simplereach.com
  • data news Social-Networks worldview

    The Native Advertising Leaderboard

    You take the latest marketing buzzword, native advertising. Then, you figure out how to make it interesting. Native advertising is all about listicles, so how about a leaderboard? And we’re a social data company that measures content, so let’s rank it by total shares. Make it look good. Make it interactive. And there you have it, The Native Advertising Leaderboard. A nice little marketing stunt…

    Then we started digging into the data we were aggregating. We saw that on just a few dozen publishers, there were nearly a thousand sponsored articles. We found that there were more than a hundred different brands behind those articles and we still hadn’t uncovered all of them. And the brands that were involved didn’t fall into a neat bucket. They ranged from consumer appliances and energy to shipping.

    We kept digging and realized that different publishers performed better on different social networks. One brand might do really well on Facebook, but not nearly as well on Twitter, while another brand would experience the exact opposite.

    Native Advertising might end up being just another buzzword, but seeing all this data brought together, it’s hard to say there isn’t something really happening. And regardless of what we end up calling it, there are a few things that are here to stay.

    Content is the new creative asset for digital brand advertising. Brands will continue to create content to engage with the audiences they have earned on social networks. Publishers will lend their brand, audience and, in some cases, their voice, to helping brands become content creators. And agencies will take part in not only the creative process, but the shifting dollars to distribute content to the right audiences across devices and networks.

    So on national Native Advertising Day, at the Native Advertising Summit, we’re excited to announce The Native Advertising Leaderboard!

    (I hope this does wonders for our SEO.)

  • news

    There Are Many Like It, But This One Is Ours

    It all started with the Fredrik, a bucket of metallic green paint, and a tough decision. We’ve been sharing offices with some great companies over the last few years, but with a number of new hires it was time to get our own place. It took several weeks of searching before we finally found The One, and set about making it ours.

    An offhand joke about painting the columns gold led to several Home Depot trips, A/B testing metallic paint samples, and learning more about the subtle variations of white paint than any of us ever wanted to know.

    Building the desks.

    Zanab testing paint samples in the jungles of Home Depot.

    Ed had been aggressively scouring Craigslist for the Fredrik, a model of standing desk that is no longer in production. After hoarding a critical mass of desks, we decided to pull the trigger.

    Shortly after Thanksgiving we grabbed the desks we had collected from the greater New York area, put our computers under our arms, and moved one block north to our new digs.

    Building the desks.

    Kenn and Ed assemble a desk while Eric works. And yes, we all stand.

    In the first few weeks, other than desks, the office was pretty open and empty so we filled it with mind-controlled helicopters and nerf guns. But then came the fridge which we stuffed with tallboys, followed by a teal couch in a shade impossible to describe in words.

    The space has since filled up with proper office things and we’ve settled into the ebb and flow of daily startup life. But in the interim something interesting has been happening: the emergence of our own culture.

    Things like Burger Friday, where our lunches have become pilgrimages in pursuit of delicious burgers, complete with mandatory punishment for failing to follow the rule: everyone must order a burger.

    And Thursday talks, which are basically informal presentations, cover a wide range of topics from bloom filters to the art and science of espresso pulling. Russ recently gave a talk on ballistics, Roman the fundamentals of bitcoins, and Heyjin the visualization intricacies of d3.js.

    A Thursday talk on design wouldn’t be complete without pizza and wine.

    Outside of weekly events, there are smaller patterns, jokes, and incessantly trolled music videos that have come and gone. Hubot is an integral part of the team and Nagios is chatty as ever, while an endless debate rages over Guild Wars vs. World of Warcraft, and one of us continues to insist that the J.J. Abram’s Star Trek really wasn’t that good.

    When we’re not hunting down burgers or crashing helicopters into our faces, the office is blanketed in the quiet hum of concentration and keyboards. There’s a curious rhythm to it all as we move between work and play, the bursts of laughter and activity a counterpoint to the intense focus of building something amazing.

    Hard at work.

    It’s been fun to see the culture evolve in the short six weeks we’ve been at the new place. The process has been completely organic, a product of the interests and interactions of a very talented group of people. Looking forward to what 2013 will bring.

    Good morning! Have a donut.

  • data Social-Networks

    Santa Got More Shares Than The Apocalypse

    After surviving the Mayan Doomsday unscathed, and enjoying the holiday season, some of us at SimpleReach began to wonder what sharing patterns had looked like on content related to the now laughable (actually it was always laughable) Apocalypse. So we decided to take a look into our rich trove of social data and analyze how sharing activities on networks were influenced both before the impending doom and after it was safely past. We were also curious about what kind of impact the potential end of the world had on a more prevalent topic – Christmas and the holidays. We figured that different topics would be less likely to have overlaps in terms of social activity, and we would be able to see clear and independent trends.

    To get started, we created an extensive list of relevant keywords for both topics and queried our content database for matches in tags, categories and article titles. Each list consisted of about 15-20 of the most common keywords:

    • Christmas, xmas, Santa Claus, …
    • doomsday, apocalypse, mayan prophecy, …

    To begin our analysis, we divided social activity for the two category sets into three time-based buckets: the week leading up to the 21st, the 21st, and the week following. We focused the research on activity across three of the top social networks – Twitter, Facebook, and StumbleUpon. Broadly, this is how social activity behaved across those networks:

    Actions and referrals across networks

    Different patterns in activity are immediately noticeable.

    For example, on Twitter there were more Apocalypse related social actions and referrals on the 21st than in the entire preceding week. The number of social actions on that day were 1.33 times the total combined actions for the past week, while referrals were a bit higher at 1.67 times. This wasn’t the case with Facebook, although Facebook experienced the highest activity on the 21st. Curiously, on Stumbleupon, there appears to have been more interest in the Doomsday after the predicted day in terms of actions! There were 19,953 social actions in the week after compared to 15,973 actions in the week before – around 25.2% more activity.

    To follow up on our hunch about Christmas articles being affected by Doomsday, we looked at pageviews on Christmas related articles in the week before the 21st and on the 21st itself. (In the week after the 21st there was obviously a lot more Christmas activity.)

    Comparing Christmas and Doomsday refs

    As expected, social actions were greater on Christmas articles than Doomsday articles across all networks. However, on the 21st, pageviews for Christmas content across Twitter and Facebook were lower than the corresponding numbers for articles about the Apocalypse. Doomsday was obviously a hot topic on Facebook and Twitter, temporarily overtaking Christmas in terms of pageviews. For StumbleUpon, in contrast to Twitter and Facebook, the pageviews on the 21st were higher for Christmas related content than Doomsday related ones.

    Social actions are important because they typically drive referrals from a specific type of traffic source, so we took a closer look at trends in referrals over our selected time periods. The types of referrals we are looking at are an aggregate of pageviews driven by social media (social referrals), and pageviews driven by search (search referrals). While there are always visitors who reach a site directly (i.e. no referral), we are not including them in this analysis. We plotted referrals on Doomsday content and Christmas content on a day over day basis for a comparative illustration in traffic trends.

    Daywise referrals
    For articles about the Apocalypse, we noted that search was more influential than social in terms of driving traffic. On the 21st, there were 2.09 times more search referrals than social. On the other hand, when it came to Christmas articles, people were more likely to get there through social channels than through search. In particular, on Christmas Day, we saw 3.94 times more pageviews from social compared to search. Interestingly, for Christmas related content search referrals peaked on Christmas Eve, while social referrals peaked on Christmas Day. Our conclusions are anecdotal, but we like to imagine search referrals on Christmas Eve correspond to last-minute gift-finding panic, while social referrals on Christmas Day correspond to the love (or hate) of said last-minute gifts.

    All in all, we’re glad that everyone survived the apocalypse (and the holidays). While we realize there is no longer cause for concern, we’ll keep monitoring the social channels. Just in case.

  • data Social-Networks

    Turns out, Klout does work

    Well-known, recognizable publishers tend to have an impressive social media presence, with Twitter followers and Facebook fans in the millions. Tweeting a single article can result in an overwhelming deluge of responses, which is awesome, but how does a content creator decide where to reciprocate engagement?

    We work with a lot of large publishers and have seen this problem again and again, so we thought we’d help out with some smart filters. The question is, what do you filter for? We threw around a number of ideas: sentiment analysis to find the most passionate, oldest first, newest first, or even filtered by those with awesome profile photos. In the end, we realized that perhaps there was an even better approach; partner with a company already working on this problem – Klout.

    Klout’s goal is to measure influence, which is a tricky endeavor and not always clearly defined. We needed something a little bit more tangible to measure our results against, so we focused our analysis on determining what would help publishers drive Twitter traffic. We analyzed just under 3 million tweets containing links from approximately 5,000 publishers. As it turns out, the overall volume of tweets is the single largest determinant of what drives traffic from Twitter back to a site.

    Users with higher Klout scores have, on average, much higher retweet rates than users with lower Klout scores. For example, a user with a Klout score of 70 generates, on average, four times as many retweets as a user with a Klout score of 35.

    retweets by Klout score

    Basically, users with higher Klout scores are better able to propagate and amplify messages through their social networks.

    In order to properly conduct our research, we needed to settle on a definition of a retweet. Since there are a few, we decided to start out with the official Twitter definition. As we got deeper into our work, we realized that we needed to include not only the more visually common definitions like “RT” and “via,” but also some of the more esoteric formats like “MT” (modified tweet) and “HT” (which can mean either Hat Tip or Heard Through). For more information on the official and unofficial definitions of a retweet, see this Twitter post.

    The following chart shows a breakdown of retweets by various formatting options:

    relative popularity of RT formats

    We found that simply using the retweet counts provided by Twitter missed retweets where people add the letters RT and sometimes commentary to someone else’s tweet, like the format shown below:

    rt RT format

    Twitter only recognizes retweets where the original tweet is not modified:

    official RT format

    We were surprised to discover retweets using the RT @twitterhandle format are more popular than the official Twitter retweet format — by a factor of 1.46. That means for every 100 tweets in the official Twitter retweet format, there are 146 unofficial retweets in the RT format.

    Ultimately, in terms of driving traffic, it doesn’t matter if it’s a tweet, official retweet or an unofficial retweet. It’s the overall volume of tweets that determines referral traffic.

    Our Solution

    Nerding out data is great, but we wanted publishers to be able to do something practical with the results we found. We thought it would be cool to show them all of the Twitter activity surrounding an article:

    content show page

    Using the results of our analysis, we built out a real-time dashboard for each article that includes every tweet that mentions that piece of content, sorted by Klout score. The prioritization of tweets allows publishers to sort through the deluge and engage their audience. As the social media ecosystem grows, so does the need the for these types of filters.

  • Technology

    A Big Stage for Cassandra

    The tools to handle big data are a hot topic these days and part of a growing market. In processing 100 million social actions every day, we needed to tap into some of those tools to support our big data at SimpleReach. To give content creators the market-leading view into what’s working on social channels, we decided to leverage several novel techniques for storage, processing, analysis, and display. Amongst the many tools we utilize, Cassandra is the one that does the heavy lifting for our massive volumes of data. We’ve learned many hard-earned lessons along the way and were excited to share some of them at the third annual Cassandra Summit on August 8, 2012, with both myself and Russ Bradberry, our Principal Architect, presenting.

    It was great to see our enthusiasm for Cassandra matched by so many others. The attendance was outstanding, with over 800 people talking about the latest trends and use-cases from the experts. This year’s attendance more than matches the previous 2 years combined. The speakers brought a depth of big data know-how, with showings from Netflix, Disney, Ebay, Ooyala, Sourceninja, and more.  We were excited to see how many of them (like SimpleReach) have contributed work back to the community through open source.  There were also technical deep dives into the internals of Cassandra given by the developers. And Datastax, the excellent host of the Cassandra Summit, even provided a lounge with popcorn, cookies and beer where you could ask the experts questions.

    Though we learned a lot from the content, and enjoyed the oddly cool Japanese drum show to kick off the conference, the highlight for us was getting to share our experiences at SimpleReach with the Cassandra community. At SimpleReach we love Cassandra for it’s ability to handle high-volume, high-velocity data ingestion and do so while scaling linearly.  And even though we know that it’s not the right tool for the job all the time, it’s done very well for us where and when we’ve needed it.  Here is the talk I gave about Cassandra and the SimpleReach use-case:

    The way we often talk to Cassandra is through one of our many Node.js applications. Helenus, the Node.js driver for Cassandra created by Russell Bradberry was also featured in a lighting talk.  Russ talked about the performance boost with running queries through Helenus and Node.js:

    We were also fortunate enough to be able to share the story of SimpleReach’s product and not just the technology on The Cube:

    The 2012 Cassandra Summit was a great opportunity for SimpleReach to show the world that our novel approach isn’t just in the way we view social media and its interactions, but also the technology that supports it. We’re excited about continuing to push the limits of big data tools, alongside this emerging community, for years to come.

  • data Social-Networks Technology

    Facebook Bots Are NOT Stealing Your Ad Spend!

    Bender Bot

    In a recent Facebook post a rather frustrated individual was noticing that the number of referrals (as measured by his analytics software) wasn’t matching the number of ad clicks that facebook was reporting.  This led the person to do an expansive investigation, including writing his own analytics software, to find out why this was happening. His conclusion was that dirty nasty bots were clicking the ads and therefore, the analytics software wouldn’t pick the referrals up.

    There were a few false assumptions made in the post. The first is that the traffic coming in had JavaScript disabled.  If this was the case then the JavaScript analytics software would not detect the incoming traffic and therefore would not be able to log the result at all.  The second false assumption is that bots can not execute JavaScript, or that bots can not pass the referrer. Bots can do anything your browser can do, and can hide it just as easily.  I’m sure that if this individual looks at their analytics again, they will see a disproportionately high amount of direct traffic.

    So if it’s not evil bots, then what’s happening?

    My best guess is that there is a data loss due to protocol transfer. Facebook gives you the option of secure browsing, just click on “Security -> Secure Browsing” in your Facebook “Account Settings” area.  This will enable you to keep the communication between you and Facebook secure so that others can not see what you are doing.  Other social networks like Twitter also offer this. Google Plus even forces it and makes you browse securely.

    When you browse a secure site (as shown by the https in the url), you can remain confident that what you do on this site is secure.  Here is where it gets tricky, when you browse from a secure site to a non-secure site (known as protocol transfer) all information stored about your session is lost.  This includes the place you came from (known as the referrer).  When you click on an article from Facebook while having secure browsing turned on, the fact that you came from Facebook is lost.  This is how things are intended to work as according to the rules of the internet.

    What’s the solution?  Well there is no 100% solution. Some social networks such as Twitter internally transfer you to the correct protocol before redirecting you to the intended site. This method preserves the referrer.  Others like Google Plus have stated that they will not be doing that.  Publishers have several different workarounds; though none of them will get all of the information.  One option is that you could add a tag to your links, such as http://simplereach.com/blog/better-than-real-time?ref=facebook.  Another option is to make sure all your content is hosted on a secure site. The main issue with both of these is when people cross link the article on a different social network or site, which can give you misleading results.

    I’d be willing to bet that the only Facebook referrals this individual sees are users that have not enabled secure browsing on their Facebook accounts.  While secure browsing is a necessary part of the internet, there’s still a number of people who don’t understand it’s implications which often results in a misattribution of traffic.

  • worldview

    Better Than Real-Time

    Want to predict the future? Track the cause, not the effect. When content links are shared on sites like Facebook, Twitter and Pinterest, there is predictable resulting traffic. If you know how to correlate social content sharing into the traffic it later drives, you are way ahead of real-time on-site data. By the time a piece of content is trending in real-time, an opportunity to engage with your audience has already been lost.

    Imagine your website is an island resort and, as a good manager, you want to track the guests on your island and tailor your business accordingly. One approach is to count all the people moving around the island right now. There’s value in that real-time data, however you’re always playing catch-up with each new influx of visitors. You need time to figure out who they are, where they came from and what they want.

    A better approach, and what SimpleReach does, is to count the boats approaching the shore. After observing hundreds of thousands of boats and the resulting visitors, you know by the boat how many people are onboard, where the boat came from, and the varying hours it takes each boat to unload its passengers. You’re now seeing your visitors before they arrive on the shores of your site.

    You start to notice additional things about the arriving customers, how long they stay and what sorts of things appeal to them. You notice that when you get an increase of boats from one destination you also get more from another destination, as if their networks overlap. You begin tracking the features and marketing that appeal to different destinations, and how many visitors each combination drives.

    A Justin Bieber tweet? A giant container ship commandeered by sixty thousand shrieking, teenage girls who stay for 37 seconds and buy funnel cake. Your great uncle’s oddly compelling geology tumblog? A tiny dinghy, but one that shows up with very engaged visitors and more each week. When you track these boats at the dock over a period of time, you know your customer before they know you. This is what SimpleReach does for some of the largest content creators in the world – analyze every share of their content across every major social network or social content aggregator, and tie those shares directly to on-site traffic.

    Though tracking causes rather than effects makes ample logical sense, Team SimpleReach takes a data-driven approach to evaluating our hypotheses. We ran various predictive models across 2 million articles over the course of several months, and showed definitively that predicting article-level traffic over the coming hour was meaningfully improved in explanatory power and reduced in variance when social actions were significant inputs. In other words, if you’re not tying social actions to its resulting traffic, you’re missing critical insight into the largest traffic-driving force since search.

Company
Connect
Contact
  • 41 W. 25th St, Floor 8
  • New York, NY 10010
  • info@simplereach.com

<3 NYC