3 New Ways to Measure the Social Web
Post by Tim Trefren (Co-founder of Mixpanel, Inc.) guest posted at
When most people think of web analytics, they think about pageview tracking;
basically, measuring which pages on a website are being viewed. Pageview
tracking is a well-established technology, but it’s no longer meeting the
needs of many of the most well-known companies in social media. Companies like
RockYou are spending tons of resources
building their own internal analytics tools.
There’s a reason for this: Social media is highly competitive, and the biggest
advantage you can have is data. To improve and grow, these companies need to
gather as much information as they can, and they need more than simple
In the following sections I will cover three of the most important things to
measure for social applications.
1. Funnel Analysis: Measuring Conversion Rates
One critical kind of analysis that social apps require is called Funnel
Analysis. This is a way of measuring conversion rates, which is the lifeblood
of all applications. The term “conversion rate” refers to the total number of
visitors who came to a site, compared to the number of visitors who did a
desired action (such as creating an account or purchasing an item).
What Funnel Analysis gives you is a more granular way of analyzing conversion
rates. Instead of simply looking at signups divided by total visitors, you
figure out the steps that have to be taken to get a user to sign up and
measure the individual conversion rates between steps. As you can see from
the image above, there’s often a pretty steep dropoff between each step,
giving you the namesake funnel shape. (Note: the image uses made up stats and
is for illustration purposes only.)
This more granular look at conversion rates can have surprising results. Let’s
take a look at Twitter’s signup funnel:
- Hit homepage
- Go to signup page, fill out registration form
- Browse suggested topics
- Add e-mail friends
- Search for someone
As you can see, the signup process is pretty complicated, and will benefit
from detailed analysis. We might find, for example, that there’s a huge
dropoff rate (a “dropoff” occurs when many of the people who made it to one
step don’t make it to the next) at the “Add e-mail friends” step. Once we’ve
discovered a dropoff rate like this, we have to figure out the root cause. The
dropoff rate at the “Add e-mail friends” step could mean that users are unsure
how to continue, causing them to leave, or they might not want to add their
e-mail information. We would have to test to make sure.
Ultimately, Funnel Analysis is about finding and improving trouble spots in a
website. With continual analysis, changes can be measured and ideas can be
tested over time.
2. Engagement Tracking: Measuring What People Do
As I mentioned earlier, pageview tracking is becoming less and less
relevant for many web companies. Instead of the basic unit of measurement
being the pageview, they are starting to track more directly relevant things,
like the actions people are taking. Twitter, for example, may want to know how
many tweets the average person sends and what they are searching for, not how
many pages they viewed. Pageviews are just a way of approximating the
information we really want, and as the web grows more interactive, they become
less and less relevant.
Think about this: Sites exist today on which you never actually change the
page. These are highly interactive sites, but they are impossible to track
with pageviews, so traditional analytics tools are useless.
This will only become more common as time goes on and more companies develop
highly interactive applications and adopt
AJAX loading techniques.
3. Visitor Retention: How Many People Come Back?
This next technique measures a fairly complex but extremely valuable metric
for successful web applications.
You can think of Visitor Retention as a measure of how “sticky” your site is.
What we’re really measuring is the percentage of people who come back again
and again. The most common way of approaching this is to look at a group of
users from a single time period (a week, for example) and track their behavior
Here’s an example of a retention table that should help clarify things:
Each row shows the weekly retention rates for a single group of users
(sometimes known as a “cohort”). The first row, for example, is the cohort
seen between December 7 and December 13, 2009. We can see that 15.15% of the
users in that group came back after 1 week, 13.4% after 2 weeks, and so on.
This is crucial information, particularly for social applications, because
most of the value lies in the size of the community. An application with low
retention is like an empty shell — many installs but few active users — and
you don’t want to build an empty shell. You want a thriving, vibrant
Retention is a huge factor in building a strong community for a few reasons:
You don’t have much of a community if everyone is a newcomer (so more old
users is a good thing), and the nature of retention is such that you get
disproportionate returns on any increases you make. Without going into too
much detail, an example would be that increasing retention by 33% might give
you 50% more users in the long run.
Twitter is again a good example for us, as the network has been plagued by
low retention rates.
Twitter may seem successful now, but their low retention rate is troubling. In
the past, companies that seemed to be extremely successful (think early
Facebook apps) ultimately lost their edge because they couldn’t retain their
It’s entirely possible that Twitter itself could be a fad. With such low
retention, I wouldn’t necessarily be surprised — but it is still too early to
There’s a lot to learn about analytics from the frontrunners in social media.
The intense competition has resulted in many new and innovative ways to track
and analyze visitor data.
We covered three such concepts in detail today: Funnel analysis, which lets
you track conversion rates across whole parts of your site, engagement
tracking, which is becoming more relevant than pageviews, and visitor
retention analysis, which helps you understand and optimize the number of
repeat visitors you get.