Keeping Up with the Algorithms: What You Need to Know About Each Platform's Sorting System
Algorithms now dominate the social media marketing discussion - almost everywhere you engage, within almost every platform you use, machine learning and data sorting is used to decide what it is you see.
The evolution of social algorithms has undoubtedly made things harder for social marketers. The development of algorithms follows much the same progression as the technical elements of SEO - in order to show searchers the most relevant responses to user queries, and weed out spammers, Google has had to constantly update its algorithm to stay one step ahead.
Now social platforms are in much the same position, which adds a more complex, technical element to the social marketing process.
So what do you need to know to ensure you’re on top of the latest social algorithm shifts, and how can you best prepare yourself, and understand the focus of each platform to maximize your performance?
Here’s an overview of the current state of each platform’s algorithm, based on what we know of their workings.
The most influential social algorithm is Facebook’s. The Social Network was the first to implement an algorithm feed, in response to the flood of spammy posts and junk which had inundated the app as its popularity rose.
Essentially, Facebook’s algorithm aims to show you the content you’re most likely to engage with - though obviously there’s a lot more to it than that.
To best understand Facebook’s algorithm, you first need to get your head around the basic engagement equation Facebook uses.
Facebook’s News Feed chief Adam Mosseri explained this in an information session at their F8 conference last year, comparing their selection process to choosing a meal for a friend at a restaurant:
“Let’s say I’m waiting for my wife – her name is Monica - for lunch at a restaurant, and she’s running late and she calls me and says ‘order for me’. And that moment, I have a problem to solve - I have to figure out what to order Monica, and we can break that problem down into steps. The first thing I need to do is check what’s on the menu – what are my options? Next, I need to consider all the of the information that I have so that I can make an informed decision about what to order – things like ‘does she like fish?’, ‘Is it lunch time or dinner time right now?’, ‘What’s good here?’. From there, I actually, in my head, make these lightweight predictions, things like ‘Would she enjoy the salmon?’ ‘Would she think it’s weird that I ordered her a chocolate soufflé for breakfast?’ And then I have to make a decision.”
This process, says Mosseri, is essentially utilizing an algorithm within your own head – Facebook’s system simply transfers these thought processes to a machine, shifting the same core elements into more technical factors – Inventory, Signals, Predictions and Scores.
Here’s the same calculations with Facebook actions replacing the restaurant options.
Facebook measures each post against these parameters, based on your individual usage patterns, which, along with a range of other measures, enable it to come up with a ranking score for each post.
That score then dictates where each will appears in your feed.
In basic terms, Facebook measures the likelihood that you’re going to take a particular action which their data suggests is the most engaging, then shows each story to you based on that probability.
Over time, however, Facebook has regularly changed the focus of the algorithm to better uncover particular measures – for example, Facebook’s most recent major News Feed update puts more emphasis on posts from friends and family, and content that generates conversation between two people.
That then shifts the algorithm equation, putting more weight onto the measures relating to comments and shares. This means that they get an increase – maybe a share goes from 0.2 to 0.4 in the weighting, a subtle shift, but one that would have a big impact.
Every single tweak Facebook makes in this regard leads to significant change, because the scale at which Facebook operates (more than 2 billon users) dictates that the platform drives a lot of referral traffic.
For users, the changes may be subtle, but for Pages, not so much.
This is how you need to consider the Facebook algorithm - the core equation relates to your relationship history with each poster (person or Page) and the likelihood you’re going to respond (based on your history and how other users have reacted). How Facebook weights that second element is relative to what they see as most relevant for boosting on platform engagement at the time.
Twitter’s algorithm is less advanced than Facebook’s, but they are putting increased focus on their machine learning efforts to boost engagement.
Users were initially outraged that Twitter would even consider implementing an algorithm, but the data has since shown that the system has helped the platform boost performance – as algorithms have on virtually every platform.
Twitter’s main aim with their algorithm is to show you more content you may be interested in – similar to Facebook, though Twitter is more about showing it to you in addition to the other tweets you might see, as opposed to ‘instead of’.
The brevity of tweets works to Twitter’s benefit in this regard – while Facebook only has so much room to show you posts each day, Twitter can show you almost every tweet, due to the real time nature of the feed.
You’ve no doubt noticed Twitter’s algorithm impacts – when you first log on each day is when it’s most prominent, highlighting tweets you may have missed, which are basically the most popular tweets from the people you follow.
But Twitter’s also been working to boost engagement by showing you tweets that people you follow have liked, or responses to tweets from those you engage with. The idea with this is to highlight potentially relevant content you may not be aware of, but may like, if you see it.
Not everyone likes this, but as noted, their efforts are working to increase engagement. It’s hard to argue with the data.
From a social media marketing perspective, the key thing to note is that Twitter’s working to show users more tweets based on engagement and interaction.
For one, that means you need to, ideally, post more engaging tweets, but even re-tweeting and liking your own content has been shown to have some impact (though minimal). That also means that engaging with people who mention your @handle or reply to your tweets can also boost exposure – if you like a reply, for example, that then has a greater chance of showing up in the feeds of your other followers.
Another important note with Twitter’s algorithm is that it’s not as sophisticated as Facebook’s, and can therefore be more easily influenced. As such, smaller actions can mean much greater exposure – it’s worth considering the various ways in which you engage with tweets, and with your Twitter followers, and how those actions can relate to potentially increased reach.
The hot social platform of the moment, Instagram too has implemented a Facebook-style algorithm – and when we say Facebook-style in this context, it’s most definitely true, considering The Social Network owns both platforms.
Like Facebook, Instagram’s algorithm is based on your likelihood to engage – Instagram engineer Thomas Dimson went over the various factors considered in a presentation last year.
As you can see, Instagram’s algorithm weights the content based on engagement factors - which include not only your search and engagement habits, but also your messaging behavior - in order to show you more relevant content matches.
How much weight, exactly, Instagram gives to each is not known, but its these elements that the system takes into account, along with overall engagement levels on each post and whatever element Instagram’s looking to optimize for to increase time spent in app.
As with both Facebook and Twitter, this means you need to focus on prompting user engagement, triggering action from your audience in order to increase engagement and generate more direct connection. That means considering on platform actions - liking comments, for example, and replying to the same.
Instagram also highlights the most relevant search matches within their Explore section, putting further emphasis on engagement.
Yes, LinkedIn also uses an algorithm, though like Twitter, it’s not as advanced as Facebook’s.
The aim of LinkedIn algorithm is the same - to boost on-platform engagement - and they take into account your historical response factors and connection strength (i.e. connections in common, workplace history, etc.) in order show you the most relevant results.
A good example of how LinkedIn’s feed algorithm works is their comment sorting tool – LinkedIn recently published an update on how comments are displayed on each post, based on engagement.
This gives you an idea of the variables they can utilize in their feed algorithm also – again, which measures they choose to weight depend on what their data shows them is the best for boosting time spent.
In my experience, LinkedIn’s algorithm needs some work – I’ll often see posts from weeks ago in my feed. But they are refining their data, and working to show you more relevant content. It’s worth considering how these factors relate to your posts, and what you can do to boost each.
And the last platform we’ll look at is Pinterest and its evolving algorithm to uncover more relevant Pin matches.
As with all platforms, Pinterest uses an algorithm to uncover more related content, and keep you searching - but their system has been upgraded from its initial matching efforts to better highlight relevant content.
For example, in its initial iteration, Pinterest’s system used board names as a proxy, but that wasn’t always effective – people might post different things to different boards for completely different reasons.
Pinterest provides this example - in the image below, there's an image of a lion couple cuddling which has been saved to boards called 'Animals' and 'Wild animals'. As such, Pinterest would use that as context, and provide related matches as shown. Even though these images aren't similar in terms of content, they are similar to the other images posted to boards alongside the first.
Due to that inconsistently, Pinterest devised a new system last year which utilizes deep learning from across the Pin network to not only examine related board content, but also what other users most commonly engage with immediately after viewing the first Pin.
As such, Pinterest’s system is now more aligned to user behaviour, showing you relative Pins based on activity, not on the actual data gleaned from the Pin itself.
That’s a much more effective process for Pinterest, which is used by people looking for similar products and recipe ideas – Pinterest is also getting smarter with its image recognition based recommendations based on usage of its Lens visual matching tool.
Essentially, Pinterest now references common user behavior to lead people to the next most likely Pin match, as opposed to the content itself – which is why it’s worth familiarizing yourself with Pinterest’s ‘Guided Search’ recommendations if you want to ensure your optimizing your exposure.
As you can see, algorithms are used in a wide range of ways across the various social platforms, though all with the same goal – to keep you on platform for longer based on usage trends. The more they can match usage habits with recommendations, the more they can increase engagement – algorithms are not an exact science, they’re an ongoing experiment, which is why it’s important to stay up to date with the latest shifts.
Follow Andrew Hutchinson on Twitter