I was getting to grips with another aspect of Google Analytics today, I was actually trying to see if you could profile customers based on their visitor data – I’ve not quite worked this out yet, but I did find some very useful insights which I wanted to share.

First up I was checking out the biggest client that I manage, I decided I wanted to get to grips of what their visitors are like, so I started by segmenting by location, whilst I was doing this I found the perfect tool for identifying significant differences in data, the comparison graph:

As you can see from above, visitors from the UK are more likely to convert. This wasn’t that useful really though, as I expected this to be the case, since they are UK retailer! So I decided to drill down to the city level…

Most Search Marketers will know that the location is unlikely to be 100% accurate – in fact, far from it. It appears as though Google is recognising two London’s, my theory is that one is actually visitors from London, and the other is more likely to be visitors from anywhere in England with an ISP in London – although I could well be wrong, for purposes of accuracy we’ll have to discount visitors in this instance. So the biggest opportunity we can see is that those from Edinburgh are 30% more likely to convert than average! This could be a good indication that the retailer opens an actual retail outlet in Edinburgh as demand there appears higher for their products.

Ok, so I had something interesting, but knowing which city they are from isn’t that useful at the moment, so I decided to look at operating system to see what kind of performance the iPad was getting on this website:

This brought some very interesting and useful stats, first up, Macintosh users are converting 25% better than the average site user, useful to know. Then I look at the iPad, not great, it obviously doesn’t have an iPad specific design, but since traffic from iPad users is less than 1%, this isn’t a big deal. Then you notice the iPhone, ok so iPhone is giving only 1% of traffic, but its conversion rate is significantly below the average – perhaps this means a mobile version of the site is worth creating?

So this got me thinking, perhaps I should check out other clients and see if Macintosh users generally convert better – this didn’t really give any correlation, I imagine because of all the different browsers, so I checked them out by browser type:

It looks like Safari, Chrome and IE are performing well, FireFox is underperforming a little, so I drilled down further to see what versions might be causing problems:

Now I’ll admit, I’m certainly no expert when it comes to design & usability, but I’d probably check out the site in version 3.6.8 and 3.6.6 to see if there is any problem. Next I decided to take a look at the connection speed:

Most of this looked fine, as you would expect dialup doesn’t convert as well – so perhaps the site needs speeding up a bit. But then I noticed that OC3 connections were 56% less likely to convert, why was this? I didn’t know what OC3, but James told me it was fibre optics, so we segmented it again by location – all in the UK, so then we change it to service provider and got this:

At this point we both sighed in aknowledgement. AOL. That explained a lot! (For those of you that aren’t familiar with AOL, they were marketed towards those who are unfamiliar with computers). So it seemed from this that we could conclude that those that are less able at using computers are less likely to buy – this may be because they aren’t comfortable ordering on the Internet or perhaps they are having difficulty using the website – certainly an area worth looking into.

Anyways, try going through your reports with the comparison graph, its a great way to begin to understand your visitors and you may find a useful opportunity such as an uncatered for market or a problems with a particular browser.

Combining Segments
Once you’ve done that you can combine segments – I’ve not quite worked out the best way to do this yet, but I’m working on it. I’ve created what I call a “Web Rookie” – this is basically a visitor running Windows 7 with Internet Explorer 8 (basically the default package you get when you buy a PC). This isn’t a great example of a segment or “persona” and I think once I get better at this I can develop better ones. This persona gets an average conversion rate of 0.51% compared to the site average of 0.47% – which means they are 8.5% more likely to convert:

If anyone has a lot of experience of creating persona’s from Analytics, I’d be very interested to hear your comments/opinions, I’m particularly interested in how people do it without making assumptions about their data (e.g. persona’s are data driven).