Methodology
Where the data comes from, and how far to trust it.
Every trend tool claims it spots things early. This page shows how Lolik actually works, and the public log shows whether it worked.
The sources
Live data, pulled on a schedule, cached so your usage never hits a provider quota.
Trending and niche video charts, view counts and channel signals, refreshed several times a day. View counts also power the outcome tracking behind Receipts.
Live trending searches across 8 English-speaking markets, with spike detection and velocity.
Real news stories from hundreds of outlets, deduplicated and tracked from first appearance.
GitHub repository momentum and community chatter feed the Signal Vault as early-confirmation sources.
How the scoring works
A trend is interesting when attention grows faster than the content serving it. That gap is the arbitrage score: high demand, low supply, time still on the clock.
Every pre-breakout alert is labeled with its phase: early (accelerating below its prior peak) or momentum (at its all-time high). The label is computed from the signal's own history, never assigned by hand.
Every score carries HIGH, MED or LOW confidence, mapping what the system observed against what it estimated. A number without its confidence level is a number you can't trust.
How outcomes are measured
Each morning the Signal Vault emits phase-labeled pre-breakout alerts. From that moment the alert is frozen: its subject's view counts are re-read daily from the platform's own API and appended to a permanent outcome table. An alert with fewer than two follow-up observations shows as "measuring", never as a claimed win.
See the live prediction logWhat we refuse to do
- No fabricated numbers. If a data source is missing, the field shows as missing, never as a plausible default.
- No winners-only marketing. The prediction log at /receipts keeps every tracked alert, hits and misses alike.
- No claimed hit rate before the sample is statistically meaningful. The log fills in public until then.
- Features that aren't built yet are labeled as such on the features page. Nothing is sold before it exists.