[Thank you Dr. Di Cook for featuring this blog post among the Project:Data Stories on your ETC 1010: Data Modelling and Computing webpage].
I am voluntarily helping a friend, Carl Lanore, with Facebook advertising for his business, Super Human Radio Network – featuring the health / physical culture-focused podcast, Super Human Radio. This is a provocative podcast with content I do not always find agreeable but often find intriguing and inspiring with respect to exercise and lab testing analysis.
Here are data visualizations created with my in-development but functional FBadstats R package for the month of August.
With a budget of only $200 / month, I have aggressively experimented with a variety of ads (creatives, audiences, and objectives) in pursuit of low-cost desirable results. My Facebook Ads work with the business owned by my wife and me, Impackt Squared Solutions, LLC [now focused on her online corporate career advancement training course, Six Figure Spirit – Power Moves], led to my contrarian belief that for most objectives (e.g., clicks, post reactions and shares, page likes), Facebook ads will generally indicate long-term quality of performance in a day, and often within a few hours. For Super Human Radio though, I have often let ads run for at least 3 days (the most frequent minimum duration I have seen suggested) at $3 / day.
Ads discussed in this post were generally given about 12 hours before deciding on their continuation.
When Carl posted a Facebook live video (“Dog Food Wars: Should Your Dog Be Vegan?” that quickly earned a high number of comments, reactions, and shares, I boosted the post as soon as I could. Doing so from within the main Facebook system instead of Ads Manager creates a mess in Ads Manager because every combination of boosted post and audience creates a separate campaign. However, I did use my mobile phone because this is the fastest technique for boosting (advertising) a post and I was not able to get to my computer. I help with Facebook advertising on a volunteer basis, for now, and work fulltime as a data scientist.
The “Like Page” audience (Super Human Radio Network has about 8000 Facebook likes) easily shined the brightest, scoring post reactions, comments and shares whereas the other audiences (Like Page + Friends and then a mix of 1% Lookalike Audiences subset to those into CrossFit or Bodybuilding) scored zero after a few dollars had been spent. The latter audiences have sometimes been good performers, with the “Like Page + Friends” being a surprise given its massive 3 million-plus person size.
So I focused on the “Like Page” audience (Ad Set “SHR_DogFood_Vid_SHRPageLike_RP”), made its campaign name “SHR_DogFood_Vid_Multiple_RP”, and quickly duplicated the Ad Set in Ads Manager so I could experiment with more audiences and compare their performance against my best performer. As you can see, the amount spent ($20 for the “Like Page” audience”) represents an unsustainable amount given the $200 / month budget. Carl approved the experiment. My justification: Less was spent last month and this video represented something special. It also was an opportunity to try something different and communicate to him a strategy for the future.
Carl has expressed a desire for more shares and the Dog Food video delivered in that regard while also engaging consumers with Carl’s on-screen presence and increasing our video-viewing warm advertising audiences (e.g., “Watched 10+ Seconds of one of your videos”). More generally, our primary objective is to increase listens to each episode and I wanted to explore focusing Facebook ads spending on developing true fans versus paying for, perhaps, a single article’s or episode’s consumption. A challenge exists in tracking multiple listens because of Facebook pixel limitations. So we have to work based more on our intuition than otherwise.
Among the audiences I tried, the other standout is at the top of the screenshot: SHR_DogFood_Vid_SHRLkDog_RP (1% Lookalike Audience for Page Likes = the 1% of the Facebook audience in the U.S. that most resemble those who already like the Super Human Radio Facebook page, subset to those interested in dogs) . This outperformed a couple warm audiences (e.g., had viewed at least 25% of one of the page’s videos) as well as Lookalike Audiences resembling those who had viewed a significant proportion of one of the page videos. Strong performance for the Page Like Lookalike represents an important accomplishment. It suggests the page now has enough Facebook likes for us to analyze that audience and derive insights.
[EDIT:I ultimately decided to continue running both the aforementioned audience (relevance score = 8) and the 50% video-viewed Lookalike audience (relevance score = 9) for another day at $10/day because I liked the number of people in the latter who were watching a significant proportion of the video. These are also individuals likely to become true fans and enhance our video-viewing warm audience.]
Focusing on delivering page likes, we ran an ad that earned a decent cost per page like of $0.82 when using a 1% Lookalike Audience (those who had engaged with the Facebook page versus the aforementioned liked) subset to an age group and some DMAs (Designated Market Areas) that had performed well with respect to page likes. I identified these DMAs by using my free (open-source) FBadstats R package, which aggregates ads data found in Ads Manager-exported CSV files and then reports the best and worst breakdown groups.
Super Human Radio is a health-focused show but I have not observed those who like the page to stop liking the page if they won over by this manhood-focused ad, which features proud father Carl’s own words. I assembled the slideshow using images found on his page.
I tested the same ad with a similar audience except one was for a 1% Lookalike for those who like the page and one was for those who had viewed at least 50% of a page video. You can see at the top that the Like Page lookalike won by a large margin this weekend. This provided further evidence that we have now reached a point where first experiments should target those who like the page or a Looklike for those who like the page.
I want to backup a bit and talk about a page like ad I tried a couple weeks ago. Carl developed a 30-second video I call “DriveTime” that I tested on a cold audience to generate page likes. It did not perform well, only earning one page like with $6 spent. The audience was in the same high-performing page like DMAs referenced above, numbered 530k people and had performed well for a variety of ads. The audience qualities included: under the Lookalike Audience (1%) for those who had engaged with the Super Human Radio Facebook page – age 25-54; Interests: Military, CrossFit, Police, Good Men Project, Physical Culture.
We only spent $10 between two audiences before I concluded on too poor of performance to continue. We do not have the budget to hope for poor performers to improve as the Facebook algorithm refines ad delivery.