[Paper Review] Follow Us and Become Famous! Insights and Guidelines From Instagram Engagement Mechanisms

Pier Paolo Tricomi, Marco Chilese+2 more|arXiv (Cornell University)|Jan 17, 2023

Digital Marketing and Social Media4 citations

TL;DR

This paper proposes an interpretable machine learning framework to predict Instagram post engagement (likes and comments) using a large-scale dataset of 10 million influencer posts across nine categories and five influencer tiers. By combining deep learning for feature extraction with interpretable models, the approach achieves up to 94% F1-score and reveals category- and tier-specific guidelines for creating highly engaging content based on visual, textual, and structural features.

ABSTRACT

With 1.3 billion users, Instagram (IG) has also become a business tool. IG influencer marketing, expected to generate $33.25 billion in 2022, encourages companies and influencers to create trending content. Various methods have been proposed for predicting a post's popularity, i.e., how much engagement (e.g., Likes) it will generate. However, these methods are limited: first, they focus on forecasting the likes, ignoring the number of comments, which became crucial in 2021. Secondly, studies often use biased or limited data. Third, researchers focused on Deep Learning models to increase predictive performance, which are difficult to interpret. As a result, end-users can only estimate engagement after a post is created, which is inefficient and expensive. A better approach is to generate a post based on what people and IG like, e.g., by following guidelines. In this work, we uncover part of the underlying mechanisms driving IG engagement. To achieve this goal, we rely on statistical analysis and interpretable models rather than Deep Learning (black-box) approaches. We conduct extensive experiments using a worldwide dataset of 10 million posts created by 34K global influencers in nine different categories. With our simple yet powerful algorithms, we can predict engagement up to 94% of F1-Score, making us comparable and even superior to Deep Learning-based method. Furthermore, we propose a novel unsupervised algorithm for finding highly engaging topics on IG. Thanks to our interpretable approaches, we conclude by outlining guidelines for creating successful posts.

Motivation & Objective

• To understand the underlying mechanisms driving Instagram engagement beyond likes, including the growing importance of comments.
• To develop a transparent, interpretable model for predicting post popularity using real-world, large-scale data.
• To identify category-specific and tier-specific patterns in visual, textual, and structural features that drive likes and comments.
• To propose an unsupervised method for detecting highly engaging topics on Instagram without requiring domain-specific labels.
• To provide actionable, evidence-based guidelines for influencers and marketers to design high-engagement content in advance.

Proposed method

• The study uses a dataset of 10 million Instagram posts from 34,000 global influencers across nine categories and five audience tiers.
• It leverages state-of-the-art deep learning models to extract visual and textual features from posts, which are then used as input for interpretable machine learning models.
• A supervised learning approach with interpretable models (e.g., decision trees, linear models) is used to predict engagement (likes and comments), achieving high F1-scores up to 94%.
• An unsupervised topic modeling approach is proposed to detect hot topics in each category and tier by analyzing textual and visual content patterns.
• Statistical analysis and feature importance ranking are used to interpret model outputs and derive actionable guidelines.
• The dataset and code are released to support reproducibility and future benchmarking.

Experimental results

Research questions

• RQ1 What visual, textual, and structural features most strongly predict likes and comments on Instagram posts?
• RQ2 How do engagement patterns differ across categories and influencer tiers (e.g., micro vs. macro influencers)?
• RQ3 Can an interpretable machine learning model predict post engagement with performance comparable to deep learning while remaining transparent?
• RQ4 What are the most engaging topics in each category and tier, and how can they be detected without labeled data?
• RQ5 How can the insights from this study be translated into practical, data-driven guidelines for content creation?

Key findings

• The model achieves up to 94% F1-score in predicting Instagram post engagement, outperforming or matching deep learning baselines while remaining interpretable.
• Likes are primarily driven by visual features such as image quality, color, and subject type, while comments are predominantly influenced by caption content, including questions, calls to action, and emoji usage.
• Influencer tier significantly affects engagement patterns: higher-tier influencers show more category-specific behaviors in content design.
• The most engaging posts in the 'Pets' category feature warm-colored, indoor or natural outdoor scenes with highly cute animals, especially horses, exotic animals, and dressed cats.
• Travel posts with young female subjects, low male presence, and positive sentiment in captions generate more likes and comments, especially when featuring seaside or holiday themes.
• Captions that include questions (e.g., 'Which outfit do you prefer?'), calls to action (e.g., 'Tag a friend'), and hashtags at the end with line breaks are significantly more engaging.