One of the most interesting sessions to me at Web 2.0 Expo was fellow analytics blogger Avinash Kaushik’s session called “Click the Big Red Button : Tips & Techniques for Optimizing Conversion and A/B Testing”. I was looking forward to it because even though I’ve been big into testing for years, I still feel like the majority of the web world is so far away from embracing testing to improve web results.
Avinash is also embarking on a new adventure as an independent consultant after leaving his analytics job at Intuit. He’s currently speaking, has a book coming out, and consulting for Google Analytics. The following is an overview of his presentation.
The top metric that most people are looking at is increasing their conversion rate. So how do we increase conversion rate?
1. Segment, segment, segment.
Avinash wisely pointed out that looking at a site’sA overall total conversion rate is not that easy to take action on. By using web analytics you can segment conversion rate into all kinds of groups. You can segment by referral source, segment by landing page, segment by search terms, etc. Working then to improve the conversion rate of these segments is easier to monitor and improve.
2. Conversion rate needs friends
Don’t look at conversion rate in isolation. Look at average order size, visitors, bounce rates, and other metrics in conjuction with your testing on conversion rate.
3. Don’t look at conversion rate in isolation
Similar to the above two points, you need to put context around conversion rate numbers. For example, by offering your product for half price you’ll most likely increase your conversion rate. That’s great right? Well, no, because you’re making less revenue.
4. Fall in love with abandonment
Abandonment is the only way people can leave your site and hurt your conversion rate, so naturally improving the abandonment rate on each page of your website will help your conversion rate.
5. Determine the true opportunity
It’s impossible to actually convert everyone. Part of your audience is going to be landing on your site by mistake, some of the users are just crud and will never be customers, but what real opportunity can you get? You need a good idea of your true audience to go after it.
So what is the golden answer to all this? Testing.
Avinash said people far too often use the HiPPO strategy on their web decisions. HiPPO stands for “Highest Paid Person’s Opinion”. The HiPPO for your website is just one opinion, no matter how smart they are. You need to test to let the real users decide what works best. 80% of the time the HiPPO (or you) are wrong about what’s going to work best on your site. And you must continue testing, what’s great today, is stale tomorrow.
What methodologies can you use to test?
Methodology #1: A/B Testing
Hopefully most of you are familiar with A/B testing, and Avinash outlined the pros and cons of this strategy, but said if you’re doing no testing today then you should start with A/B testing immediately. He gave a great example of how when he was at Intuit they tried out this amazing and sexy Flash shopping cart solution. Avinash thought it was great, it kept users from having to refresh pages, it was slick, it was easy to use, and he saw no way it could fail. In an A/B test with Intuit’s old normal shopping cart, it failed miserably. Why? Users apparently were used to the old normal way.
Essentially, A/B testing is simple, easy to do yourself, but it’s limited and you have to keep it simple to be accurate.
Methodology #2: Multivariate Testing
Multivariate testing is a testing format that allows you to test multiple parts of your page at once. It allows you to test alot, most tools provide great additional tracking, you can visual results pretty well, but it’s more complicated. Avinash gave an example of how multivariate testing improved the downloads of Picasa on Google’s Picasa download page by 30%.
The prime tools in multivariate which all have different strengths:
- Offermatica – The most marketing focused.
- Optimost – More technically focused.
- Sitespect – Don’t need to implement javascript code, they do it by packet sniffing between the server and user.
- Google Website Optimizer – The only free tool, but have to be using Adwords.
Methodology #3: Experience Testing
This is the art of testing sets of multiple pages independently. Such as funneling Windows users vs. Mac users. There are no current tools that do this, so you have to be sophisticated and do this yourself. However, this has the most power and can lead to the biggest results.
A great talk overall, and Avinash was fun and engaging while he spoke. I think he has a bright future as an analytics evangelist.
