little girl at the beach on vacation
Could the secret to more human marketing interactions be technology? PHOTO: Leo Rivas-Micoud

Marketers have been swamped for years now with emerging technologies which help them better identify prospects, target and re-target them, personalize campaigns and squeeze every ounce of value out of every advertising dollar spent. 

The trend isn’t limited to the world of digital advertising: it's consistent across all owned and paid channels — the web, email, social, ecommerce, OOH and everything in between. Every web page, every email, every social post, every ad is being personalized. 

However, the rise of machine learning promises to deliver personalization at a whole new level of precision and at unprecedented scale. The question marketers need to start asking is not, what do I do with all that data — machine learning is already solving this — but rather do I have the right message(s)? Are we presenting customers the best content possible?

Machine learning is transforming industries from entertainment to security, from commerce to transportation, but the biggest near-term change machine learning will bring to marketing will be around content marketing.

Big Data, Big Content

Data scales, but content doesn’t. 

Until now.

Big brands with millions of customers have an almost unlimited need for content and no scalable way to create and manage it. Branded content is expensive to create, more expensive to customize and honestly struggles to cut through the noise. A 2016 Beckon Research report noted only 5 percent of branded content gets noticed.

Brands understand who to target — what their customers' passions, personalities and interests are — but ultimately send only a handful of generic, branded messages and offers. Machine learning offers the opportunity to sift through, analyze, sort and, crucially, understand existing content at a pace impossible for humans.

This matters. More personalized, engaging content is ultimately cheaper AND performs better. There is, after all, a lot at stake. Back in 2014, Nielsen’s Global Consumer Trust Index reported 92 percent of consumers trust earned media more than owned media.

How Machine Learning Changes Content

For most brands, compelling content is out there, and more is being created every day. Machine learning can match the thousands (or millions) of pieces of content —often produced by your own customers — that surround your brand and find other like-minded customers to share it with. 

For example, a travel company identifies a prospect researching a family trip to the beach. 

You know you've seen hundreds of thousands of happy families on Hawaii vacations recently and want to present the most beautiful images and videos of Hawaii possible. Billions of images and videos are created every day. How do you surface the best ones?

You don’t, but machines do. And they do it in three ways. 

1. Computer Vision

People post more than 85 million photos to Instagram every day, making it impossible for marketers to identify what’s in those images. Computer vision technology can scan each piece of visual content, decipher its contents and automatically apply relevant tags. 

By combining computer vision software with language, location and sentiment analysis, enormous volumes of visual content can be processed, filtered, categorized and prioritized.

For example, here’s an Instagram photo I shared on a family vacation in Hawaii.

Instagram photo of Cassidy family vacation in Hawaii

Without human involvement, computer vision technology can identify that the photo features “kids,” a “pool,” “swimming,” “sand,” “palm trees” and “deck chairs.” That it includes a 'group' of various ages with a general emotion of ‘happiness,’ that it’s 'SFW' (safe for work), that the comment is in English and that the image was posted from the Grand Hyatt resort in Kauai. 

Previously, marketers either missed huge amounts of visual content or had to manually tag and classify these photos, requiring vast resources and time.

All this metadata can be applied to the content, meaning it’s ready and waiting for marketers to inject it into any personalized web page, email, campaign or even pushed into an asset manager.

Remember that prospect planning a family beach trip? Here’s your Facebook ad.

Facebook ad mockup featuring photo of father and his two children on holiday in Hawaii

Companies report increases in conversions and engagement by implementing authentic, user-generated content into Facebook ads, with some as high as 400 percent.

2. Predictive recommendations

As your marketing team starts selecting content for publishing, machine learning algorithms can start identifying common characteristics of these content items. 

Do photos with people get published more often than those without? Are there common hashtags that your brand appears to prefer — say, #selfie or #foodporn? What about still images versus videos?

Then, once the content has been published, algorithms can prioritize the content generating the most engagement. If Instagram ads featuring families swimming get better traction than those of a beach volleyball game, then swimming photos will be recommended more often.

All of this data — collected from the interaction behavior of the marketer and the audience — teaches the algorithm to identify and eventually recommend the best content to use. This presents massive time savings for marketers who don’t have to spend tagging and categorizing content, let alone spend resources staging photo shoots or evaluating creative copy.

3. Real-time optimization

The final piece is perhaps the most important. 

Analytics and performance metrics around each content piece can be automatically funneled back into machine learning algorithms so they not only understand what types of content you prefer, but also which content converts. 

This way, if certain content pieces aren’t actually driving traffic, generating leads or resulting in sales, the technology can and will automatically recommend you replace a certain post with another one. The system then learns and repeats.

Machine learning is unlocking the treasure trove of user-generated content to automate content marketing, providing the highest-performing content with the least effort. 

Machine learning is starting to deliver on the promise of personalization at scale. And in a strangely ironic way, it’s technology that will make marketing more human than ever before.