Trend Engine: how we find trends before they explode

A look under the hood of Content Hunter: how the Trend Engine collects signals, separates real trends from noise, and alerts you days before the peak.

The Trend Engine is the core of Content Hunter that answers the main question: what will be viral tomorrow? In this article we explain how it works — no marketing magic, just mechanics.

Step 1. Signal collection

Every hour the Trend Engine refreshes data on your sources: creators, hashtags and keywords across Instagram Reels, TikTok and YouTube Shorts. For every video it records views, engagement, growth velocity, sounds used and visual patterns.

Step 2. Anomaly detection

A creator's baseline is their median views. When a video exceeds the baseline by 5x, 10x or 50x, that's an anomaly. One anomaly is not yet a trend. But when the same pattern (hook, format, sound) produces anomalies for several independent creators, the probability of a trend rises sharply.

Step 3. Scoring and clustering

The Scoring Engine rates every anomaly on dozens of factors: growth velocity, freshness, spread across niches, hook type, audience reactions. Similar videos are grouped into clusters — that's how a 'trend' is born, not just a list of videos. Every trend gets a forecast: growth, maturity or decline stage.

Step 4. Alerts

When a trend is confirmed and still early, you get a notification: which mechanic is growing, example videos, a breakdown of why it works, and a ready script for adaptation. On average this happens 3–7 days before the trend peaks.

Why this is hard to do by hand

A person can track dozens of creators. The Trend Engine tracks thousands of sources simultaneously and never gets tired: anomalies are computed with the same precision at 3 AM and at noon. That's why our users systematically enter trends before their competitors.