Challenge

Pinterest sought to help advertisers increase effectiveness of ads on its platform.

To do it, Pinterest needed an objective, data-backed set of creative best practices across all video and static ad units, since what they had to date were anecdotal and subjective.

Pinterest then sought to offer advertisers the ability to score the effectiveness of their ads before they launched campaigns with a new feature that produced a “creative effectiveness score” of an advertiser’s ads before a campaign went live.

Solution

• Built a sophisticated, trainable AI model that uncovered objective, data-based factors for what makes an effective ad on Pinterest. Factors ranged from having close ups on faces in beauty ads to tips on where to place CTA buttons on video ads.

• Leveraged advanced AI techniques like large multimodal model iteration to determine ad features for performance, image detection, video parsing, NLP analysis, and more.

• Validated all creative best practices with a diverse group of creative professionals to ensure they made sense and were actionable.

• Built an MVP prototype of the “creative effectiveness score” feature to test with advertisers.

Impact

• Trained thousands of advertisers on how to make more effective campaigns.

• Ads will become more engaging and will drive greater conversion.

• The advertiser experience will improve with better tools to make effective ads.