How we measure trends
Transparent observations instead of invented search volumes.
Discovery
Several times daily, the collector requests recent popular public Roblox Marketplace results across multiple categories. Discovery samples paid items and excludes official Roblox items from keyword extraction. Repeated keywords and phrases are extracted from item names. These are automatically discovered item topics, not verified user search queries. The experimental query endpoint is off by default; any permitted results are attributed separately.
Interest index
The 0–100 index measures weighted topic presence in observed recent rankings. Weight for zero-based position r: 1 / (1 + r / 8). A topic's summed weights are divided by the number of unique observed items and multiplied by 100. This is a relative estimate. Changes in the returned sample and ordering affect it.
Growth and opportunity
Growth compares real observations near the start and end of a selected period. Without a suitable baseline, a dash appears. Opportunity requires observed daily growth and a nonempty, uncapped competitor sample: capped growth (0–200%) / 2 × (1 − sample size / 31). It is a research heuristic, not an earnings forecast. No score is calculated when the API indicates another results page.
Competition and sorting
Competitors are items observed on the first page of a topic query, not the entire market. A ≥ sign indicates further pages. Default ordering uses Roblox's provided recent ranking with a time filter. Rankings, favorites and purchases are separate metrics. Growth and publication-date sorting apply within the current sample.
History, freshness and limits
Target cycles are every 2 hours for items, 6 hours for topics and daily for scores, with bounded load and retries. Larger watchlists rotate through a queue; each observation displays freshness. Item history is retained for 90 days, topic history for one year. A favorites-only update does not refresh price freshness. Momentum is the difference between consecutive daily growth rates in percentage points. A surge or unusual decline means the latest hourly change differs by at least 3 standard deviations from preceding changes within 14 days. At least 8 usable intervals are required; gaps exceeding 9 hours are excluded, and a gap in the latest interval makes the signal unavailable. These heuristics are sensitive to sample changes. Gaps are not filled. Exact sales appear only when the official response actually supplies purchaseCount. A public 3D model is not assumed available; official 2D previews are used.