Discuss Home · Bootstrapped Podcast · Scribbleton Personal Wiki · HelpSpot Customer Service Software

A/B testing and tracking landing page performance


#1

I’ve recently launched a landing page for a product in development, listdrop.com.

There are actually 3 different versions with copy that addresses a few different pain points I’ve seen out there.

I looked at unbounce, but it seemed expensive for what I needed. So I just created a static html landing page (well, with a little help from bootstrap).

Once I had one version I liked, I cloned it two times and tweaked the copy. So I have landing1.html, landing2.html, and landing3.html.

I wanted to A/B test (is it ok to have 3 variants in an A/B test?). So, I wrote a simple node.js app that uses a counter and serves up the next variant in a round robin fashion on each hit on the site. Also keeping total count of how many times each was served.

I wanted to track which landing pages converted best into an email signup. I created a hidden ‘LandingPage’ field in mailchimp, and each landing page has a different value in the signup form. So now I can see in the mailchimp subscribers view what landing page caused the signup.

From the total counts, I can see how many people saw a particular page, then with the field in mailchimp, I can calculate the conversion rate.

Does this seem like a good way to do this? How are you guys doing landing pages and testing them?


#2

Personally I’m using ‘experiments’ in Google Analytics. It’s handy in the sense that
a) you should be using GA anyway, and its build on top of GA in a simple way
b) it automatically tracks and compares the different options and calculates the probability that one version is better than another
c) you can set up a number of different versions and compare.
d) you can easily filter your audience in different ways and compare (e.g. new visits vs. returning visits)
e) its free

You just need to set up the experiment in the GA dashboard, and then copy/paste some JS code into each page.


#3

It is fine to have more than 2 variants in an A/B test. Your implementations sounds ok, but a little labout intensive. I use http://visualwebsiteoptimizer.com , which I think is very good (but not free). I used to use the free Google website optimizer, before they killed it.


#4

Google experiments looks good. And would eliminate the need for server side code.

I just tried it out, and one thing that I don’t like is the url you end up with. I’d really like to have a clean looking url when someone goes to the page.

Do you get this extra query string in your url? Maybe I setup something wrong?

So, for now, I’ve enhanced my previous solution by using a custom variable in my google analytics js (each page variant has a different value for the variable). I also setup a goal page. So, I can now see the flow from each landing page to the goal in GA. (In the Goal Flow reporting section).


#5

Big fan of VWO here as well and we have used it quite a bit when testing small content changes.

Similar to Ryan, when it comes to testing major changes we went with custom code since it made it much easier to do major swaps.