[TWTR] Twitter

One possible explanation for why Twitter has been slow at developing new products is that for many years this was a muscle it didn’t have to exercise.  Twitter got so big so fast, receiving national attention in SXSW less than a year after it was separated from a failing podcasting directory in 2006, that its founders didn’t really have time to understand what Twitter was.  Users and developers figured it out for them.  They grabbed hold of Twitter as an abstract communications concept and concretized it into a product with global reach.  Twitter thought of itself as a text-based social network where one could provide status updates to friends. But users had other ideas in mind.  Barack Obama used Twitter to mobilize supporters; celebrities to share personal details and promoted films; businesses to communicate with customers and run promotions; the rest of us to access breaking news and discuss world events with strangers.  Twitter’s real-time conversations around shows and events made it a natural complement to media properties, who began incorporating tweets into their coverage, amplifying Twitter’s reach and virtuously drawing in more content creators seeking the widest possible audiences¹.

Some of Twitter’s key innovations weren’t even developed by Twitter.  Users introduced conventions like hashtags for topics and $ signs for stonks. Third-party developers seized on Twitter’s open API to build thousands of programs and mobile clients².  

Twitter acquired two popular ones, Tweetie and Tweetdeck, in 2010 and 2011, and launched a native photo-sharing service in competition with those developed by yfrog and Twitpic, coincident with a broader strategic mandate to cease support of third-party apps: 

“…developers ask us if they should build client apps that mimic or reproduce the mainstream Twitter consumer client experience.  The answer is no.”

(Source)

Twitter couched its decision in language supportive of user experience – “we need to move to a less fragmented world, where every user can experience Twitter in a consistent way” – but developers understandably felt burned.  After devoting considerable time and resources plugging holes in Twitter’s service, they were being told “thanks guys, we’ll take it from here”. Whereas many software companies start as apps and aspire towards a platform, Twitter started as a platform, almost by accident, and became an app. 

As Twitter added some no-brainer features like the ability to upload photos and share tweets privately, and explored ways of making it easier for users to sign up and giving them reasons to do so (for instance, by dropping them into relevant timelines as fast as possible), by 2014 it also had its sights set outward.  Besides driving traffic to its own property, Twitter was looking to disseminate content created on its platform to third party apps and media.  As part of this strategy, the company began to invest in the picks and shovels for mobile app development.  Just as Google acquired Android in 2005 to gain distribution for search as the world transitioned to mobile, Twitter stitched together a development toolkit, Fabric, that it believed would foster a healthy universe of mobile apps, creating more surface area for Tweets and more inventory for its advertisers to buy against.  

A mobile developer might use Fabric’s Crashlytics and Answers to stabilize and test their apps and to analyze performance. Twitter conveniently integrated with those development tools with a Supply Side Platform (SSP), MoPub³, through which publishers could expose their apps to ads, and a Demand Side Platform (DSP), TapCommerce, through which Twitter advertisers could buy inventory across thousands of sites and mobile apps⁴. And with access to user emails, Twitter could precisely re-target (“determinisitically match”) users across external properties.  It also launched Twitter Kit, an API that allowed developers to easily embed and create Tweets within mobile apps – EPSN’s mobile app, for instance, could display tweets from notable athletes alongside their proprietary content. So essentially, Twitter offered developer tools for free as a loss lead funnel into MoPub and Twitter Kit, where they monetized through ads.

Twitter wanted the largest audience in the world.  Whether that audience consumed Tweets on Twitter itself or alongside commentary on CNN or ESPN or as part of a third-party app experience seemed less important at the time.  

But Twitter would direct its gaze inward again.  In the 3 years leading up to his ouster, CEO Dick Costollo had spent $1.5bn of cash and stock on a wide range of acquisitions.  Jack Dorsey, upon assuming the reigns as CEO in 2015, promised greater focus on bigger, fewer projects.  Twitter shut down the mobile app development tools, Crashlytics and Answers, and worked to hone user engagement and monetization on Twitter proper. Most of the issues that investors gripe about today were identified by Twitter around this time. Management talked about making it easier for visitors to immediately enjoy Twitter without requiring that they build follow graphs.  It used look-a-like modeling, location, mobile contacts, and demographic data to construct more relevant timelines for casual browsers.  It relaxed the strict reverse chronology of user timelines, placing more emphasis on relevancy vs. recency, and organized tweets by popular topics.  

Twitter’s incremental product improvements were supposed to accelerate the rate at which its 500mn monthly unique logged-out visitors converted into logged-in users.  In 2014, management offered a compelling picture of how much incremental revenue could be realized by nudging the ratio of Daily Active Users to Monthly Active Users (a proxy user engagement) from 48% to just 51%.  But DAUs grew in fits and starts and 5 years later, DAU/MAU declined to below 40%, way below Facebook’s 66%.

With ad revenue tied to user growth and engagement – the fewer users there are engaging with Twitter, the less valuable Twitter is to advertisers and the less data available to accurately target users – Twitter’s lackluster monetization mirrored its uneven user growth. But it also reflected a more general lack of progress in ad formats and technology. In 2015, five years after launching launched Promoted Tweets⁵, Twitter’s ad program was still missing basic features, like the ability to automatically convert a campaign to the creative with the most engagement or attribute actions taken on third-party sites to ads view on its platform.  They had only just launched direct response and self-serve advertising in earnest.  

So they started fixing this stuff, inking attribution deal with DoubleClick (Google) and prioritizing mobile app install ads (and direct advertising more generally⁶) as they pursued the 7mn SMB Twitter accounts for ad opportunities. Twitter even started to test commerce.  But Twitter has struggled to stand up scalable ad tech infrastructure to support self-serve direct advertising by long tail of small businesses and today, most of Twitter’s ad revenue still comes from top-of-funnel brand advertising large marketers.  Third-party ad revenue has stagnated and commerce initiatives have gone nowhere.  The last 5 years have felt like Groundhogs Day, with Twitter harping on the same product and monetization initiatives again and again with sporadic progress: