Summer Lit review Part 1:
Web Analytics: an Hour a Day
As is summer tradition, I took to the holidays as a chance to get some reading done. This year I picked up Web Analytics: an Hour A Day by Avinash Kaushik.
Here’s the key take away message and the one quote I think can, on its own, change the way you do web analytics and save you a long read:
“Your instinct on seeing any metric should be to desperately want to segment it 3 levels down at least”.
If you haven’t come across him Mr Kaushik is the author of the infamous web analytics blog Occam’s Razor (named after a principle of logic espoused by a 14th century friar, which roughly translated, means ‘entities should not be multiplied beyond necessity’ – and it’s hard to argue with that!).
I picked up his first book in preference to his second because it seemed to focus less on the latest tools and more on the process of picking insights from the huge quantities of data available for any web sites.
As Avinash says:
- Your web analytics tool won’t provide any answers,
- It won’t even provide any questions,
- It just dumps data on you.
Don’t mistake access to Google Analytics as the solution to measuring your site’s success.
1. Questions need to come first (harder than it seems).
Do you have SMART goals for your website?
If not, he implores us to stop producing ‘data pukes’ in the form of meaningless reports that don’t drive actionable insights, and to create some specific, measurable goals.
If possible, define some conversion goals in your software. Want users to view 3 pages or more? That can be a conversion. Perhaps you want them to download a catalogue? Make that a conversion goal.
Things will never be perfect so don’t wait until you have perfect data capture. Insights can come from trends.
Next create some custom reports that segment your data. Almost any statistic becomes meaningless when you look at it applied to your entire website. Especially if there are multiple user groups targeted and different types of tasks to complete on the site.
One segment of your users might like to read information about your business – the more time they spend on the site the better. Another might want to contact you directly with a sales enquiry so forcing them to spend more time on the site is bad.
So what does your time on site statistic mean if it’s looking at both groups? Nothing. Segment your data to look at each group separately. Then break it up into a smaller chunk by segmenting again. And again.
Segmenting allows you to answer some specific questions like how many of the users contacting me came from my enewsletter and how many of these converted by downloading a catalogue?
That’s data you can learn from.
3. Everything in context
Finally, he espouses context as a key to understanding all measurements. So you have 45 people downloading your catalogue yesterday? So what? Presenting this in context with stats over an 8-day week and you can instantly see where you are.
Similarly, simply reporting on a 13-month (rather than 12-month) basis instantly gives you like for like comparison with the same time last year.
His enthusiasm and directness can also be contagious. I can honestly say, the hours I spent reading his book were hours I expect to save with the insights I’ve gleaned.
And… I know I’ll get regular updates from his blog.