Methodology

TrendWatch reads daily app-store chart snapshots from free public sources and turns them into rank history, momentum, and — once enough history exists to calibrate against — download estimates. Every number here has a stated limit; nothing is presented with more precision than the source supports.

Sources

Apple RSS v2

Public top-charts feed — rank only, no download counts, top 200 per feed/category/country.

iTunes Lookup + reviews

App metadata and a review-count sample — proxy for review velocity, not a full review corpus.

google-play-scraper

Unofficial Play Store scrape — install buckets(e.g. “1,000,000+”), not exact counts; used to calibrate the estimator, not shown directly.

Google Trends

Relative search interest, 0–100 per query — not absolute volume.

Wayback Machine

Historical chart snapshots used only to backfill before this site's own collection began — sparser and less regular than live collection.

Why estimates are bands, not numbers

No public source gives exact download counts. The estimator fits a power-law rank-to-downloads curve per platform/country/category, calibrated against Play Store install buckets and (for iOS, which has no install buckets) a review-velocity transfer. Every estimate ships as a low – mid – high band with a method_version, and a stratum without enough calibration data is marked lower confidence rather than hidden or guessed.

Honesty rules

• Charts show bands, never fabricated point values.
• Gaps in collection render as gaps — a dashed bridge with a tooltip stating the cause, never smoothed or interpolated away.
• Rank axes are always inverted (rank 1 at the top).
• A backtest gates the estimator: it must place at least 70% of held-out install-bucket crossings inside its band, or the bands widen until it does.