ClickBaton
Analysis

Read the momentum, not the noise

Every tracked dimension scored against its own baseline — the full leaderboard behind the trends page and the market boards.

Site-wide: Declining -41.6%

18dimensions scored
0rising
0falling
30dprimary window

Market momentum

All markets ranked by z-score — how many standard deviations the current window sits from the baseline mean.

MarketDirectionChangeZMomentumConfidenceSample
GitHub Emerging +41.1% +0.81 +1.41 High 1,818
Research Papers Emerging +55.8% +0.79 +1.56 High 550
Product Hunt Stable -3.4% -0.11 +0.97 High 5,709
eBay Declining -100.0% -0.18 +0.00 High 2,814
Hacker News Declining -100.0% -0.49 +0.00 High 256
Stack Exchange Stable -17.8% -0.69 +0.82 High 3,296
News Pipeline Declining -40.2% -1.04 +0.60 High 8,279
App Store Declining -67.2% -1.61 +0.33 High 3,404
X (Twitter) Stable +0.0% — — Low 0
User Submitted Stable +0.0% — — Low 0
Reddit Stable +0.0% — — Low 0
npm Stable +0.0% — — Low 0
NewsAPI Stable +0.0% — — Low 0
Integration Signals Stable +0.0% — — Low 0
Customer Language Stable +0.0% — — Low 0
Company Registries Stable +0.0% — — Low 0
Business Evidence Stable +0.0% — — Low 0

Cross-market theme radar

What the classifier keeps tagging across all markets, most recent window.

Low 25,222 Clean 13,156 English 13,017 Neutral 10,375 Praise 4,596 Complaint 2,646 General Discussion 2,594 Announcement 2,125 Positive 1,803 Medium 835 Negative 595 Pain Point 577

Methodology

How these numbers are made — no black boxes.

Boards on this page are computed by the metrics engine on an hourly schedule. For every tracked dimension it compares the current window against a long-running baseline, expresses the shift as a percentage change and a z-score (how many standard deviations the current window sits from the baseline mean), and assigns a direction with a confidence grade that reflects sample size. The primary window shown here is 30d.

Dimensions with thin samples are downgraded to low confidence rather than hidden — an empty-looking board usually means thin data, not missing data. Newest snapshot per dimension wins; stale rows are never mixed in.

Turn analysis into evidence

The raw signals behind every board sit one sign-in away in the evidence layer.