At ShareProgress, we’re big fans of using data science in our work. But when sharing our excitement with others, we’ve discovered that not everyone knows what we’re talking about.

To help clear up some of that confusion, we’ve decided to create a few posts explaining some key concepts of data science. First up: the A/B Test!

I’m so freakin’ excited!

Here’s how it works: Let’s say you have a bunch of people you want to invite to your birthday party.

I don’t know if the Olive Garden will be able to seat us all together.

You have two ideas for an invitation email: one about birthday cake and one about presents.

Instead of just going with the one you like best, you randomly choose a few people and send them the cake email, and then you randomly choose a few more and send them the presents email.

Cake vs. Presents

You wait a while, and then you see which people have RSVP’d for the event. If more people RSVP’d from the cake email, you send that to everyone else who you’re inviting. If more people RSVP’d from the presents email, you send that one instead.

If you’re only inviting a few friends, you might not get enough RSVPs to be sure which email is better. But if you invite hundreds or thousands of people, there’s a good chance you’ll discover that a lot more respond to one version over the other. In that case, you’ll know with confidence which one is the best choice.

Everybody loves cake!

For nonprofits and companies, the same approach can be applied to the emails they send: asking supporters to make a donation; selling subscriptions to online services; telling people they should upgrade to the newest smartphone.

“I hear this one comes with a 40 megapixel camera, an 8″ screen, and the ability to time travel four minutes into the past.”

A/B Testing can actually be used for asking people to take any type of action, as long as you can send different messages to small, random groups and determine whether those people did or did not take the action in response.

It’s possible run an A/B Test with more than just two options, using different groups for each one.

And it can even be used for differences beyond just language, like how information is presented or what imagery is shown.

That’s it! Now you know about A/B Testing — you can visit the links below for more information, and check back soon for more data science concepts explained through animated GIFs.

Helpful Links:

• A/B Testing (Wikipedia)
• The A/B Test: Inside the Technology That’s Changing the Rules of Business (Wired)
• Is Social Media A/B Testing Worth It? (ShareProgress)