The Importance of Post-Campaign Creative Analysis
June 22, 2017
In 2016, creative production constituted 24% of marketing budgets. This year, content creation costs are slated to make up more than 34%, and marketers are not optimistic about these costs decreasing any time soon. As consumers’ attention becomes fragmented and distributed among numerous digital channels including Facebook, Instagram, Pinterest, Snapchat, Twitter, YouTube and LinkedIn, creative becomes more important and more challenging than ever before. But the cost of producing creatives to reach people on each of these platforms shouldn’t increase every year—it should be more scientific, more deliberate and extremely data-driven. One way to achieve this is by implementing post-campaign creative analyses to uncover insights and inform content production in the future.
Creative pre-testing is a great strategy when you’re analyzing a small number of creative. Brands can run media at low levels of spend to test creative variations against prospecting audiences and identify top-performers. This allows you to strategically scale spend against top-performing variations of your creative. It also, as the name implies, takes place before you run your campaign.
But brands have a wealth of data from campaigns run in the past. Mining this data allows you to answer questions and extract insights, without having to travel back in time to initiate a controlled test. SocialCode recently tagged and categorized hundreds of Facebook video ads based on style and length for a CPG brand, in order to draw meaningful insights. This analysis showed that videos built specifically for social, which delivered impact in a matter of seconds, definitively drove the best results. The brand saw increases in ad recall, awareness and purchase intent far above CPG norms, suggesting that social-specific video content resonates with audiences despite its shorter length. These learnings can be applied to future campaigns across the brand’s portfolio.
Additionally, machine learning taxonomy allows you to programmatically label images in a consistent and hierarchical manner to determine which elements of a creative (e.g., tone, composition, etc.) lead to stronger performance, without conducting these analyses manually. Machine learning, using image labeling tools such as Cloudsight, Google Vision and TensorFlow, involves creating an image labeling framework to identify objects, logos, and text in creatives.
For example, the CloudSight API recognizes, captions, and classifies the details of an image within seconds. It can differentiate between celebrities and tag images accordingly (ad featuring Matt Damon versus ad featuring Brad Pitt). Google’s Cloud Vision API quickly classifies images into thousands of categories (e.g., “sailboat”, “lion”, “Eiffel Tower”), detects individual objects and faces within images, and finds and reads printed words contained within images.
Using TensorFlow’s deep-learning framework, SocialCode was able to divide hundreds of creative images from a pet food brand’s Facebook campaigns into two categories: cats and dogs. The image labeling technology could recognize a Labrador Retriever in an image and classify it as a dog, or a Siamese and categorize it as a cat. This resulted in 549 dog images, and 392 cat images, each with corresponding performance data. Upon analyzing the bulk performance of both categories, SocialCode was able to report back to the client that its Facebook ads that featured dogs drove a higher click-through rate than its cat ads.
Click here to download the full data-driven creative guide and learn how to effectively use data to cut down your content production costs and efficiently utilize your non-working media dollars to drive better outcomes.