Snapchat is using machine learning to work out who is most likely to swipe up on an ad

Snapchat has begun using a new machine learning technique to determine which of its users are most likely to swipe up and engage with an ad on the app, AdAge reported.

This month, Snapchat began rolling out what it is calling "goal-based bidding," the company confirmed

The targeting technique allows advertisers to tell Snapchat whether they value ad engagement - such as swiping through to install an app, visit a website, or watch a movie trailer - versus simply reaching the maximum number of users.

While advertisers using goal-based bidding will continue to be charged in the same way - on a CPM (cost per thousand impressions) basis - Snapchat will optimize ads to be delivered to the portion of their target audience that has the highest propensity to swipe on ads.

The new targeting option is currently available through the Snapchat API (application programming interface) it launched earlier this year, which allows advertisers to buy Snapchat ads through the third-party ad tech vendors they use to buy ads in other apps and elsewhere on the web.

Snapchat says around 20% of its advertisers are already using goal-based bidding and that it has helped them achieve up to 40% efficiencies in cost-per-swipe and increased ad view time (although it didn't specify by how much).

While goal-based bidding is an iterative improvement to advertising optimization on Snapchat, it marks another effort the company is making to bring its targeting options more in line with other digital advertising platforms.

Currently the only goal advertisers can optimize towards is swipes, but Snapchat says it's likely to add more goals in the future, which could include app downloads or ecommerce sales.

As one unnamed agency executive said to AdAge, this is what advertisers are really craving from any advertising medium: The ability to optimize their advertising toward the likelihood to achieve actual business outcomes.

Analysts at eMarketer predict Snapchat will generate $935 million in revenue in 2017, having taken early marketer criticism about its lack of measurement and targeting options on board and moving to roll out new ad formats, targeting, analytics partnerships, its API, and expanding its global sales team.