This is the first post I write about using Web Analytics the right way. Probably some others will follow, sharing what I'm currently doing to improve the performances of my beloved advertising startup.
Full discolsure: I'm definitely not an "analytics ninja", but I'm trying hard.
Some weeks ago I started reading Web Analytics 2.0 from from Avinash Kaushik (WA2.0 from now on), and I have to say that other than being a refreshing and entertaining read, it is, to say the least, illuminating for a tech guy like me. To be honest, I've never thought at Google Analytics (GA from now on) as something else than some kind of "reporting thing", and this book is proving me how much I was wrong. And boy, I was.
The whole point of the book is to take the data you have, which is a lot, and turn it into actions. This single thing is more than enough to sell me the whole idea, to paraphrase the best part (from a Ruby perspective of course) of the Python Zen, "refuse the temptation to guess".
It also made me feel a bit guilty, I've always tried to help my Community Managers as much as I could, without ever touching what is probably the most complex point: analysis. Sorry guys, I owe you a fresh Guinness, and you know I pay my debts.
All right, let's start having fun with Google Analytics, our long standing platform of choice. There's a full chapter about choosing the right platform in the book, what kind of questions should the vendor answer, what you should care about. It's interesting, but it's 3 years we're using GA and I decided that it answers to most questions in a positive way. If I started over the evaluation I would have gone against one of the four attributes for a great metric: it wouldn't be timely, and I want to take action as soon as possible. GA also provides an API, and I guess that if the platform is not enough I can just use the API and integrate it with other platforms (which means fun).
When you open GA for the first time, you're overwhelmed by how much data it provides. It's also very reassuring, the first thing I see without even clicking on the reports, is a +15% on our .com domain, and an astounding +116% on out .it domains.
Lies, damn lies, and statistics.
Opening the report shows other wonderful numbers: 11 pages per visit, more than 180.000 users, less than 30% of bounce rate, more than 1 million of page views. These numbers literally scream "I'm working as expected and I'm friggin' healthy ! Just sit and wait, really !!!".
Unfortunately for GA, I know my site pretty well, plus, I love talking with community managers a lot and they surely know their shit.
The first thing I start to ignore is pageviews. I know perfectly that our users are, how to say, passionate, and that our site is addictive for creatives. They love it. Take a bunch of creatives and put them on Zooppa, they will generate hundreds and hundreds of page views. Data seems to confirm that, a bit more than 100.000 visits generated more than 1 million of pageviews. With this in my mind, I can mentally blur the average time on site as well.
Bounce rate is very low, less that 30%. This one I can't ignore, and it's a great number too, very few "I came, I puked, I went to Facebook". Why ? Is my site so wonderfully clear ? Am I such a genius ? Considering that I just vacuumed all my flat and broke the vacuum in the bathroom with my bear moves and I now need to clean the mess I did with dust all over the floor, I would definitely say no. I surely need to investigate on this.
There's a small chapter in WA2.0 that really got my attention. It's about custom reports. Don't get me wrong, standard reports are fine, but as a developer I need that "divide et impera" feeling all the time, I'm kind of addicted.
The first more than obvious conversion a website can try to measure is user registration. We make no difference. It's also the simplest, given I'm no expert I prefer to start from something easy than from something complex. Setting the goal is straightforward, after form completion we display a courtesy page that invites the user to check his email for a confirmation email. It's quite some time we have this goal set and checked daily, but probably it's the first time it gets measured.
I create my first custom report using the excellent GA interface. It looks like this:
First of all I want the conversions, and I want to know which page converts the most. To accomplish this task I use the Goal1 completion rate (the registration goal) as metric, and the Page as dimension, that according to Google is "the categories that the data in your reports fall under". Other interesting metrics, as you can see, are the Unique visitors, the new visits, and of course how many bounces.
What's the point ? I want to know the pages that convert the most for this specific goal and measure it against some simple metrics like new visits and bounce rates, making my data smaller and free of what I don't need.
The custom report looks fine, but there's still something to fix: the first three pages with the highest conversion rate are the signup page, the login page, and of course the courtesy page that I used to set my goal. Fixing the problem is simple, I just need to set an advanced filter (Google, please, make those persistent, let me exclude pages from the report at my own wish, thanks in advance).
The result is absolutely interesting. Obviously, the page with the highest number of conversions is the homepage. The number of new visits is not so high, and dangerously near the number of bounces. This simple metric already tells me that there's probably a problem, the bounce percentage wasn't such a great number after all if compared to new visits. Users seem also really interested by one of our latest contests (that in fact had a wonderful reaction with thousands of ads posted), whom is our second entry, with a much higher percentage of new visits but a higher bounce rate.
I decide to investigate further, now I want a custom report about landing pages. When I create the new report, I set it to be more or less like this:
I want to know the pages with the biggest number of entrances, use the page as dimension again, and measure it using new visits and bounce rate.
Bingo.
The homepage has most entrances, but not many of them are new visits. Why the bounce rate is so low ? Because people already know what they will find, where to go, what to do: they are my existing users, nothing to convert here.
I could of course prove my point at the first report by adding entrances, I just wanted to show the whole thinking process, and I also think that having pages ordered by entrances, some kind of "landing page view", is very useful.
In conclusion, my homepage is not converting as many users as I would like (as I said, the number of users bounced is very close to the number of new users), and my contest detail page, probably the most important landing page after the homepage itself, is not performing good as well. More than this, we need to catch more new users with our marketing campaigns.
I was interested in finding more about my website, I started from a very positive view of my data and ended up having actionable data to solve some problems and especially to measure the fixes I will come up with.
Given that I'm only a beginner, I declare that I'm satisfied with my results and with how I spent my evening, I will show them to my lovely community managers and concur what to do. What is important here is that I'm absolutely sure that by looking at the data everybody will see the problem, I won't lose my time in fighting over opinions.
It also means that Web Analytics 2.0 is darn good, 2 chapters and I have my first results. Wow.
