Methodology & accuracy

Fortnite: the Epic Fortnite Data API (Ecosystem) — public, first-party, no auth. Peak CCU and unique players at up to 10-minute granularity; 7-day live retention (we archive permanently). Returns null below 5 unique players.
Roblox: Rolimons for discovery, then the public Roblox games + votes API for concurrent players, visits, favorites, and like ratio. Deeper metrics are owner-only — first-party connect is the only path to them.

Raw snapshots are rolled up into 5-minute, hourly, and daily aggregates. Dashboards read only pre-computed rollups. Values are measured, never modeled.

Trend score blends CCU velocity (week-over-week change) and acceleration (change in velocity), each z-scored against the game's own 28-day baseline. Lifecycle stage (emerging → accelerating → peaking → declining → dormant) follows transparent rules on those signals — each call shows its evidence.

Estimated monthly impressions ≈ average concurrent players × session/turnover assumptions. Media value = impressions × a reference CPM range (currently a documented $2–8 band; real cited CPMs pending). Suggested integration fees = media value × observed deal multiples, always shown as a range with the number of comparable deals.

A 0–100 score from transparent, evidenced factors: CCU volatility, scale, and unlicensed-IP graph links. Each factor's contribution is shown so a brand can defend the pick. The IP-risk methodology and specific IP determinations are under counsel review before public use.

Difference-in-differences: the campaign game's change from its 28-day baseline, minus the change of a matched control cohort (similar games that didn't run the campaign) over the same window. Reported with a 95% confidence interval — “not significant” is an honest, first-class outcome.

As first-party connections accumulate, we publish estimate error (MAPE) by metric and segment here, and improve models against ground truth. This section fills in as the calibration set grows.