Introducing much better segmentation

The images in this post are out of date.

Today we are announcing a new feature called Segmentation that will enable you
to analyze your data retroactively and perform even more complex data analysis
then ever before–of course it’s all in real-time. This is a huge step towards
our overall goal: To help the world learn from its data.

Segmentation is extremely flexible and therefore it’s often tough to describe
which problems it may solve because it solves so many. To help introduce this
new feature we’ll go over some simple use-cases that, hopefully, will excite
you.

A really simple example is if you have an ad campaign on Facebook and you
would like to understand the demographic of users who ultimately purchase.
Here we’re curious and filter on users who have seen a specific ad campaign
that we’ve created and are only female purchasers. More importantly we are
curious to see the age distribution or breakdown of that type of customer.

This is extremely powerful because not only are you seeing deep insights into
your data but you’re doing it in real-time on potentially a large set of data.
It’s generally a pain to wait to see the result of a question you have about
your data because as marketers or analysts we generally ask our questions
iteratively until we have that “Ah ha!” moment.

Keep in mind that Ad Campaign, Gender, and Age are defined by us. You can
define any segments you’d like. For example, if you were a social game you
could define character type, source, and item purchased instead.

Segmentation now accepts any kind of data that may be numerical. If you know
the number of invites a user may send during their sign up process to increase
virality of your product, you can now see a statistical distribution of that.
This is extremely powerful and helps you understand your virality. In general,
an average may not be truthful enough.

Other examples of numeric data you can send us: age, page load time, number of
items selected in a list, anything with money involved, coins or credits, etc.

We also support true/false values and have plans to support dates soon which
will let you ask questions such as “Everyone who signed up after this date.”

If it wasn’t already apparent this type of analysis is fully retroactive. You
do not need to define any queries/questions up front as we will look back at
data you’ve already sent us and analyze that.

We believe iterative data analysis is what our customers want and is a great
way to find insights in your data.