How do you craft information architecture when users have the power to categorize data?
There are a lot of photos on Instagram — over 40 billion, according to Instagram’s press page. So how do you organize so many pieces of data, and better yet, how do you drive user engagement by creating an information architecture that gives meaning to those photos while increasing trust?
Instagram emerges from the embers of Burbn
Instagram was founded in October of 2010, but it began its life as a much different app, one called Burbn — a Foursquare-like check-in app. Kevin Systrom, the creator of Instagram, explains:
“We actually got an entire version of Burbn done as an iPhone app, but it felt cluttered, and overrun with features. It was really difficult to decide to start from scratch, but we went out on a limb, and basically cut everything in the Burbn app except for its photo, comment, and like capabilities. What remained was Instagram.”
The original app, shown in an early demo video above, was very similar to what we see today. The core data consisted of photos, user name, date, comments, likes, photo captions and optional location data (linked through the Foursquare API).
A walkthrough of the original app shows that a new user signed up for Instagram by providing a first and last name and email address and choosing a username, password and profile photo. Next you were prompted to find friends on Instagram by connecting Facebook and Twitter. From there you could view your feed and see your friends’ photos, leave comments, and indicate likes. To add photos of your own, you could use the Instagram camera app or grab a photo from your camera roll, but only square photos were allowed at the start, keeping with the Poloroid/instamatic feel. From there you added your filter — photo editing was limited to a simple filter application in the early days. Next up — the “what?” (photo caption) and “where?”(link to a location, which Instagram implemented via the Foursquare API). And finally, sharing. Instagram linked with Twitter, Flickr, Facebook, Tumblr, and Foursquare at the beginning, allowing users to share across multiple services in one simple step.
Instagram helped users create beautiful photos from their then-subpar smartphone cameras. Photo filters were hot and Instagram delivered them beautifully. This social sharing lead people to think of Instagram as “a Twitter for your photos”.
Hashtags as an Architectural Improvement
The comparison to Twitter was soon to become further cemented when less than four months after its launch, on January 26th, 2011, Instagram added hashtags. Within the next year the number of photos exploded to 150 million and the number of users increased to over 10 million. Now all those photos were being categorized organically by the users and a whole new element of sharing and discovery was opened up. Users could click on a hashtag and see all photos with that tag and each hashtag had its own rss feed — bringing Instagram to the web. This was also the beginning of search in Instagram, since hashtags were searchable.
While the hashtag was giving structure to the data on Instagram, Systrom, its creator, was still hoping to find a new way to organize all these photos in a meaningful way. Near the end of 2012, after Instagram was purchased by Facebook, he talked about this desire for a new type of categorization, saying, ”Hashtags are a great first stab at that, but over time we’re going to come up with better ways of letting people curate experiences”.
This was only a couple of months after the release of Instagram 3.0 which included the new Photo Maps feature and geotagging. Has the geotagging feature become this “better way”? An AdWeek article from 2014 stated that “just 5 percent of Instagram posts tag a location.” While recent statistics are difficult to find, a quick survey of 20 random Instagram accounts revealed that only 3% of photos were tagged with a location.
So what do you do when attempts to impose your own categorization on the data fails? Instagram’s engineering blog notes that “in 2013…there was no engineering team dedicated to data infrastructure or products that relied on data processing like ranking and machine learning.” After putting together a team, they started to focus on search, Explore, account suggestions and analytics. As a result, in June of 2015 they released the new Search and Explore which focused on real-time trends.
The problem of creating meaning from so much data has much to do with the sheer number of photos. Instagram states on their engineering blog:
“Building a system that can parse over 70m new photos each day from over 200m people was a challenge.”
One thing is certain — it’s the hashtags that drive the system. The details on how the instagram data team determined trends can get a bit complicated (Sd(h, t) = SM(h) * (½)^((t — tmax)/half-life)) but the equation relies heavily on hashtag data.
At the same time, another group of engineers was working to determine how to find meaning in the emoji hashtag which was to be released in April of 2015. Driven by user activity, Instagram said:
“digital language has evolved such that nearly half of comments and captions on Instagram contain emoji characters. And earlier this week, Instagram also added support for emoji characters in hashtags, which allows people to tag and search content with their favorite emoji #🎉.”
All of these changes to the organization and categorization of photos lead to release of the “top post” navigation added to the hashtag feeds. This was a huge change which promoted photos, taking them out of the normal, chronological feed and making discovery of new accounts easier.
Hashtags are here to stay for the foreseeable future and Instagram has built their architecture around them. While there are challenges to this approach, Instagram is clearly keeping up with the trends, announcing last September that they have hit 400 million users. And according to a report from Credit Suisse, “Instagram is turning out to be a revenue-making machine, just like its parent company, Facebook Inc. The photo-sharing app is expected to bring in revenue of $3.2 billion in 2016.” Those hashtags seem to be working out for them so far.
Back in 2011, Instagram, like many apps, released an API to help developers use the data and innovate based on their platform. While there were some exciting projects based on the API, the beautiful and simple interface of Instagram kept people using the app, rather than turning to third-party solutions.
While there was some excitement around a few projects which used the Instagram API, the lack of trust it instilled in users was too great and Instagram announced changes that would limit the usage of the API during this past year stating that this would be done “to improve people’s control over their content”. This seems understandable, especially after a malicious 3rd party app was found stealing user passwords.
In fact, with this change Instagram reiterates a statement they made back in 2012:
“I want to be really clear: Instagram has no intention of selling your photos, and we never did. We don’t own your photos — you do.”
Fostering trust is a constant struggle in an environment where the users are allowed to set the categories and then use those as they see fit. There will always be a percentage of users gaming the system, applying hashtags that will tap into the algorithms that identify trends to get their photos the most views.
In the future, unless there is an innovation that takes us away from using hashtags and into a new method of categorization, Instagram will have to innovate to stay ahead in monitoring hashtag usage. Currently Instagram appears to be struggling with the monitoring of the 80 million photos that are being uploaded every day. Surely they can’t have eyes on every one of these photos.
In March of this year, Instagram stated that they would be making changes to user feeds saying:
“You may be surprised to learn that people miss on average 70 percent of their feeds. As Instagram has grown, it’s become harder to keep up with all the photos and videos people share.”
The concept of taking the feed, which has been chronological from the beginning and introducing an algorithm that would change the order of photos was met with protests from the community.
Another question that Instagram will have to tackle in the future is how to deal with hijacked hashtags. Recently there have been many news stories talking about how Instagram has been banning certain tags. “The social network typically has both a “top posts” and “most recent” section for hashtags. But if a hashtag is being abused, Instagram will remove that “most recent” section and only display top posts”. Clearly, censorship is no way to instill trust, so the challenge is how to make sure you don’t get porn or spam when you click on a hashtag (such as #kansas, which, believe it or not, is one of the censored tags) while keeping Instagram free and open.
The Future of Instagram
So where can Instagram go from here? With over 80 million photos being uploaded daily, they have a huge data set to analyze. While much of their time has been spent finding trends in hashtags, It seems that little analysis has been done on the photos themselves.
With innovations in machine learning, scientists are training systems to determine what make a photo memorable or beautiful. With Instagram’s vast stockpile of photos and their corresponding like button tallies, they may be able to advance this innovation even further. Just imagine if, like facebook which changed their like button into “reactions”, they extended their like button to include a “beautiful” indicator. They could then take the most beautiful-ed photos and analyze them for common traits.
Using this knowledge, Instagram may be able to help you choose which photos you should upload, or make suggestions on how to improve photos to make them more attractive or attention-grabbing.
Perhaps one day we’ll open up Instagram and see a feed of photos individually tailored to our own idea of beauty. And isn’t that why we visit Instagram — to be inspired by beautiful photos?