Reading reactions about the Instagram purchase by Facebook this morning (read the WSJ announcement here, Instagram’s CEO Kevin Systrom comment here), I can’t help but quickly comment. You might read this analysis elsewhere as I guess I am not the only one seeing through the logic of the price paid: $1bn. For a 2 years old, 12 people team, with no technology innovation whatsoever. Wow. Market forces at play.
Most comments I have read so far focus on the product aspects of the acquisition: “With Instagram, Facebook is going to have a much better photo sharing app than what it has developped on its own”, or anything along the same lines.
However I believe that the real reason for such a deal is not technology or product, but eyeballs and TSO. Both on laptops and on mobile phones. At 16% (US), Facebook is a clear leader. TSO, the amount of time a user spends online within the boundaries of a given service, is one of the most strategic variables for Facebook, as it is what drives their advertising revenues. Like with old-days television. (Facebook doesn’t monetize yet their mobile app through advertising, as they say they have not figured out yet a good way to do it — most mobile advertising is too intrusive to the user experience, and drives the user away from the app in most cases. But Instagram — a fully mobile app so far — should easily transfer their experience onto the larger screens, where Facebook knows how to push ads).
Who needs technology? Virality + Stickyness + Growth
In reality, the deal today is all about the virality and stickiness of the Instagram experience which helped them secure significant mobile TSO: they have managed to built a community of users who return multiple times daily, regardless of whether they are on FB or not. They engage quickly with the service virally, and remain engaged throughout the day for months. Add rocket growth in the past few months, a sure sign that a service has gained acceptance: the Instagram user based grew 30 times in the past year, to 30 million last month. A winning formula.
Kudos for a brilliant setup
The smart move was Instagram’s when they chose to not integrate their app with Facebook last year. It must have taken balls to resist the temptation, but they made it. Because the service was growing independently of Facebook, capturing TSO and growing their user base very fast, they became a threat.
With the $50mm round they were raising at $500mm value a few days ago, they were on the path to becoming an unaffordable threat, in both senses. Facebook had one bullet, and they fired it quickly. Investors at $500mm valuation would not have agreed to selling at a small premium. Hence the offer that made the deal, at a x2 valuation: $1bn.
Instagram clearly outsmarted Facebook: their $50mm round announcement was a smartly designed, well executed bait.
4/09 3:33pm update: On 2/29 Facebook announced their mobile advertising strategy. It has not been rolled out to everybody as I write this. Venture Beat has a good analysis here: “Facebook’s mobile ad strategy is a risk for Facebook and its advertisers (analysis)” (http://venturebeat.com/2012/02/29/facebooks-mobile-ads/)
In one simple fake video, Google has created a level of over-hype and over-expectation that their hardware cannot possibly live up to. — Blair MacIntyre, Georgia Tech —
Ever since my technology adolescence days, when I was studying data structures, machine learning and knowledge-based systems at UTC (Compiegne, France), data has been core to all the hardware and software businesses I was involved in. Whether it was UNIX servers (and the new data structures, file systems and databases they brought with them), business activity monitoring, telecommunications billing systems, speech recognition, or mobile image processing, data was central.
Startups founders I advise these days will tell you how much of a b*** I am about the data model when developing a new service. With all due respect, those who start development with the UI and think they will just plug in some database to support their service are in for serious surprises — not to mention costs.
The latest example that comes to my mind is Vitogo, a self-tracking workout app for iPhone. Great interface, great design, but… it doesn’t work. According to the company, the app is now undergoing a “major rewriting” that has been taking months so far. I suspect the data model (or lack of a well though-out model, from inception) is central to the issue. Anyone familiar with what is going on at Vitogo, please comment or correct my assumption.
So when the industry started to talk about Big Data a few years ago, I got Big Interest. I won’t spiel here a tutorial about what Big Data is, for fear of boring most of you. Enough to quote Noreen Burlingame in her well-written, concise “Little Book of Big Data” (2012; available here in a kindle reader version):
"Every day of the week, we create 2.5 quintillion bytes of data. This data comes from everywhere: from sensors used to gather climate information, posts to social media sites, digital pictures and videos posted online, transactions records of online purchases, and from cell phone GPS signals — to name a few. In the 11 years between 2009 and 2020, the size of the "Digital Universe" will increase 44 fold. That’s a 41% increase in capacity every year. In addition, only 5% of this data being created is structured and the remaining 95% is largely unstructured, or at best semi-structured. This is Big Data."
One of the core emerging technologies for dealing with Big Data is Hadoop. Apache Hadoop is an open source platform build around Map/Reduce, a “seed” technology developed by Google (and inspired by LISP — no wonder I like it… there’s no denying one’s childhood love) to address their needs for indexing and analyzing web data, and dealing with the flow of searches hitting their data.
Several companies have sprouted up, that enhance / simplify / complement Hadoop with specialized layers and applications. A good overview of the current Hadoop landscape (suppliers and major users) can be found on Wikipedia here. And then there’s the analytics world, with the likes of ClearStory Data and Palantir Technologies.
Among these players, Hortonworks is particularly notable. Hortonworks was created by Yahoo! and Benchmark Capital in 2011 to take on the Yahoo! contributions to Hadoop. It is an independent company and one of the rising stars of Big Data, together with Cloudera and a few others.
The legend says that the Hadoop name comes from that of one of the original authors’ son’s toy elephant. Hence the elephant in the Apache Hadoop logo. How inventive of Yahoo! and Benchmark to have called Hortonworks after Horton, Dr Seuss’ elephant… I guess ‘Babar’ was protected by a strong copyright.
Why did Dr Seuss name his character Horton?
I will offer that when “Horton Hears a Who!”, the book featuring the nicest elephant on Earth, was published in 1954, Dr Seuss might have been influenced by the elephants of the Horton Plains in Sri Lanka. A paradisiac high plateau at 2,100 meters of altitude in the center of Sri Lanka, the Horton Plains were home to a large population of elephants until their extinction in the late 1940’s due to over-hunting by the British who then occupied the region.
From Big Data to Serendip (Sri Lanka), the connection is now established. Serendipitously, thanks to Yahoo! and Benchmark Capital.
"—- you don’t reach Serendib by plotting a course for it.
You have to set out in good faith for elsewhere
and lose your bearings … serendipitously.”
(John Barth, The Last Voyage of Somebody the Sailor)
Conference: the Quantified Self -
I will attend the Quantified Self conference to be held September 15 - 16, 2012 at Stanford University. Will you join me?
From health and activity monitoring (fitbit, withings, runkeeper to name a few) to collecting and using data about our own activity, the sky is the limit. I see tremendous value coming from, beyond the collection, the analysis of this data. ‘Big Data’ meets the individual. We will detect patterns, correlate data to develop better habits, increase performance, predict outcomes, or diagnose (and hopefully cure) pathologies or to alter damaging habits. We will analyze behavior, usage, and reactions based on actual data, taking the guesswork out of marketing, planning, or choosing a resource / tool. Once the data set is large enough and we have found a metastructure that allows for it, we will be able to extend these to multi-individual / group / population level analysis. The sky is the limit, this is a true revolution in the making.
Also www.palantir.com and many, many more.