(image)
“We don’t put an algorithm between you and your feed.” – Twitter exec Adam Bain, March 2013
“Please do.” Me, today
Twitter has always appealed to tinkerers, to makers, to the people who first took up blogging, who championed RSS and HTML in the early days – you know, the people who created the open web. And because of that, Twitter has always had a strong dose of egalitarianism in its DNA. Twitter expresses that DNA in a particular way: it never decides what you might see in your feed. Whenever you come to the service, you are presented with everything. It’s up to you to figure out what’s valuable.
Compare that to Google, which decides what content you see based on your search query or, more recently, your location (and tons of other data), or Facebook, whose impassive algorithms sift through a sea of friends’ updates and determine what the service, in its ineffable wisdom, decides you will see. Both of these giant companies have, at their core, the idea of editorial judgement – they decide what you see, and for the most part, you have no idea how they made that decision, or why.
Twitter makes no such distinction. And this, of course, has always been both its declared strength and its obvious Achilles heel.
For it is in making editorial judgements that the edges of a media product emerge – and to most of us, Twitter is a media product (it’s certainly an advertising product, which to my mind makes it a media product as well).
In the coming months, I expect Twitter will finally execute a major shift in its approach to our feeds, and roll out an algorithm, not unlike Facebook’s EdgeRank, which consumes the raw material of our feeds and process them into a series of media products that redefine our experience with the service. Doing so will solve for three of Twitter’s most critical business problems/opportunities: Its vexing “I don’t get Twitter” issue, its slowing user growth and engagement, and Wall Street’s ongoing uncertainty around how far the company’s current advertising model can scale (IE, whether it can grow to Facebook or Google level revenues, currently orders of magnitude larger).
Three years ago I wrote Twitter and the Ultimate Algorithm: Signal Over Noise (With Major Business Model Implications). My main argument was that Twitter has to figure out how to make my feed valuable to me – a point I’ve been talking about for years. It would take a lot of math, a lot of algorithms, and a lot of trial and error, but ultimately, I wanted Twitter to surprise and delight me each time I came back, and there’s no way a raw feed could do that. In short, I argued that it was time for Twitter to create algorithmically-driven editorial voice, one that presents me media product(s) that extract maximum value out of the feeds I followed.
It’s fair to say that three years later, Twitter hasn’t done what I wished for. Back then, Twitter wasn’t a public company, and its ad business was in its early stages. But today Twitter is a $24 billion public company with strong advertising revenues tracking at more than a billion dollars a year. So what do I know?
Well, I know that the problem still exists, and there’s no way Twitter can grow into (and beyond) its current valuation, much less compete with Facebook and Google, if it doesn’t tack into the waters of editorial judgement. This means Twitter has to stare down its existential DNA problem – it has to be willing to put itself between us and our feeds.
And I think there’s all sorts of opportunity in doing so. I think nearly everyone wants Twitter to try, and while I have no inside information, I’m pretty sure that Twitter is working hard on doing just that. Ever since the company made it clear it didn’t want developers creating consumer facing applications that built new interfaces for the consumption of tweets, the responsibility for creating that value lies squarely with Twitter.
But even as the product and engineering folks at Twitter labor to create these new interfaces, there’s no need for the company to abandon its core philosophy of showing us everything – that should be a mainstay (and differentiating) feature of the service. We just want media products on top of those feeds that mine the best stuff and present it to us in a way that keeps us engaged, provides us significant value, and thereby keeps us coming back. This of course would solve for quite a few other pesky problems – user growth and engagement chief amongst them. Oh, and it’d create the kind of media product that’s rife with signals of user intent – exactly the place where new Twitter ad products can thrive.
Earlier this year I argued that Twitter might encourage a class of “super curators,” a kind of crowd sourced approach to solving the problem, but that’s not enough. For Twitter to grow at Facebook or Google like rates, it has to build a media product that is automated, but feels uniquely “Twitter-y.” And to me, that means making something that exposes its inner workings to its users, and lets those users customize their consumption in ways that can be shared, celebrated, and even commercialized. In Who Owns The Right to Filter Your Feed?, I wrote “No one company can boil the ocean, but together an ecosystem can certainly simmer the sea.”
It’s my hope that Twitter lets its tinkerers, makers, and users help make it better and better. The company’s roots are as a user-driven service. Users came up with hashtags, retweets, and other core Twitter features. One of its most valuable assets is its open DNA – and it needn’t abandon that to create an algorithmically edited version of its main product. In fact, given all the suspicions both Facebook and Google have fostered because of their black box algorithms, a more open approach could be a great strength for any new Twitter product. Show us why your algorithm created a particular media product, and let us play around with making it better. I’d bet that plenty of folks would love to do just that. I know I would.
The way you describe it, it sounds like a binary choice between having unfiltered feeds on the one hand, and Twitter-knows all algorithmic filtering on the other.
There is a middle ground, though. That middle ground is to give us, the user, tools to algorithmically filter our feeds ourselves.
The problem with Google-like filters is that it decides what it thinks is important to you, and does not give you any ability to give feedback on or alter that decision. Say you are traveling overseas on a business trip, but are looking up something in California for when you return in a few days. The Google-like approach will see your overseas IP address and automatically give you those overseas local results first, rather than results for California. And there is no way for you to override that automatic algorithmic decision. Google always knows best.
The middle ground here is for Twitter to adopt algorithmics, but to do so with the attitude of the user, rather than Twitter, knows best.
However, given the fact that none of the other web biggies do this, that not only Google, but FB, Bing, etc all force their editorializations on the user without the ability for the user to give explicit feedback and control, doesn’t give me hope that Twitter will do it, either. Thus, it will probably either remain the firehose, or else turn into something Googlian.
I agree with the middle ground. I want to be able to do what Sara Watson suggests, but easily, and also, I want to either share the work I do – or better, have Sara share her approach, and allow me to use and edit it. Twitter should facilitate that!
From JG’s comment above: “That middle ground is to give us, the user, tools to algorithmically filter our feeds ourselves.”
We already have that to some extent. We can create lists. I don’t understand why Twitter seems to be burying that functionality. I have multiple lists for the topics I want to follow. I keep my main feed to less than 200 accounts so I can stay on top of it.
I think curating/auto-filtering will ruin Twitter.
I like the idea of lists, but it’s still too much user work. When I use them, I’m constantly having to remove and/or add people, and keeping the lists clean and focused takes a lot of effort. What would be nice would be if I could do something like seed my list with a few people of interest, and then have Twitter algorithmics automatically suggest people to add to (or remove from) the list. So that I am not always having to remember every person myself.
Additionally, it would be nice if these algorithms weren’t just black box, but instead when Twitter makes the suggestion, it doesn’t just tell you WHO it is suggesting, but WHY it is making that suggestion. Y’know, like how Netflix says that it is suggesting this movie to you because it is a romantic comedy, or because it is a period action film, or whatever.
Then, that would give you the power not only to add or reject people from your lists, but to instruct the algorithm, transparently, as to why you like this suggestion, or don’t like that suggestion.
..so at the end of the day, the user still has full control over each and every list. There is no algorithm with so much power that it actively hides certain things from you, as happens in (for example) the Facebook feed.
But there should be an algorithm that helps you manage your lists, dynamically and at will, and with full control and feedback.
I see where you are going with this, but I think the easier solution (and one that already exists) is to self-curate with lists and keep your follow numbers low. 1435 is a lot of people to follow. I’ve found that I need to keep my main feed around 500 to keep it relevant and manageable. Lists are my best friend, including one for about 50 people whose tweets I don’t want to miss.
Hi Sara! Damn spam software almost got a great comment. I don’t think your approach, which I envy but fail to do myself, scales for enough people that it drives an ecosystem of value beyond the one motivated individual.
John, I thought I was the algorithm that edited my Twitter feed and consumption thereof. Admittedly for that to work I need to keep my following relatively small and my lists up to date. So I’m almost certainly missing a good deal that I would otherwise be quite interested in. But I still have only so much bandwidth and I’d prefer to keep that under control: more quality, less quantity, or more signal, less noise.
How about Twitter offering an on/off algorithmically-generated feed?
Twitter shot itself in the foot when it turned against developers to build out the twitter eco-system. Facebook is doing all it can to foster its developer community. Since Android came on the scene, Google has become developer-focused. If Twitter re-focused on the developer community would any developer trust them? Nope.
I agree to a point – many developers are confused or stung by Twitter’s shift over the past few years. But if the company creates a strong value proposition for developers, I think they’d come back.
I disagree. It’s time for Tweetdeck to do a better job of integrating other services besides Twitter. I suppose Twitter Corporate bought Tweetdeck to prevent G+ from having access to such a great tool. Without Tools Twitter sucks. Trying to get what you want on Twitter.com is a joke. Why do it?