ARTFindr

Designed With 😇 :

• Aws
• JavaScript
• Nextjs
• React
• Vercel

ART Finder is a web application designed to simplify market analysis for brands, providing Effective CTAs with their conversion rates, Best hooks with their engagement rates and many other metrics such as sentiment analysis, hot trends etc.

Basic Functionality of ART Finder includes

Data Processing: The application uses LangFlow to manage the workflow, sending a detailed prompt to the Perplexity API.

Structured Output: The response is formatted into a structured JSON using Gemini's controlled generation feature.

Dashboard Display: The structured data is sent to the dashboard via an API call, where it is displayed for user interpretation.

The user-friendly dashboard presents the findings in an intuitive interface, allowing brands to gain valuable insights into market trends and consumer preferences.

Technologies Used:

React.js: Builds the interactive user interface.

Next.js: Provides server-side rendering and routing capabilities.

LangFlow: Manages the workflow and interfaces with the Perplexity API for data analysis.

Perplexity API: Processes detailed prompts to generate comprehensive insights.

Gemini: Ensures consistent JSON output through controlled generation.

API Testing: Ensures reliable communication between components.

By integrating these technologies, ART Finder offers a seamless experience for brands to conduct effective market analysis.

Further updates to ART Finder will include functionalities like Retrieval Augmentation Generation using Astra DB by DATASTAX

GitHub Link 🔗

• https://github.com/abhinandanwadwa/findart

Deploy Link 🔗

• https://artfinder.abhinandan.me

Problem it solves 🙅‍♂️

• Task 1: Automated Research and Trigger Finder (ART Finder) Objective: The objective of ART Finder is to streamline the research phase of ad creation by automating data gathering and analysis. This tool will: Identify user pain points and triggers from multiple data sources such as Google, YouTube, Reddit, Quora, and app reviews. Analyze competitor ads and strategies to uncover high-performing hooks, CTAs, and content formats. Generate actionable insights and suggestions to help marketers craft effective, user-centric ads. Key Features: Comprehensive Research Automation: Scrapes data from blogs, forums, social media, and app reviews. Analyzes YouTube videos and competitor ads to identify trends, pain points, and effective solutions. Actionable Insights Generation: Summarizes key triggers and user problems. Suggests best-performing hooks, CTAs, and solutions tailored to the topic and audience. Reference Dashboard: Provides direct links to scraped YouTube videos and competitor ads for easy validation and inspiration. Visualizes insights with graphs, word clouds, and sentiment analysis. User-Centric Interface: Simple input fields for topics and brand guidelines. Intuitive dashboard showcasing insights and recommendations at a glance.

Challenges I ran into 🙅‍♂️

• Data Gathering: Scraping diverse platforms like YouTube, Reddit, and Quora was challenging due to differing structures, anti-scraping measures, and rate limits. Data Quality: Filtering relevant insights from noisy, unstructured data required advanced NLP techniques. Competitor Analysis: Accessing and analyzing diverse ad formats and CTAs consistently was difficult. Sentiment Accuracy: Contextual sentiment analysis, especially for sarcasm and slang, required fine-tuned models. Visualization: Designing intuitive dashboards with meaningful visualizations like graphs and word clouds was a balancing act. Scalability: Processing large-scale data efficiently while ensuring real-time performance was resource-intensive.