Twitter’s Timeline Could Get (More) Algorithmic

Twitter has changed significantly over the past 8 years. With every new product manager and leadership cadre there have been attempts to make it more accessible to new users and more sticky for existing ones. But the biggest changes may yet be ahead as Twitter considers altering the core of its product: the timeline.

Update: Twitter CFO Anthony Noto made some comments at a conference this week confirming what we’d heard, that a more algorithmic Twitter timeline is coming.

If you’ve been watching the company for a while — or have been a regular user of the product — then you know Twitter is always experimenting. So much so that It’s become almost a running joke amongst the reporters that cover changes to Twitter. If there is a new thing someone is seeing then it’s most likely a Twitter experiment.

Twitter VP of Engineering finally just wrote a blog post about it that Twitter PR points everyone to when they inquire about new experiments. “It’s rare for a day to go by when we’re not releasing at least one experiment,” he wrote.

I’ve written before about the way that Twitter was collating and interpreting the data from these ongoing experiments, but many of them have been centered around user growth or ‘engagement’ — getting people to interact with tweets and content.

Even now, Twitter is conducting a host of experiments in dozens of ‘buckets’ of thousands of users. Recently, faves of people who users follow began showing up in feeds of activity — effectively curtailing the use of the fav as ‘shorthand’ and placing close to par with the ‘retweet’. And Twitter has already experimented with injecting other content into your timeline, as we reported recently.

Not every ‘fav’ translates into a tweet being shoved into your timeline, instead an algorithm is used to determine the ‘velocity and momentum of interest’ in the tweet. If they’re significant in your network, then Twitter will surface them for you, even if they would not normally have appeared in your feed.

If this sounds familiar, then you may be thinking of the Magic Recs experiment, which turned into a built-in feature that sends you push notifications when a tweet, account or hashtag has gotten a lot of play in your network. But, we’re told, the team behind the fav experiment is separate from Magic Recs — though both have the common thread of using an algorithm (an equation for interest, so to speak) to determine which tweets to surface.

But we’ve been informed by sources that Twitter isn’t done experimenting yet. In fact, a new effort internally will be touching many aspects of Twitter’s consumer products — including the timeline.

Specifically, a more ‘algorithmic’ timeline that chooses things to share with you based on your interests and interactions and those of your network.

Until recently, the Twitter timeline was an inviolate representation of who you chose to follow — and what they chose to share — in reverse chronological order. Then, ads and promoted tweets started getting more prominence, and the ‘reply’ mechanic started mucking around with the chronology of the timeline, bumping up old tweets when they were responded to.

Now, Twitter is set to go further, with a more ‘algorithmic’ approach to users’ timelines

Personally, the first thing I check when I hit Twitter is not my main timeline or the Discover tab but my Notifications tab. This is where the ‘life’ on Twitter is for me. People replying to me, favs being thrown back and forth, people mentioning me when they think there’s something I’d be interested in.

I’d personally be just fine if the main timeline was melded more with things like this, as they make Twitter feel more engaging. I’d be willing to bet that it’s hard for most new Twitter users to get this kind of feeling and experience from the platform, especially with how difficult it has been for them to nail the on-boarding experience.

Years after I first wrote about the poor initial impression Twitter gave to new users, the company just rolled out a new on-boarding process that improves on some things but still misses the mark in others. It still focuses, for instance, on lists of people who it feels you should follow, enforcing a minimum of follows if at all possible.

“Forcing follows like this seems like the wrong approach, since enjoying Twitter is about curating your timeline,” says Owen Williams in a recent piece about Twitter’s new on-boarding method. I agree that forcing follows is a bad idea, but I don’t think that forcing ‘curation’ right from the get go is wise either. Another solution would be to ask about interests and present users with an algorithmically driven timeline that allowed them to get up and running right away — curating their own follows and culling content they didn’t like over time.

Sources inside and outside the company I’ve spoken to emphasized to me repeatedly that the biggest challenge Twitter faces is how to show the ‘good stuff’ of Twitter to people who haven’t built a decent timeline.

As any long-time twitterer knows, the service is very much ‘what you make of it’ and poor choices in initial followers (or difficulty understanding how to make those choices) likely kill off many potential users before they get hooked.

In the on-boarding process, Twitter could use signals (who you picked to follow initially, who your friends are via contacts etc) to generate an automatic timeline of content that showed you stuff that wasn’t completely dependent on your ‘good taste’ in follows.

As one person put it to me, Twitter could see that you followed a particular player from, say, the Yankees — but not one who used Twitter particularly well — and it could show you other, more prolific or ‘engaging’ Yankees players in your timeline without you having to follow them or other people in your feed having to explicitly share them via retweets or favorites.

The key, of course, is all about timeline relevancy.

Twitter’s power comes from the fact that it provides and up-to-the-second snapshot of the world shared over the web. Facebook’s algorithm focuses on the most read (engaged) or most likely to be read content.

Twitter’s should — and I say this hopefully, not factually — focus on delivering content that’s ‘right’ for the right user at the right time. That’s both in content and, more importantly, temporally.

Thresholds for how ‘old’ the packages of content we formerly called tweets can be before they’re no longer ‘relevant’ will have to be tweaked and tuned carefully.

Before you get up in arms about this kind of algorithmic content shifting your timeline around too much remember that retweets, one of the earliest additional functions of Twitter, already offer a mechanism for presenting older content as ‘fresh’ in your feed. The difference being that one is initiated by another human you’ve chosen to follow and the other by a server in a room somewhere.

If Twitter is able to honor the same kind of relevancy in older content that RT’s do (humans believing they’re still current) then its algorithmic timeline could actually make for a more compelling experience. Selfishly, as a Twitter ‘power user’ of sorts, I would love a toggle that gave me the ‘plain Jane’ timeline as it exists now — but that adds complexity so who knows.

This new effort is said to be headed by Twitter’s VP of Product Daniel Graf, a former Googler in the Maps department. I’ve spoken to people inside and outside the company who have had both complimentary and doubtful things to say about Graf’s new approach and his understanding of Twitter. But that’s not out of the ordinary with any incoming product head. Graf was hired earlier this year to replace long-time chief Michael Sippey. A recent Businessweek article from Sarah Frier and Brad Stone reported that ‘user interface’ changes were in the works, driven by Graf.

I’m sure that some will react negatively to the concept of an algorithmic timeline — especially when you consider that the approach will likely be used to show you more ‘promoted’ content as well as organic content. But I think that it’s an important step in making a Twitter that can attract and retain a billion or more users at some point in time.

If you consider the ‘design for the 80%’ philosophy that many companies apply to making a mass appeal product, then Twitter’s current core concepts are in even more need of a rethinking. With billions of potential users out there and only ~280M of them on board so far — who do you design for?

Recent events like Israel and Gaza, Ferguson and even the World Cup have demonstrated to me that there is a real, powerful benefit to a Twitter that can show media and text in conjunction with real-time reporting and news — and that is designed that way from the ground up. The challenge would be to not give up the unique qualities that made Twitter the place to be for those events, not Facebook — mostly the real-time and seemingly ‘unfiltered’ nature of the service.

Will I miss the simple, straightforward timeline? Sure. But am I convinced that this will ‘ruin’ Twitter? No.

Could it? Absolutely. The timeline as a concept was so powerful, so concise, that it powered the first 7 years of Twitter without any real structural changes. That’s extremely rare in a product, especially in today’s pivot-happy startup environment. But it also makes it that much more difficult for Twitter to make large changes without upsetting core constituents.

Selfishly, I don’t want Twitter to change too much, and if it does I want it to emphasize personal conversations and content over something that’s interesting to the rest of the world but not me. But it’s an incredibly tough challenge and there are no easy answers to what’s best for the ‘next billion users’.

Twitter did not respond to inquiries about upcoming changes to its product.