1. The Journey: Breaking Free from Recycled Spreadsheet SaaS

If you spend any time in the creator economy or digital product space, you’ve seen the ads: expensive monthly SaaS subscriptions ($79 to $199/month) promising exclusive access to "secret breakout YouTube niches."

For a while, I was tempted. But after inspecting how most niche research platforms operate, I realized a frustrating truth: many of these tools sell overcrowded spreadsheet lists driven by shallow heuristics. They look at basic subscriber-to-view ratios, wrap them in a dark-mode dashboard, and charge a recurring fee.

I asked myself a simple engineering question: Why am I renting black-box insights when the underlying data sources—Google Trends time-series datasets, YouTube Data APIs, and millions of un-mined viewer comment threads—are publicly accessible to engineers?

So instead of paying for another subscription, I decided to build it myself. The result is NicheRadar—a full-stack niche intelligence platform designed to function as an early-warning feeder system:

Upstream Signal

1. Detect Demand

Spot rising queries on Google Trends (BigQuery + API alpha) before creators flood the YouTube search results.

Downstream Supply

2. Verify Saturation

Scan YouTube Data API v3 to confirm whether incumbent channels dominate or if small creators are breaking out.

Qualitative Goldmine

3. Mine Pain Points

Parse thousands of viewer comment threads with Gemini AI to uncover exact unanswered questions and commercial gaps.

2. High-Level Architecture: Engineering the NicheRadar Stack

Building an autonomous research engine requires an architecture that orchestrates scheduled ingestion, asynchronous third-party APIs, cognitive AI structured extraction, and clean data presentation.

System Flow Diagram Multi-Stage Pipeline

1. Demand Ingestion

Google Trends BigQuery + Workspace CSV (`top searches.csv`)

2. Backend Engine

FastAPI Normalizer + Noise Filter (Excludes Celebrities)

3. Qualitative AI

YouTube Data API v3 + Gemini 3.5 Flash Comment Mining

4. Editorial Canvas

PostgreSQL (RLS) + React Paper-White UI Terminal

How Each Layer Operates:

• Noise Filtering & Normalization: Raw Google Trends feeds are 70% sports scores, holidays, and celebrity drama. Our Python engine applies aggressive domain regex rules (`STOP_PATTERNS`) to drop one-day events (`4th of july`, `celebrities`, `sports scores`), preserving high-signal software, AI, business, and educational niches.
• YouTube Content Density Checking: Once a query passes trend validation, the backend scans YouTube Data API v3 to measure creator competition (`youtube_content_density`) and incumbent market dominance (`saturation_score`).
• AI Cognitive Layer (Gemini 3.5 Flash): We integrate Google Gemini via structured JSON (`ConfigDict`) schemas to categorize topic clusters and classify thousands of viewer comments by underlying intent.
• Editorial Research Frontend: Built with React + Vite + Vanilla CSS tailored to a calm paper-white canvas (`#fcfbf9`), replacing noisy marketing charts with actionable research cards.

3. The Power of Mixing Quantitative Velocity with Qualitative Audience Frustration

If there is one technical takeaway from engineering NicheRadar, it is this: Quantitative data tells you where people are looking. Qualitative data tells you what they are desperate to buy or learn.

Signal Dimension	Data Source	What It Reveals
Search Velocity (+%)	Google Trends BigQuery	Early-stage interest spikes before mainstream YouTube saturation.
Creator Saturation (%)	YouTube Data API v3	Whether a niche is dominated by giant incumbents or open to newcomers.
Audience Pain Points	Comment Threads + Gemini AI	Specific unanswered questions, friction points, and feature gaps.
Buying Intent Rating	AI Commercial Classifier	Direct evidence of willingness to pay for software, templates, or courses.

By joining quantitative search gain (+%) with qualitative comment gaps, NicheRadar computes a holistic Opportunity Score (0–100):

Composite Scoring Formula

Opportunity = 0.35(Velocity) + 0.25(YouTube Fit) + 0.20(Pain Points) + 0.20(100 - Saturation)

When a topic combines rising Google search demand, low creator saturation, and high qualitative comment frustration (unanswered questions & commercial buying intent), NicheRadar flags it as a breakout target. You don't just discover a keyword—you discover the exact script outline, software feature gap, or product angle the audience is waiting for.

LN Author

Lorenzo N.

Lead Systems Architect

Published by ZoraPro Operations Lab • Updated August 2026

Lorenzo is the founder and lead systems architect at ZoraPro, with 8+ years of engineering experience across distributed data pipelines, BigQuery analytics, and practical AI agent deployments. He builds inspectable data architectures that replace opaque SaaS tools.

GitHub Profile LinkedIn Twitter / X

Why I Built NicheRadar: Engineering a Full-Stack YouTube & Google Trends Intelligence Engine

Quantitative Velocity + Qualitative Audience Frustration