Company Blog

Community Measurement: New Analytics for All

Building a community is hard. Buckets of blood, sweat, and tears go into it. For over 2.5 million sites, Disqus is a part of that community building process. But sometimes it can feel like guesswork. We’re now making that arduous process of increasing readership even easier: announcing analytics for all publishers.

Every site can benefit from analytics, and with recent infrastructure upgrades we’re now able to offer these services to everyone. Previously, only a select number of publishers that paid us a fee every month received comprehensive analytics that included metrics like most replied to comments and the social login breakdown of their users.

The initial cut of our new analytics cover two key areas:

• Commenting: number of comments, number of votes, and top comments
• Revenue: money earned from using Promoted Discovery

Admittedly we’re starting simple, but it now allows you, for example, to see how posts you make coincide with how many comments are posted. Based on our analysis of the last 2 months of data, the most comments are posted on Wednesdays and the least on Saturdays. See if your community’s trends match up!

Over the next few months we’ll continue to add insights like which commenters are new versus returning and traffic referrals from within the Disqus network.

Check out your new analytics now. Have feedback or a specific metric you’d like to see? Let us know in the comments below!

Trying out this Go thing…

Last Thursday, May 16th, we shipped our first Go project into production.

Disqus has a history of using one tool for the job and hammering through everything. Until now, that tool has been Python. This mentality has been extremely beneficial for us since we are a small team. Everyone is able to jump in to any piece of our stack with minimal effort.

So what was the problem?

Our realtime was having issues. It was slightly short of… realtime. The original realtime backend was written as a pretty lightweight Python + gevent service that handled a few basic tasks. Realtime consists of four components: a queue, a nozzle, a transformer, and a publisher. Everything was Python, except for the queue, which was something implemented on top of Redis.

The rate of messages has since increased, and our backends were having a bit of an issue scaling to our needs. We had 4 servers, each at maximum capacity, and our end-to-end latency was at best a few seconds. At worst, minutes. At peak, we process 10k+ messages per second.

Realtime is a pretty critical component to Disqus, so we decided to try something a bit different.

Why Go?

Go was initially very attractive to us. The language felt very natural coming from Python backgrounds, and the performance approaches C levels. The goroutine model and channels are very easy to work and immensely powerful to manage concurrency.

In roughly a week’s time, I went from initial commit to shipping replacement backends while only having a cursory level of Go knowledge. To me, that’s highly impressive. Our realtime end-to-end latency is on average, less than 10ms, and currently consuming roughly 10-20% available CPU on one machine at peak.

Overall, this should yield a much more responsive experience for our users, and gives us a lot of room to grow in the future. I look forward to trying out more Go and seeing where it can fit into our stack.

THE FUTURE

Would I use Go again? Absolutely! We were very happy with our results and can only imagine it getting better the more we learn. I highly recommend anyone giving it a chance and checking out what it has to offer for you.

I look forward to trying out Go for more projects and start contributing back libraries.

Does this sound fun to you? Looking for a job writing Go? Come work with us at Disqus!

A Discussions Editor for Mere Mortals

We’ve been busy building lately. So what better time to introduce one more thing?

It’s called the Discussions Editor and it allows you to 1-click update attributes of any discussion on your site. For example, you might update the title or link associated with a discussion in Disqus to keep it up-to-date after changing it on your site itself.

Simply click into any cell in the editor, enter the desired information, and hit enter. The attribute will be instantly and automatically updated. It really is that easy.

You can learn more about this useful new tool at the Discussions Editor F.A.Q. Enjoy!

Time — Is On Your Side (II)

Last year we talked about our goal to turn time spent into time invested with the new Disqus.

We continue to be mildly obsessed with this concept of time as a top metric we should be thinking about. In fact we have been working with others like Chartbeat, who have embraced this as well with their Engaged Time metric. Their data also reaffirmed some of our own that show, on average, over half of all visits now scroll down to Disqus.

More recently, we got word from sites that they were seeing marked differences in time spent when using the new Disqus, similar to what we saw a couple years ago with the older version. This time, though, we wanted to see if we could go past anecdotal or internal data, and instead get numerically reliable measurements based on a public standard. We also wanted to see if we could find samples that would show results for a number of commenting systems, not just Disqus.

We concluded comScore numbers tend to be the most standardized, but since they don’t provide section-specific data, we needed to use sites for which overall traffic is highly skewed towards pages that have commenting. We therefore turned our attention to popular blogs that get a sizable amount of traffic to ensure the sample size was big enough and external factors wouldn’t skew results too much. This left us with 5 of the Technorati Top 100 that met the above criteria, and had used Disqus and at least one other common comment system in the past year.

Graphing out their comScore data for the past year, we found an across-the-board increase in average monthly time spent per visit when switching to Disqus (Red State, Talking Points Memo, The Next Web in September 2012), and a decrease when switching away from Disqus (Engadget, also in September 2012):

Techcrunch data, which we had going back a few years, was especially interesting. They have tried four different platforms in that span of time, for two months or more each, so we could get a fairly comprehensive breakout of engagement by platform:

So while a number of factors go into website technology decisions, and one should never be too conclusive about causation, the trends we saw were very encouraging. We spend a huge amount of effort thinking about what attributes meaningfully result in higher engagement, as defined by a user’s time spent with a given community. We’ve found that focusing in on our “3 Re’s” — Retention, Reactivation, Recirculation — has gone a long way in measurably achieving this goal. In fact, after recently hitting 1 billion unique monthly visitors, we also discovered Disqus now accounts for over 10 billion minutes of time spent each month by those visitors in aggregate. That’s more than 20,000 years! Naturally, we want to make sure it’s time well spent.

If you’d like to hear more on this and will be in the New York area next week, our friends at Chartbeat will be hosting us at their offices during Internet Week. We’ll share some more learnings about optimizing for time-based engagement. Click here to RSVP while space still remains.

What’s Cooler Than a Billion Monthly Uniques?

Last month, Disqus achieved a significant milestone: our network hit a billion monthly unique visitors. No matter how you slice it, it puts Disqus in a select category of ubiquitous web services that millions of people use everyday.

Some other supporting data includes:

• 7 billion monthly pageviews
• 100 million user profiles
• 2.5 million site installs
• 1 million Wordpress plug-in downloads

But in and of itself, taking meaning from numbers this large is hard. And of course, Disqus isn’t a destination site so the numbers don’t lend themselves to easy comparisons. For the team at Disqus, they simply remind us that with great scale comes great responsibility. That far reaching impact is what motivates us.  

To mark the occasion, we wanted to do something a little different. First, we launched a new product today called AudienceSync. See our tandem post on that here. It enables users to port their profile information to their favorite sites’ registration systems. It’s great for publishers who want to tap into the largest audience of commenters on the web.

Then we wanted to dive a little deeper into our data and look below the surface of 1 billion monthly visitors. Comments and the community dialogues they make up are poorly understood and there’s little data out there about the patterns and behaviors at work. So our data engineers did some number crunching and we’re excited to start sharing some of it.

The Majority of Comments Are Tweetable

One of the first things we were interested to learn more about was a way to classify comments themselves. The character count of comments paints a picture of natural tendencies formed by other communication mediums. One of our findings is that the majority of comments fall under 140 characters: most comments are tweetable. (Conveniently, this already a feature of Disqus. Give it a try, it’s fun.) And then there’s a long tail of comments that we’ve categorized this way:

• Poems and Paragraphs: Statements more fully formed than can be captured in a Tweet.

• Speeches and Soliloquies: Fully formed arguments and ideas.

• New Posts: Readers writing their own article or full reply to the original post.

(Download the full image)

How channels like Disqus and Twitter can seamlessly pull audiences into a true discussion is something we’re continually thinking through. For publishers, identifying the hidden writing talents among their readers is another key takeaway. Readers increasingly look at comments as their chance to be in the story and as this data shows, some even want to rewrite the story.

Community Voting Habits Have Extremes

Voting on comments is an extremely popular function of Disqus. It’s a lightweight social action that keeps even those not leaving comments participating. It also serves as a crowdsourcing mechanism to surface the best comments. In this analysis, we wanted to know what it takes to earn an upvote and how that might differ according to the type of site you’re on. And again, character count told us a lot. By categorizing over 200 super active Disqus communities into 40+ categories, a wide ranging upvote-to-character ratio emerged.

(Download the full image)

What really surprised us were the extremes between fan sites. Fan sites are big users of Disqus. Fans of anything are by definition a community. They share a common passion. But in this analysis, sites dedicated to gamers and mobile fanboys showed rampant upvoting. Readers there aren’t so much having fully formed discussions as they are likely trading one-liners. Whereas sports fan sites were on the opposite end of that spectrum, with basketball sites showing an average character length per upvote of 42 and hockey a whopping 253 respectively. (It is playoff time after all. Passions are running high.)

Better understanding the breadth of community types on Disqus will be a focus of future research we do. Seeing and engaging with the galaxy of sites that use Disqus is already a reality. It’s called Gravity. Check it out if you’ve haven’t already.  

Replies Are an Active Community Metric

We get asked a lot what makes a great community. There are many components to a community. But here we wanted to start to lay the foundation for tangible data that moderators and bloggers could look at to judge just how active or vibrant a community is. So we did an analysis of our most active sites and looked at one key indicator: replies. Replies to comments (and replies to replies) are a sign that community members are interacting, that they’re talking with each other. In this analysis, sites focused on music, politics and entertainment were the runaway most active winners.

(Download the full image)

You’ll see more data releases from us in the future. If there’s anything you’d like to see us examine, share a comment below.

So What is Cooler than a Billion Monthly Uniques?

Finally, we wanted to mark this occasion by sharing a little love to you, our users. We don’t always get things right, but everyday we hear great feedback from you, our own community. That’s what’s cool to us.

So we put together a little video about you and for you. Enjoy.