How we work

A statistic is useful only when you can tell what it counts, who it covers and where it came from.

Start with the source

Our U.S. coverage includes NTIA participation aggregates, NCI HINTS adult visit frequency, limited Pew platform-adoption excerpts, CDC high-school frequency and one Gallup teen time estimate. Eurostat provides separate European comparisons; original company disclosures support supplementary worldwide platform briefs. Every important number keeps its geography, denominator, period and source. Editorial excerpts are distinguished from imported datasets; none is a live user counter.

Preserve the measure

We distinguish people, accounts and devices, and separate survey adoption from monthly active users, daily active users and advertising reach. Company family-of-apps totals cannot stand in for an individual platform.

Survey population and question wording are part of a measure. NTIA internet users aged 15+, Pew adults aged 18+, HINTS civilian noninstitutionalized adults aged 18+, CDC students in grades 9–12 and Gallup teenagers aged 13–19 remain distinct. Frequency of visits does not establish hours spent. European participation among individuals aged 16–74 is another separate series. A “reviewed” date never replaces the original reference period.

Import, check, release

Dataset imports enter staging. Their validators check the relevant dimensions, missing values, ranges, source flags, evidence hashes and method changes. A review record is required before publication. NTIA and Eurostat import published aggregates. The HINTS process reads one public-use survey item and its supplied weights, writes only national aggregates and retains evidence for the calculation; respondent records are not published.

Published files are immutable and versioned. A failed import leaves the last good release available. The site builds from a checked-in snapshot, so it does not depend on a live source API for each page view.

Make calculations inspectable

Derived statistics preserve input observation identifiers, an operation and its result. Time comparisons require consistent metrics, populations, geographies and methods, and must not cross a source break. Missing or confidential values cannot enter a calculation.

We use percentage points for differences between percentages. The HINTS point estimates follow the supplied codebook; their standard errors are SocialUsage calculations using NCI’s 50 replicate weights and documented jackknife method. NTIA standard errors, Gallup’s standard error and CDC confidence intervals are source-reported. We label these different origins rather than treating every uncertainty measure as our own analysis.

CDC’s matching 2023–2025 hourly-use comparison is a supported U.S. example: its published test found no statistically significant change. We attribute that test to CDC. Our single-wave NTIA, HINTS and Pew selections cannot be joined into a trend. Descriptive differences alone establish neither significance nor causation.

Be honest about review

This first edition was assembled and checked with an automated assistant. Its evidence was cross-checked against original sources; it has not received independent human editorial sign-off. No expert author or reviewer is implied. Publication dates describe editions; data periods describe the underlying evidence.

Corrections and source revisions

A correction creates a new release and an entry in the revision log. Original versions remain citable. A source revision, a change in methods and an error in our presentation are recorded separately.

Report a possible error to [email protected]. Include the page or release URL, the claim in question and any supporting source. Avoid including personal survey records or sensitive information.