- Posted by John Lahr
The ability to uncover subtle truths about what an audience responds to is powerful for a number of reasons. First, it takes marketing beyond vanity metrics of performance and deeper into discovering what triggers an audience to actually convert. No matter if that conversion is watching a video, completing their check-out on an ecommerce site, or attending a sponsored viewing party for the big match at a local bar.
Second, the types of subtle triggers that move an audience to conversion are often unknown to the audience themselves and therefore they are highly effective, but by nature are not obvious and nearly impossible to find through manual social media analysis. This is where Artificial Intelligence can provide a game changing competitive advantage, since it is purpose-built to find connections in massive data sets that humans can’t see. When combined with machine learning, AI can even refine its own hypothesis and continually test new ones, empowering marketers to make creative decisions based on actionable data instead of biased opinions.
Finding Signal in the Clusters
One of the most effective tools for identifying what elements of content resonate with an audience is Cluster Analysis. Cluster Analysis is when data is divided into group that are meaningful, useful, or both, and has long been an important tool in the social sciences, biology, statistics, and pattern recognition. In marketing, clustering content based on elements like objects in pictures or videos, primary colors, or post copy, and looking at that element’s impact on performance can reveal unexpected trends that greatly alter how an audience will react to your content.
We recently had an opportunity to dig into building a campaign strategy using cluster analysis when our client, Visit Utah, did a cluster analysis of their Instagram page through Cortex’s Content Grader. What they found was that posts about Skiing were the lowest performing posts on their page. With the ski season quickly approaching, and a large ad campaign already approved to promote Utah’s 14 resorts, making sure their content strategy was top notch was critical.
Case study – Visit Utah, Ski Content
Identify what content should be created in advance of the 2017 ski season to maximize performance on Instagram.
- We began by looking at the clustering analysis of Utah’s Instagram, and why their ski content ranked so low.
- Their highest ranking content were photos of red rock geological features, which performed 48% above average. With pictures like this who can be surprised?
- Big Sky photos containing mountain ranges were right around the average, performing just 3% above average.
- Snowy mountain ranges came in a bit below average at -10%
- Skiing came in dead last at 48% below average … ouch.
- We reasoned that since only 22% of the US population are skiers, it shouldn’t be surprising that only a small subset of Utah’s audience would be interested in skiing photos, while anyone can appreciate the beauty of red rocks or landscapes. This would explain the mountain shots performing better than skiing ones.
- The next step was to find what photos skiers found most engaging. To find out we looked at 40 instagram pages of the top ski resorts, tourism bureaus, and publications to get a representative sample of content and audiences. You can see some of the pages, as well as the number analyzed from each here:
With all this data loaded into Cortex, we were ready to run our cluster analysis, and what we found was fascinating. These are some of the highlights:
- The best thing you can have in a ski picture is a Fir, Pine, or general Conifer tree. This cluster performed 90% above average
- After the Fir trees, performance increases level off significantly.
- Clusters 2-8 average perform 23% – 18% above average, with the medians for each cluster approaching the average performance the lower in the range you go. Some features of these clusters:
- The less people in the picture the better. Nature shots or photos with only 1 person outperformed group shots, ranking higher in overall performance and within their clusters.
- “Party” type photos showing celebrations came in right around average, but the lowest performing of the cluster performed nearly 60% below average, so be careful with these.
- One sure-fire way to make sure your photo doesn’t perform well is to have it feature children. Kids and Families made up the 3 worst performing categories ranging from 27% – 50% below average.
This analysis proved immensely helpful to Visit Utah from a content strategy perspective. They now have a clear understanding of why their ski content does not perform as well due to its limited appeal, as well as what makes up the best content to reach a skiing audience during peak season.
When creating their own content and reposting user generated content they know to look for:
- Pine trees. The more the better. This is one of the more powerful subtle truths about what their audience likes.
- Blue skies and clean looking trails
- Landscapes that imply skiing, or only 1 skier featured.
They also know to stay AWAY from:
- Kids and families. Save them for the photo album.
- Groups, especially ones looking to party hard
- Blizarding conditions. It ruins the performance of what could otherwise be good photos.
Want to know what makes your content tick?