Detailed Overview:
Skills: Advertising management, Artificial Intelligence, Marketing, Strategy
Tools used: Python, OpenAI CLIP, IBM Watson NLP
Collaborators: Rohitha Yegnisettipalli, Bharath Simha, Priyanka Reddy Bathula, Tarunya Daga
Link for code: Github
Introduction:
Apparel companies often use controversy to promote their brand. High fashion integrates controversial elements into their design aesthetics and marketing choices to 'make a statement' and 'stand out', while sports apparel companies are known to leverage controversial themes to add a social element to their campaigns. In this project I analyze Nike's successful but controversial campaign featuring Colin Kapernick using Computer Vision and Text Sentiment analysis and compare the results with controversial but unsuccessful initiatives by Balenciaga and Sabyasachi to determine a 'smart' way to use controversy for marketing.
Problem Statement:
Brands which leverage controversial marketing aren't able to use it smartly in their Ad campaigns
My Solution:
A combination of Computer Vision and Text Sentiment analysis to measure the controversial 'aspect' and smartly optimize content validation to get positive outcomes when using controversial marketing.
Project Goal:
I aim to use these results to create a Business Intelligence tool that can be used by marketers to profile vendors, validate content and utilize Computer Vision and Sentiment Analysis to optimize their decisions and successfully navigate controversial marketing.
In November of 2022, Spanish fashion house Balenciaga dropped an ad campaign that featured children holding Teddy Bear plushes. This would have been another one of their infamous and viral marketing campaigns except this time, the Teddy Bear was clad in a BDSM costume which, due to its extremely inappropriate nature, made the ad campaign more controversial than just 'infamous'. Balenciaga should have known better than to have an inappropriate pairing of themes in its Ad. Sure, the company was known for regularly utilizing controversial subject matter, especially with its political stance on war and themes of mass displacement and emigration, to inspire and promote its latest collection of clothes but to go so far as to bluntly feature child abuse is what came as 'shocking' to many a customer. What ensued was a lot of public back and forth between the company, its ad vendors, production designer, and celebrities whom they had endorsed. Balenciaga put up statements about their mistake and briefly tried to shift the blame onto the production designer, but in the end, quietened to just apologizing. Many celebrities who had previously supported the brand were quick to cut all ties with it. This made me wonder, Fashion and apparel were always known for controversial marketing campaigns, especially nowadays when it is harder than ever to grab people's attention. But have companies been able to smartly leverage this tool historically, and can we analyze the impact of said controversial campaigns on these brands using AI to identify successful campaigns and what made them succeed? Can these insights be used by marketers who want to leverage controversial marketing in the future?
My Product:
Content Validation and Vendor Profiling Tool
Self or Vendor's Images
Self or Vendor's Text
Optimized content for publishing
Visual representation of my solution. The diagram to the left represents image profiling using OpenAI's CLIP to validate content (which can also be applied to profile vendors). The diagram in the center shows IBM's Watson for text sentiment analysis and the final image on the right shows the result of the application of both of these tools to use the 'right' marketing content. One thing to note is that the role of a 'human' is required at every stage of the process with more emphasis on analytically informed decision-making in the final stage when most managerial marketing decisions are expected to be made.
My Approach
To study the use case of my tool, I have conducted two experiments, each of which will analyze Images and Text from campaigns that have used controversy. The response from the public post these campaigns will be analyzed using a sentiment analysis of the news headlines of the top Google search results.
Experiment 1 (Computer Vision Analysis):
Methodology:
Open AI's CLIP uses Zero Shot Image Classification to connect text to images. For my first experiment, I will run an identical set of words on two images from Nike and Balenciaga's controversial Ad Campaigns and analyze the scores from the two outputs.
Words run through CLIP:
Tools used: Python, Excel and Figma
Image Profiling: Nike
48%
INSPIRING
39%
EYE-OPENING
5%
SERIOUS
4%
SAD
Analysis powered by
Inference:
Looking at the output from CLIP, we see that from the words provided, the words INSPIRING and EYE-OPENING have the highest percentages of 48 and 39 percent respectively.
Point to Note:
Out of the set of words, the words with the majority score are words with positive reinforcement. This shows that the controversy being leveraged is left to the context behind the poster but isn't being depicted in the poster itself.
Image Profiling: Balenciaga
Analysis powered by
CHILD
71%
BDSM
9%
SERIOUS
7%
BONDAGE
3%
Inference:
Looking at the output from CLIP, we see that from the words provided, the words CHILD has the majority of the match percentage with a 71 percent match, and BDSM, SERIOUS, and BONDAGE make up the rest of the match.
Point to Note:
Here we should note that CHILD showing a high percentage becomes a problem when the same image also gives words like BDSM and BONDAGE as matches even with low percentage matches.
Experiment 2 (Text Sentiment Analysis):
Methodology:
IBM Watson's Text Sentiment Analysis is used to analyze the text in Sabyasachi's Instagram post about women's day and Nike's Colin Kaepernick Ad. Scores from the analysis are used to compare the two campaigns
Tools used: IBM Watson, Excel and Figma
“If you see a woman overdressed, caked with makeup, armoured with jewellery, it is most likely that she is wounded. Bleeding inside, silently. Heal her with your empathy because nothing can replace human warmth . Not even the most precious of jewellery.”
Sentiment analysis of text used by Sabyasachi for Women's Day:
jewellery
woman
makeup
empathy
human warmth
Sentiment scores for each of the words:
Inference:
Studying the Sentiment Analysis results we notice that a few key words which are closely associated with the Sabyasachi brand like woman, jewelry, and makeup all have a negative sentiment. This bodes very critically for a fashion company that is geared toward women. If the text had been screened through the IBM filter beforehand, this problem could have been identified and avoided.
Sentiment analysis of text used by Nike for the Colin Kaepernick campaign:
Inference:
The Sentiment Analysis of Nike's text is Neutral in its tone. This is again indicative of how Nike maintained a neutral tone in its text and used positive and 'inspiring' imagery but let the controversy more or less speak for itself.
Impact Analysis
To study the impact of these advertisements on the brands, I have done a google search and run the first 50 web results which were news headlines through IBM Watson's Sentiment Analysis to gauge the sentiment score for each of the companies. These are the results.
Sentiment analysis of Google search results of news headlines:
