Instagram co-founder Mike Krieger reveals methods driving social media site’s massive growth

With over 400 million users, Facebook's photo-sharing app goes from strength to strength. But what are the secrets behind Instagram's success?

Sometimes numbers are so vast they become almost meaningless. But when Instagram co-founder and CTO Mike Krieger says more than 1.2 million photographs will be uploaded onto the picture sharing site during his 20 minute talk at Web Summit in Dublin today you get a sense of how socially influential his creation has become.

To put that into further context the New York Library has taken precisely 100 years to accumulate that number of photographs and postcards for sharing with the public.

Five years in and it’s hard to believe such shock was expressed when Facebook saw fit to splurge $1bn on the start-up in 2012. Today, rightly or wrongly, its Unicorn status is assured and a $35bn price tag has been placed on it since.

Today 75% of photos are taken on camera phones and the New York Times estimates a trillion photos will be taken in 2015.

So when Krieger reveals some of the methods applied to make Instagram one of the world’s fastest growing social media sites – three-quarters of the US company’s 400 million or so users are global – it pays to listen.

How does the company make sense of its vast sea of data? And why is getting the right balance between its machine-learning algorithm and human curation so crucial? Krieger shared some of the answers:

1. Have a clear purpose – and pursue it relentlessly. Krieger says Instagram’s content has to meet these criteria: It has to be timely – not about last week, last month, last year. It has to be personal – about who you follow and what’s in your feed. And it has to be at scale.

2. Establish a way to achieve the purpose. For a company so heavily reliant on its technology, the machine learning algorithm Instagram has developed is absolutely fundamental, but without some human intervention the site wouldn’t achieve it’s purpose.

Krieger says more recently Instagram has found the right balance between pure editorial and pure machine learning. The company’s first hire was not an engineer, a designer or an ops person, he says, but a community manager called Josh. To be successful Instagram needs an engaged community.

3. Find ways to keep customer churn low. For Instagram, who you follow is the biggest determinant of whether users churn. “It’s really important you get that right from the first day,” he says. One method was to allow users to follow their Facebook friends or who they followed on Twitter. But finding ways to build suggestions for users became key.

The site anointed “Kings and Queens” – super-users they thought would drive sustained interest and engagement. But without a personalised approach to cater for the interests of each user the approach didn’t go far enough. Even small tweaks depending on the user’s country of origin weren’t sufficient. Users churned as much if they followed suggested users as if they didn’t follow anyone at all.

4. Don’t allow criticism to over-influence. Krieger admits that with each change to its suite of features the site drew criticism on Twitter from loyal users. And while it may have been valid for those individuals who desperately wanted to keep the status quo, the company focused more on the data, any drop in user churn and the uplifts in traffic. Better to make many people happy most of the time than a small number very happy all of the time.

5. Keep iterating with small tweaks. Each change to the site’s algorithm and front-end display of popular content amounted to a tweak rather than wholesale transformation. So showing photos liked by users other people were suggesting and putting blog content on the popular page saw a small 10% boost in traffic – not enough to excite Instagram’s founders. Showing the photos liked by the people you follow resulted in a 5x boost in engagement overnight. Photos liked by people of photos you’ve liked saw an increase 7x better than the baseline.

Further tweaks followed with the system constantly improving itself as it learned what works and doesn’t work. Furthermore, for users looking to explore, the site began organising images into categories and sought to help people tell stories about different events. It added a human layer to the machine learning with editors analysing the data to curate the groups for users interested in, for example, specific sports – while promoting great content users may have missed.

6. Build on your winners. For a moment-sharing site there are particular events each year that see huge spikes in traffic, such as Halloween, Valentine’s Day, New Year, and Thanks Giving in the US. Krieger explained that Instagram wants to enable people to tell and share stories better. So to show users Halloween as it was happening this year Instagram went live with promotion of recently liked video clips that share a hashtag, with the most played videos rising to the top.

Mike Krieger was speaking at Web Summit in Dublin. Websummit.net

Comments

(will not be published)