We Turned Every YouTube Comment into a Business Signal
Millions of comments. Sentiment-scored by AI. Buyer intent surfaced. Brand intelligence delivered in real time.
Written by Shubham(opens in new tab), Founder at BestlaTech · Published June 26, 2026
Engagement Type
Full-Stack AI Platform: Built from Ground Up
Core Tech
YouTube Data API · OpenAI · NDJSON · Metabase
Delivered For
YourComments.ai
Client
YourComments.ai
Duration
Ground-up build
Category
AI Analytics · NLP · Creator Intelligence
Client's Markets
US
Comments Processable at Launch
1M+
Full ingestion and scoring pipeline
Report Types in Metabase
15+
Covering creator, brand, and competitor use cases
Per-Comment Score Time
~2s
Fetch → normalise → score → store
Comment-Level Granularity
100%
Every comment and reply individually classified
The Challenge
Creators Flying Blind
A YouTuber with 500,000 subscribers gets 4,000 comments per video and reads 30. The remaining 3,970 contain loyalty signals, purchase intent, and early churn warnings: unread and lost within 48 hours.
Brands Ignoring a Free Research Channel
A competitor's YouTube comment section fills with unfiltered customer reactions the moment a product launches. Most brands never look. Those that do read manually, which doesn't scale past a few hundred comments.
No Tool Built for Comment-Level Depth
Existing social listening tools score at video level. None delivered per-comment sentiment, intent detection, reply-chain analysis, and trend data: the granularity that makes the intelligence actually useful.
Use Cases
Know Your Audience Better Than They Know Themselves
Every comment on every video: sentiment-classified, intent-flagged, loyalty-scored. A creator can see which videos drive the strongest positive sentiment, which topics generate purchase-intent language, and which comment patterns indicate a subscriber moving from casual viewer to loyal advocate.
- Purchase intent language ("where can I buy", "link?", "just ordered")
- Loyalty indicators (returning commenters, positive sentiment trend over time)
- Negative sentiment clusters (early warning before a video underperforms)
- Question patterns (what your audience wants you to make next)
Your Competitor's Comment Section is a Free Focus Group
A brand analyst selects a competitor's channel or specific videos, the platform ingests all comments, runs sentiment analysis via OpenAI, and surfaces structured intelligence: top themes, sentiment breakdown, feature mentions, and emotional tone over time. No surveys. No interviews. No waiting.
- Competitor product sentiment (what customers love and complain about)
- Feature gap analysis (what users wish the competitor's product did)
- Audience emotional profile (how a competitor's launch landed in the market)
- Trend data (how sentiment shifted from launch day to 30 days out)
Start with the Data Problem. Then Build the Intelligence Layer.
Before any code was written, we mapped the exact intelligence each buyer type needed, the architectural constraints of building at scale, and what consistent AI output actually requires in practice. Those questions shaped every decision.
Define the Business Questions First
Reports were designed before the pipeline was built. For creators: purchase intent, loyalty signals, negative sentiment clusters. For brands: competitor sentiment breakdown, feature gap analysis, emotional tone over time.
Engineer for Consistency, Not Just Accuracy
We solved the structured AI output problem through prompt architecture: JSON output format, temperature control, per-edge-case fallback instructions, and confidence scoring on every result.
Build for Scale from Day One
NDJSON streaming and a per-comment PostgreSQL schema were chosen at design stage. The architecture handles 10,000 comments and 10,000,000 comments the same way: no redesign needed as the platform grows.
Weekly Demos Against Real Channel Data
Every sprint ended with a live demo using actual YouTube channels. Edge cases surfaced during demos, not after launch.
A Five-Stage AI Pipeline That Turns Raw Comments into Structured Intelligence
Every layer purpose-built: data ingestion, normalisation, AI scoring, storage, and reporting. Not a plugin, not a wrapper, not a low-code workflow.
YouTube Data API v3
Quota-aware ingestion layer fetching every comment and reply for any public video or channel. Pagination automatic, large jobs queued as background tasks, rate limits managed, partial-job recovery built in.
Cleaning Before Scoring
Language detection, spam filtering, deduplication, thread structure preservation. Only clean, structured data reaches OpenAI.
OpenAI with Structured Prompt Engineering
Deterministic JSON output per comment: sentiment classification, intensity score, intent category, confidence rating. Temperature controlled. Edge cases handled with specific prompt logic.
NDJSON Streaming into PostgreSQL
Scored results stream as NDJSON, one record per line, into a structured PostgreSQL schema. One row per comment with full sentiment metadata, queryable at any granularity.
Metabase Dashboards Built Around Business Questions
15+ report types built around specific buyer questions, not around the data model. Creator dashboards: sentiment trend, purchase intent volume, loyalty score, negative spike alerts. Brand dashboards: competitor breakdown, feature mentions, emotional tone, launch comparison.
The Full Stack: From YouTube API to Business Dashboard
YourComments.ai is not a wrapper around an existing tool. It is a full-stack AI analytics platform built from the ground up: data ingestion, NLP pipeline, storage architecture, and a reporting layer.
FETCH
YouTube Data API Ingestion
The platform connects to the YouTube Data API v3 and fetches comments and replies for any video or channel on demand. Pagination is handled automatically (whether a video has 200 comments or 200,000) with rate limit management and retry logic built in.
PARSE
Cleaning and Normalisation
Raw YouTube comment data passes through a normalisation layer: language detection, spam filtering, deduplication, thread structure preservation (comment → reply chains maintained), and structured formatting before any analysis runs.
SCORE
OpenAI Sentiment Analysis
Each normalised comment is sent to OpenAI's API with a structured prompt returning consistent output: sentiment classification (positive / negative / neutral), intensity score, intent signals (purchase, complaint, question, praise), and a confidence rating. Every comment and reply scored individually, not the video.
STORE
NDJSON Pipeline into PostgreSQL
Scored results are streamed as NDJSON, one record per line, processed and written without loading the full dataset into memory. Results land in a structured PostgreSQL schema: one row per comment, with sentiment scores, intent flags, metadata, and thread context all queryable.
VISUALISE
Metabase Dashboards
The PostgreSQL data powers a Metabase reporting layer with 15+ dashboard types: built to answer specific business questions per channel, per video, per time range, and per sentiment type.
Technologies Used
The Hard Parts
YouTube API Quota Management
The YouTube Data API v3 has strict daily quota limits. We built a quota-aware job scheduler that batches requests, prioritises high-value channels, queues large jobs overnight, and recovers gracefully without losing progress or corrupting partial datasets.
Prompt Engineering for Consistent AI Output
We engineered a prompt system that returns deterministic JSON output for every comment (sentiment, intensity, intent category, and confidence score) with temperature set to minimise variance. Edge cases (sarcasm, mixed sentiment, non-English text, emoji-only comments) each handled with specific prompt instructions and fallback logic.
Handling Scale Without Memory Collapse
We chose NDJSON as the streaming format precisely because it handles arbitrarily large datasets as a stream, one record per line, processed and written without ever loading the full dataset. Processing 1,000 comments and processing 10,000,000 comments uses the same architecture.
Building Reports That Answer Business Questions
We worked with YourComments.ai to define the specific questions each report needed to answer: for creators: 'which of my videos has the highest purchase intent?', for brands: 'how did sentiment on my competitor's channel change after their product launch?' - and built Metabase queries around those questions, not around the data structure.
Key Features
YouTube Data API v3 integration with automatic pagination, rate-limit management, and full retry logic
Comment normalisation engine: language detection, spam filtering, deduplication, thread structure preservation
OpenAI sentiment scoring pipeline: per-comment classification, intent detection, intensity scoring via structured prompt engineering
NDJSON streaming layer for high-volume data ingestion without memory bottlenecks
PostgreSQL data model storing every comment with full sentiment metadata, queryable at any granularity
15+ Metabase report types: sentiment trends, loyalty scores, intent maps, competitor breakdowns, time-series analysis
Job queue and rate limit manager: large channel ingestion as background jobs with progress tracking
Multi-channel support: analyse any public YouTube channel or video, own or competitor
Results & Impact
1M+ comments processable through the full ingestion and scoring pipeline
15+ report types live in Metabase covering creator, brand, and competitor use cases
100% comment-level granularity, every comment and reply individually classified
Per-comment scoring end-to-end (fetch → normalise → score → store) in approximately 2 seconds
Manual comment reading eliminated: replaced by automated sentiment intelligence at scale
Creator and brand intelligence previously impossible now available on demand for any public YouTube channel
YourComments.ai can now ingest any public YouTube channel, score every comment and reply individually via OpenAI, and deliver structured intelligence in real time, whether that's a creator understanding their audience's purchase intent or a brand analyst reverse-engineering a competitor's launch reception. The platform turned the most underutilised data source on the internet into a business intelligence tool.
Before vs After
| Before | After | |
|---|---|---|
| Comment analysis | Manual reading: 30–50 comments per video | Automated: every comment on every video |
| Sentiment data | None: gut feel only | Per-comment: sentiment, intent, intensity, confidence |
| Brand intelligence | No competitor comment data | Full competitor channel analysis on demand |
| Reporting | None | 15+ live dashboards: trend, intent, loyalty, competitor |
| Scale | Doesn't scale past a few hundred comments | Designed for millions |
BestlaTech took a complex idea, analysing sentiment across millions of YouTube comments in real time, and turned it into a system that actually works at scale. The dashboards gave us insight into audience intent we'd never had before. They built something we didn't think was possible in the timeline we had.
Founder, YourComments.ai
Read our verified reviews on Clutch (opens in new tab)If You Work with Audience Data at Scale, This Architecture Applies to You.
YourComments.ai is one implementation. The underlying system (data ingestion at scale, AI-powered classification, structured storage, business intelligence dashboards) is applicable to a broader set of problems.
YouTube Creators & Agencies
You publish content at scale and need to understand what your audience actually thinks. Comment-level intelligence changes how you brief content, monetise audiences, and grow.
Brand & Marketing Teams
Your competitors' YouTube channels are a live, unfiltered focus group. Comment intelligence tells you how product launches land and where sentiment is shifting: before it shows up in sales data.
SaaS Founders & Product Teams
Any product touching user-generated content at scale (reviews, support tickets, community posts, social comments) can use the same NLP pipeline. If you have a large text corpus and need structured intelligence, this is the architecture.
Want to Build Something Like This?
Whether you need a YouTube intelligence platform, a custom NLP pipeline, or AI features built into an existing product: book a call and we'll tell you exactly what's buildable and in what timeframe.
Book Your Free Discovery Call (opens in new tab)Fixed scope. Fixed price. Zero surprises.
Frequently asked questions
What is YouTube comment sentiment analysis and why does it matter?
Can brands really use YouTube comments for competitor intelligence?
How does the OpenAI sentiment scoring work at scale?
What is NDJSON and why did you use it?
Can BestlaTech build a custom sentiment analysis platform for my business?
How long would it take to build something similar?
Does this only work for YouTube, or can you analyse comments from other platforms?
What makes this different from existing social listening tools?
Who builds custom AI and NLP analytics platforms from scratch?
How much does it cost to run an LLM over millions of records?
Talk to an expert
Get expert advice from our official advisors
More of Our Case Studies
Explore our diverse portfolio of successful projects and innovative case studies that showcase our expertise in delivering top-notch solutions.
Customer Success SaaS
CAMCESS: How BestlaTech Connected 11 Data Sources into a Real-Time Customer Success Platform
End-to-end enterprise API integration for a B2B customer success SaaS, connecting 11 siloed data sources into a single real-time customer health dashboard, replacing weekly manual reconciliation entirely.
Mobile Web / Performance & Debugging
Diagnosing and Fixing an iOS Safari Navigation Crash on a Top-Tier Indian Bank's High-Traffic Website
Production incident engagement for a top-tier Indian private bank: the website's hamburger menu froze iPhones within seconds of a tap. Root cause diagnosed, patched, and verified across iOS versions in 2 weeks.
InsurTech / Web App Development
A Multi-Project Digital Suite for an Indian Insurance Broker: Premium Calculator, Claims Portal, Partner Dashboard & Lead Microsites
Multi-project engineering retainer for an Indian insurance broker: four tools delivered on one shared platform: an accurate premium calculator, a self-serve claims tracking portal, a partner/agent dashboard, and campaign lead microsites, now serving 10,000+ monthly users.
CMS Development / WordPress / Enterprise
An Enterprise WordPress CMS for a Leading CPaaS Company: 1,000+ Pages, Independent Publishing, Sub-2s Loads
Enterprise CMS build for a leading CPaaS company: a custom component-based WordPress theme, structured content architecture, and multi-language support that lets the marketing team publish independently at 1,000+ page scale with sub-2-second loads.
Legal Tech / CLM / AI Platform
Building an AI-Powered CLM Platform from Zero: Contract Generation, Clause Risk Scoring, Playbooks, E-Signature & Renewals
Full CLM (Contract Lifecycle Management) platform built from scratch for a legal-tech SaaS company (client name withheld). AI contract generation, metadata extraction and characterization, clause-by-clause risk scoring with suggested alternative language, playbook-driven review, 30/60-day renewal categorization, DocuSign e-signature, and Microsoft Word integration via a Zoho plugin.
Travel & Tourism
Building a Complete Digital Presence for Travel With Bug Travel Company
End-to-end digital transformation for a travel agency - professional responsive travel website development with WordPress CMS, online booking system, tour package management, and SEO optimization for organic lead generation.
On-Demand Services
Full-Stack Mobile App and Web Development for Mechgiri On-Demand Auto Service Platform
Custom mobile app development and web application for on-demand vehicle servicing - featuring real-time tracking, technician management, automated booking system, secure payment gateway integration, and cross-platform admin dashboard.
E-Commerce
Custom E-Commerce Website Development for BugSpeaks Gut Health Brand
Complete digital transformation from offline to online business - custom e-commerce platform development with Django, automated inventory management, secure payment integration, order tracking system, and scalable cloud infrastructure for health and wellness products.
Agritech / Smart Farming
Redesign and Performance Optimization for Letcetra Agritech
Complete frontend modernization for smart agriculture platform - Next.js development with server-side rendering, React component architecture, mobile-first responsive design, advanced SEO optimization, and performance enhancement for agritech industry leader.
FinTech / Financial Operations
How a US FinTech SaaS Unified Stripe, QuickBooks, Salesforce & Slack into One Real-Time Financial Ops Layer
Multi-system API integration for a venture-backed US fintech SaaS. Stripe, QuickBooks, Salesforce, and Slack connected into one real-time financial operations layer, cutting the month-end close from 3 days to 1 hour.
Logistics / Supply Chain
How a US Logistics Operator Unified 5+ Carrier APIs, WMS & ERP into One Real-Time Dispatch Layer
Multi-carrier API integration for a US logistics operator: parcel, LTL, and freight carrier APIs connected with the warehouse management system and ERP into one dispatch layer with automated rate shopping and unified tracking.
AI Automation / Customer Support
How a B2B SaaS Support Team Automated Ticket Triage with an LLM Pipeline: 500+ Tickets a Day, Routed in Seconds
LLM-powered support ticket triage for a B2B SaaS company: automatic classification, prioritization, and specialist routing of 500+ daily tickets, with confidence thresholds and human-in-the-loop review built in.
AI Automation / Accounting
How a US Accounting Firm Automated Invoice & Expense Processing with an OCR + LLM Pipeline: 2,000+ Documents a Month
AI document processing for a US accounting firm: an OCR + LLM pipeline that extracts line items from scanned invoices, auto-categorizes them to the chart of accounts, and flags anomalies, with a human review queue for low-confidence documents.
AI Code Rescue / Security
Rescuing an AI-Generated MVP: 15 Critical Vulnerabilities Patched Before the First Paying Customers Arrived
AI code rescue for a seed-stage SaaS startup: a security audit and hardening of an AI-generated MVP: 15 critical vulnerabilities patched, test coverage built from zero to 80%, and a CI/CD pipeline put in place before launch.
AI Code Rescue / Performance
From Crashing at 50 Users to Stable at 1,000+: Rescuing an AI-Generated App That Failed Under Real Load
Performance rescue for a consumer app startup: an AI-generated app that crashed under real traffic, fixed in 3 weeks: N+1 queries eliminated, memory leaks patched, caching added, and stability proven with load tests.
AI Integration / Knowledge Management
Replacing Keyword Search with a RAG Assistant: Grounded, Cited Answers Over a B2B Knowledge Base
RAG integration for a B2B SaaS platform. LLM-powered semantic search over the product knowledge base with grounded, cited answers, evaluated against a test set before launch. Support tickets down 40%.
AI Integration / HR Tech
AI Resume Screening Inside an ATS: Parsing, Skill Extraction & Explainable Role Matching for an HR Tech Platform
AI integration for an HR tech platform: resume parsing, skill extraction, and explainable candidate-to-role matching built into the existing ATS workflow. Screening 5x faster, with recruiters making every decision.