Andrew Chen Archives

A/B testing is fun ;-)
Generating candidates for A/B testing can be a great source of entertainment for your team. It’s easy and fun to generate dozens of potential candidates for a new headline, subtitle, picture, or other element of an important page. It’s also a great exercise in both the qualitative, consumer psychology skillsets required, as well as the quantitative set.

I’ve gather a couple rules of thumb in helping you generate good candidates for A/B testing below:

• Brainstorm the RIGHT way
• Dive down into potential customer motivation
• Go for high variariance approaches
• Test big things first, smaller things later

Let’s dive into each of these…

Brainstorm the RIGHT way
First off, not all brainstorming is created equal – you want to make sure you are going for lots of quantity, that the most senior person in the room doesn’t “run” the whiteboard, and a bunch of other guidelines that you can find in this article on IDEO’s brainstorming techniques. I generally find that after brainstorming individually on dozens of candidates, you can build on very interesting themes and start to coalesce the entire process.

Dive down into potential customer motivation
One important issue is that every product and every page within your product likely caters to multiple needs. Influence, the classic book on persuasion by Robert Cialdini, enumerates many of them. Is it:

• Reciprocation
• Commitment and consistency
• Social proof
• Liking
• Authority
• Scarcity
• etc?

Or alternatively, you may have specific ideas about value propositions or user emotions – for example, a social network like MySpace could be marketed around:

• Customizing profiles
• Socializing and friends
• Media content
• Photos and blogs
• etc.

Who knows which feature is king? The point is, each one of these potential customer motivations and values probably deserves at least one, if not several, test candidates in your A/B test. The fundamental emotions driving your product have a huge likelihood chance of altering the outcomes of your split tests.

Go for high variariance approaches
Similarly, life is too short for the safe stuff. Because of the fact that you throw away all the bad candidates and keep the good ones, it’s in your best interest to try to make the good ones as good as possible! As a result, make sure you try to go for extremely polarizing, high-variance approaches.

For example, make sure you try candidates that:

• … are aggressive and in your face
• … use different graphical elements like videos versus text versus audio
• … are varied in length, like very very long or very short
• … may offend certain subsets of your audience
• … are commanding and direct the user

Typically, in an A/B test I will usually have a control, then a candidate that incrementally improves on the control, and then a couple candidates from left-field. As you try out more candidates and learn from the process, then often times you will start going with more incremental stuff to finish your optimization.

But early on in your experimentation process, remember to go wild!

Test big things first, smaller things later
Similarly, make sure that you prioritize the your tests so that you aren’t testing subtitles and paragraph copy when you could be trying out even more extreme stuff. Things like the user flow, the layout, “hero shots,” and other factors are usually much more important than smaller things like icons or specific sub-labels for forms.

As a result, oftentimes the best thing to do is to rush out some forms to test, then make things prettier and more finalized from there.

Anyway, I hope this was helpful, and if you have more ideas, please comment and suggest more ideas!

I write a high-quality, weekly newsletter covering what's happening in Silicon Valley, focused on startups, marketing, and mobile.