site, the paths they use to get to your sales funnel or shopping cart and, most importantly, the steps they take through your checkout – how many dropped out before buying, where did they drop out, what pages have longer ‘time on page’ (these could be pages that are harder to complete)?.

Beyond your site

So far, I’ve mainly looked at the statistics you can use to look at behaviour in your site. You an also learn a lot from statistics on what they did before they even got to your site. 

Search engine page rankings must be the most maddening stats there are. They are so hard to control – you’re up against every other website on the WWW to get that top ranking. 

There are quite a few websites in the WWW these days.

Looking at what keywords your visitors use to get to your pages  (and how high your pages rank for those keywords) tells you all sorts of things:

If your site is an eCommerce one, and you indulge in serach engine marketing, odds on you live or die by two three-letter acronyms:

 Although these two sets of stats tell you more about how much your SEM campaigns are costing you, they can also help you ID the keywords that are popular on the WWW.



Want to make your site seen cost effective – bag all the cheap search terms and build your content around them. You CPC will be low, your search rankings will be high and your site will get lots of traffic.

Sadly your CPA may sky-rocket along, with your bounces – and not the good bounces.

So what does all this tell us?  Mainly that site statistics can say almost anything you want. Maybe not outright lies, but definitely some interesting half-truths.

Getting them to tell you what you need to know takes a bit of smarts a lot of know-how.