Deterministic Trend Score
Scores are calculated in code from stored observations. An LLM never invents the number. Missing sources are omitted and remaining weights renormalise — they are not filled with zeros.
Methodology
How we score and decide what to publish
The intelligence model is conservative by design. This page describes the real pipeline — not aspirational coverage.
Scoring and gates
Core rules
Entity quality
Quality assessment classifies entities (useful, noisy, insufficient data, and related statuses) using history, freshness, and evidence strength. Useful does not automatically mean public.
Public eligibility
Publication requires the quality gate and the current eligibility rules (including strength, status, and freshness). The gate is authoritative: the website and API never display non-eligible entities.
AI interpretation
Explanations summarise the same evidence package used for scoring. They must not invent statistics, sales claims, or certainty. A pending explanation does not block an eligible entity from appearing.
End to end
From observation to public page
01 · Collect
Automated collectors capture public observations from Google Trends, YouTube, and eBay on a recurring cadence.
02 · Resolve
Observations are linked to entities by exact or alias matching. Similar-looking names are not merged automatically.
03 · Score
A deterministic Trend Score weighs recency, frequency, coverage, and movement from the sources that actually exist for an entity.
04 · Quality-check
History depth, freshness, and evidence strength decide whether an entity is fit to publish.
05 · Explain
An optional AI interpretation is generated from the same evidence. Missing explanations do not block publication.
06 · Publish
Only entities that pass the public eligibility gate reach the website, API, and sitemap.
Caveats
Read this before treating a score as truth
Currently covering selected markets and data sources, with coverage expanding over time.
eBay coverage is Singapore listings only. Absence of Shopee, Lazada, TikTok Shop, or Amazon is intentional — those collectors are not live.
Public pages never invent trend cards, scores, or charts to fill empty space.
Trend Score measures observed attention and movement. It is not a forecast, demand estimate, or sales figure.