Let’s start with ‘What is an algorithm?’
Put plainly, based on a series of parameters, such as previous interactions with a platform or personal preferences, an algorithm can help to create faster, smarter and more precise experiences by deciding which content to show users or in what order.
If you’re someone that relies on your social media exposure, you can quite often think a platform is out to get you due to their algorithms making it more challenging for your content to be seen.
The aim of these companies is usually to bring people closer to the content that they deem relevant and enhancing to a user’s experience; all based on the data they have available to them.
The power of precision
Firstly, how do you define what’s relevant and valuable to your users? Most algorithms come with their pros but almost all come with potential cons that need to be taken into consideration; you can start by comparing the outcome against your product’s value proposition.
Let’s start by looking at Twitter. Value proposition: ‘See what’s happening in the world right now’. On average, 500 million tweets are posted every day. In order to provide users with the best experience, an algorithm is in place to make timeline content relevant to each user. This isn’t about hiding content but instead about bringing the most desired content to the forefront; moving away from the expected chronological ordering system. This is composed based on your previous behaviour. By interacting with the same users on a regular basis, Twitter can assume you’re interested in what they have to say and to then assume you’d likely be interested in the tweets and accounts they engage with would also be pretty reasonable.
In theory this is a great (and relatively obvious) way to expand reach from user to user; building an extremely personal network. However it does mean that you could very easily only ever interact with people who have the same views as you and never expand the conversation (#brexit). Therefore the algorithm potentially prevents users from building diverse networks and is now debatable whether Twitter does in fact provide a realistic view of ‘what’s happening in the world right now’. Aware of how their algorithms can make such a big impact on the service, Twitter continuously tests and iterates to try and find the best outcome.
“Our algorithm changes on an almost daily to weekly basis”
– Deepak Rao, Product Manager of the Twitter timeline.
Using algorithms to create new features
Algorithms don’t just have to be about enhancing timelines. If the data is available to you, they can also be used to create new features.
A great example of this is Spotify’s ‘Discover Weekly’ playlist. Potentially the most personal insight into one’s soul. 😭
For those that don’t know, Spotify’s Discover Weekly is a personalised playlist with 30 songs they think you’ll like based on your current listening behaviour. This is a surprisingly advanced algorithm, from the obvious genre tags to analysing the raw audio tracks themselves to find comparisons between songs.
As well as this, Spotify also offers concert suggestions; another algorithm built feature based completely on a user’s sole behaviour.
The power of persuasion (or manipulation…)
As well as content ordering and creating new features, you can also use algorithms and the data available to influence how you recommend content.
If you know enough about a user to recommend content, you may also be able to accurately influence or persuade their decisions when making the suggestion; helping the content appear more meaningful and thought about.
Anyone who uses Netflix is aware that their recommendations are based on previously watched titles. However a lot more goes into the curation of these recommended lists than you might realise.
Instead of creating one artwork per title, Netflix attempts to cater to users’ diversity in preferences by identifying aspects of a title that are specifically relevant to that user’s demonstrated tastes; encouraging different types of people to watch the same thing in an admissible way.
“Someone who has watched many romantic movies may be interested in Good Will Hunting if we show the artwork containing Matt Damon and Minnie Driver, whereas, a member who has watched many comedies might be drawn to the movie if we use the artwork containing Robin Williams, a well-known comedian.”
– Ashok Chandrashekar, Machine learning at Netflix
The Outcome: Netflix increases engagement with their platform with options becoming seemingly more accessible to their users — making the overall experience provided a more meaningful one.
Identifying valuable data
If you’re struggling to find ways to take your product or service to the next level and feel your offering needs some enhancement to do so, identifying the relevant and valuable data you have available to build algorithms with is a good place to start. By digging down and really dissecting this information, you can spark a world of ideation that hopefully spurs your next great ‘thing’ into fruition.
Take your current product solution > Identify every aspect of data you have surrounding your users (or ‘will have’ if the product isn’t yet live) > Join the dots and map ways in which this data can be used to add value to the end-to-end experience; whilst also building on or echoing the value proposition.