In the vast digital landscape of social media, where every post competes for attention, have you ever wondered why some shine brighter than others?

It's easy to get caught up in the numbers game, focusing solely on metadata like hashtags, engagement rates, and audience size. Yet, there's a pivotal factor often overlooked amidst the metrics frenzy: the content itself. 🔥

At Exolyt, we're challenging the status quo with an innovative approach to social media analysis. Our mission? To revolutionize how we understand and leverage content's role in online success and learn from it.

To do so, Exolyt devised a cutting-edge AI solution to redefine how we perceive social media content, particularly on TikTok. Harnessing the power of advanced computer vision algorithms, our video analysis dives deep into the heart of TikTok videos, unlocking insights that transcend conventional analytics. 

• Sentiment Analysis: Emotions drive engagement. With sentiment analysis, we track facial and vocal emotions expressed to gauge the tone of TikTok videos, unveiling the underlying feelings that captivate viewers and spark conversations.

• Text Extraction: Words carry weight, even in a visually driven platform like TikTok. Our tool extracts text from ‘within’ the videos, uncovering words used visually in the videos that are often overlooked during analysis. These words may contain captions and trending phrases that resonate with audiences. 

• Demographic Insights: Understanding your audience is key to crafting compelling content. Through demographic analysis, we unveil the age, gender, and geographic distribution of TikTok content creators, empowering creators and brands to tailor their message effectively.

• Colour and Audio Analysis: Visual and auditory elements shape the viewer experience. By dissecting colour schemes and analyzing audio patterns, our AI discerns the aesthetic and sonic signatures that elevate TikTok videos from ordinary to extraordinary.

Video AI Analysis Examples (Disclaimer: This is not the actual screenshot of the tool)

Our video analysis tool can accurately identify objects, extract texts, analyze sentiments and demographic information, and conduct colour and audio analysis within every second of TikTok video.

This holistic approach transcends traditional social media analytics, bridging the gap between content specifics and performance metrics. It's not just about how many likes or shares a post receives; it's about understanding why!

We launched a few pilot projects with our customers to initiate this project. Here are some of the findings:

a. A Global Creative Agency

As a socially led agency that studies social behaviours within online communities, cultures, and subcultures, our client wanted to reach the right people in a strategic, relevant, and effective way. To do this, our client wanted to understand a specific topic's trends, demographics, and content patterns and decode them utilizing Exolyt’s Video AI Analysis.

To do so, they identified some key brands from the luxury industry on which they were looking to gain more insights into how people wear its jewellery while also gaining insight into the types of captions and comments seen in that particular category's content.

The Project Brief included some details like: 

• Key brands: Graff, Bulgari, and Cartier, among others.
• Time window: 6 to 12 months
• Related hashtags: #luxuryjewellery 

Sample Findings from This Pilot:

• Main Hashtag distribution during the given timeframe

Source: Exolyt Video AI Analysis

• Demographic Distribution by Main Hashtag

Source: Exolyt Video AI Analysis

• Most frequent age groups by Hashtag

Source: Exolyt Video AI Analysis

• Facial sentiments overtime for Cartier

Source: Exolyt Video AI Analysis

Note: Because of the aggregation, there are more negative sentiments. Negative sentiments include more emotion classes (anger, disgust, fear), while positives include only "happy" emotion classes.

• Most used words in this category

Source: Exolyt Video AI Analysis

Some Key Takeaways from this Pilot Case:

1. Most of the content within these topics are produced by women aged 20-29, followed by women aged 30-39.
2. Almost half of the content published under these hashtags does not feature creator faces.
3. Cartier is the brand with the largest content across all age groups.
4. Vancleef mentions shrink as the age of the creator in the video increases.
5. The content produced by female users features more blanched brand distribution, with Cartier being the most common, followed by Vancleef.
6. Over time, the number of male creators in the space has risen, while content featuring no creator's face has decreased.
7. Facial sentiments indicate a sharp peak of predominantly negative emotions in creators around May 2023, followed by a peach of positive emotions in videos published towards September 2023.

b. McCann Paris

Our customer, an entity of McCann Worldgroup and a leader in the advertising sector that develops integrated marketing and communication solutions for clients in France and worldwide, wanted to study the public opinion on TikTok for one of their biggest clients from the beauty industry.  

The main objective was to understand the demographic makeup of the user-generated content across TikTok for the brand, along with the colour and sound analysis of the content, to determine which skincare products are popular, which skin tonality commonly constitutes their usage, and what the users have to say about the products.

The Project Brief included some details like: 

• Key brands: Large Beauty/Makeup Brand
• Time window: 3 months
• Related Hashtags: (brand name)
• Key markets: GB

Sample Findings from This Pilot:

• Age group distribution for the branded hashtags

Source: Exolyt Video AI Analysis

• Voice sentiments overtime

Source: Exolyt Video AI Analysis

• Facial Colour Analysis

Source: Exolyt Video AI Analysis

• Video Colour Analysis

Source: Exolyt Video AI Analysis

Note: This is a limited data sample based on videos that used the hashtag Loreal. The patterns found in this sample might not be generalizable to all TikTok videos across industries, markets, and regions.

Some Key Takeaways from this Pilot Case:

1. Gender and age group distribution for the branded topics/hashtags showcased: Most of the content for the branded topics is produced by women aged 20-29, and creators aged 20-29 prefer tagging their videos with the #brand hashtag, while creators between 30 and 39 use the #brand hashtag more frequently.
2. The creators' facial and voice sentiments maintain a stray pattern throughout the months, with a mostly positive and negative mix and very little neutral sentiment.
3. The diversity of the creator's facial colours illustrates the prevalence of lighter and warmer colours, which indicate particular skincare trends.
4. The videos on this topic predominantly have neutral background colours. Blacks and greys seem common, with orange-colored lighter backgrounds appearing in some videos.
5. Video background colour does not substantially affect video engagement in the sample we examined in this study.

Social intelligence consumers across various industries can find multiple applications of this data from TikTok. 

This goes beyond boosting brand visibility, engagement, influencer marketing, or competitive analysis, as it can be an excellent asset for qualitative research and an extension of digital ethnography.

With an entire generation shifting their social media communications channels into one platform, TikTok offers a vast untapped market research opportunity to understand their perspectives, experiences, evolving cultures, and the issues that are important to them.

The enhanced video analysis capabilities offer this opportunity to learn from the content of such people, empowering you to conduct better qualitative research and comprehensively comprehend the customer journey for various products and services.

It helps to unwrap TikTok trends that have entwined entertainment and commerce, caused cultural shifts, changes in lifestyle, and buying patterns by analyzing videos frame by frame, accurately identifying objects, extracting texts, analyzing sentiments and demographic information, and conducting colour and audio analysis within every second of a TikTok video.

It also answers the question: Why does some social media content outshine the rest? The answer, of course, lies in the intricate tapestry of visuals, words, emotions, and audience resonance woven into every post. 

At Exolyt, we're committed to unraveling this tapestry and empowering marketers, analysts, and researchers. Join us on this journey of discovery, where the power of AI meets the art of storytelling. Together, let's redefine the rules of social media success, one captivating post at a time. 

Who can join this pilot project?

Companies who would like to understand the visual content of TikTok videos, not just raw metrics, and have use cases for such analysis.

What is included in the project?

• Industry-leading AI content analysis model that can analyze each second of TikTok video.
• Free custom analysis of TikTok videos by Exolyt (participation is entirely free, as the analysis is still in the Beta phase).

How to participate in the project?

• If you are a customer, the participation process is relatively straightforward. You can directly connect with us and share your requirements or plan one with our data scientist while continuing to use the platform as usual.
• If you are NOT a customer, you can still participate with limited usage of the freemium platform plan but with full access to Video AI analysis project benefits.

Disclaimer: You do not need to log in to Exolyt to utilize this project's benefits, as the feature has not yet been deployed on the platform.

We will customize the project according to your video analysis requirements and host it in close collaboration with your company's contact person.

Your participation could shape the future of content analysis on TikTok. If you're interested in learning about this technology and contributing to developing cutting-edge AI tools, we'd love to have you join us.