Do we control the algorithm or does it control us?
Instagram’s algorithm: why do I see what I see on Instagram?
Algorithms have become as much a part of our lives as our daily Gin O’Clock and much like Gin O’Clock, their importance is often overlooked.
They help us to choose which Netflix show to binge-watch next and keep showing us that one pair of shoes that we are desperately trying not to splurge on. If you want to know about the history of algorithms or how they work, you can get yourself up to speed here.
Much to the annoyance but understanding of all of us here at One Fifty, social networks are notoriously vague about the specifics of their algorithms.
We decided to put Instagram to the test.
The Experiment
We created three Instagram accounts, all made on the same device, with all other factors the same, apart from their gender. We created a female, male and control (as we were measuring gender as a variable), let’s call them Michelle, Barack and Bo.
For seven days, we completed the same 5 actions across all three accounts:
We followed the first 10 recommended females on Michelle’s, the first 10 males on Barack’s, and the top 10 local (London) accounts on Bo’s.
In each case, we liked the first 10 ‘female/male/local’ posts that appeared on the feed.
We recorded the top 10 feed posts for each account.
We screenshotted the explore page.
We recorded the first 10 recommended accounts, no matter what they were.
This gave us lots of lovely data to analyse, and plenty of food for thought.
So, what did we find? I hear you ask.
At the end of the 7 day period, we had seen something really interesting happen. The algorithm, (you can call him Al), was relatively more sensitive to certain accounts that we followed, appearing to place a heavier importance weighting on these accounts. In other words, accounts that are recommended to us or the content that we see on our feeds, is often not entirely representative of our actions.
A little confusing, I know. Let’s break it down.
Even though we only followed accounts that were recommended to us, it was clear that Al was more sensitive when we followed certain accounts, over others. For example, for both Barack and Bo, the recommended list quickly became saturated with specific topic categories of accounts, such as Korean pop-stars and LGBT+ community accounts.
What we saw on day 5…
This was surprising, as despite these genres making up an extremely small proportion of the accounts Barack and Bo followed, they seemed to influence their entire recommended lists for the duration of the experiment.
Does this mean that these accounts could have been added in by Al to test the user’s interests? We think so. Al is likely throwing these recommendations into the ring as a way to test the waters. If these ‘test’ accounts are followed, they become a strong indicator that the user is interested in that genre as a whole, and that they should be shown more of this type of content.
This is a key reason that individual posts, rather than accounts, tend to go viral on TikTok. It’s all about falling into the #fyp trend of the moment, a theme that users are being shown due to them liking other similar ‘test’ content. For brands aiming to get onto people’s feeds, these themes are important to think about.
The experiment continues….
Unsurprisingly, as we women like to take our time, we did not instantly see the same results with Michelle’s account. We decided to continue the experiment for a few more days to see if a similar pattern occurred.
The answer is, kind of.
Whilst we didn’t see the same really obvious pattern coming through, the recommended accounts began to have extremely low follower counts, and all seemed to be followed by one account (@juliteworthy), who we had followed a few days earlier. It seems similarly to the ‘test’ genres, Al assumed that Michelle was in/or interested in Julite’s squad and then proceeded to recommend the other squad members to her.
This is something that you may have noticed with B2B accounts. If you were to follow Alex, Al may assume that you work at One Fifty, and then recommend you to follow me (a recommendation I can get behind.)
I know what you’re thinking, what does this mean for me/my business?
Find your Korean pop star (Aka, your niche). Engage with other accounts in your niche. This will likely result in Al steering users that engage with similar accounts, to yours. Be something to someone, not nothing to everyone.
The bottom line…
Knowing how algorithms work can really help you to love them, and them (Al) to love you back.
