How we read the evidence

The strength of a number depends on what went into it. We keep the sample, the measurement and the uncertainty visible.

Two ways to view the evidence

Community reported includes usable numerical length or girth changes even when a specific measurement convention is not stated. Measurements and routine context can vary. This is the default view.

Strict subset follows the preserved explicit-measurement criteria and keeps named measurement conventions separate. It contains 114 eligible outcome records. This preserved subset is not a complete strict-criteria assessment of every newly added report. Explicit measurements do not by themselves make a self-report clinical evidence.

Accounts, reports and averages

Each mean uses the release's selected observation per known account, dimension and cohort. Every usable sample is displayed. One-account values are individual reported changes; samples of two to four display summaries labeled as small samples. Five accounts does not establish reliability. Length and girth can have different denominators. The 718 eligible outcome records include earlier updates; 691 are selected across overlapping groups. We do not add every update into an average.

An account is a proxy for a person. One account can contribute to multiple devices or methods, so groups cannot be added together. Source-post counts show mentions and are separate from eligible contributors.

Mixed routines remain mixed

All recorded routines include mixed routines and records where other methods are not specified. Reported exclusive use requires an explicit sole-use statement. A change from a routine containing several methods is not credited as an independent effect of every device.

Time and rates

Timing summaries use only selected accounts with positive numerical elapsed duration. Each calendar rate is the reported change divided by that account's elapsed months, using 30.4375 days per month. The displayed rate is the median of those individual rates. Periods and rates are shown whenever usable timing exists, with their own denominator. A single timed account is labeled as an individual report.

Calendar time does not measure consistent active use. Breaks, initial changes in measurement or erection quality, regaining earlier size and temporary changes are not reliably controlled. The full-sample mean and timed-subset median period describe different samples. Do not multiply a rate to forecast a future result.

What the source cannot answer

The source is a collection of community “gains/success” posts. Selection and reporting bias prevent estimating success probabilities, typical outcomes among all users, comparative efficacy, safety or permanence. These are self-reported size changes, not verified new tissue growth.

Analysis scope

The original inventory covers 4,696 gains/success posts. This release also incorporates selected follow-up replies and a bounded review of additional comments from the same subreddit. The comment archive was screened locally to prioritize review; it was not all semantically reviewed. Remaining cases and queues are not exhausted. Source-post mention counts retain their original inventory scope.

This release incorporates 3,675 newly reviewed comment packets, bringing the recorded total to 9,291 (5,616 previously reviewed). These are review-progress counts, not eligible reports or a representative population denominator. Extraction is currently paused.

The sample remains selective and self-reported. Adding zero and negative reports does not create a representative population denominator or a clinical comparison.

Traceable, versioned data

Every detail page links the selected reports to their original Reddit sources. Anonymous report labels are stable within this release; source links may identify the original poster. We do not show personal photos or reconstruct missing before-and-after measurements.

Evidence release: gain-atlas-mvp-023, 2026-10-04. The original strict analysis and expanded release remain preserved. Website summaries are calculated from the release’s selected observations, including samples below five; they are not edited by hand.

Reading comparisons and tools

Comparison tables default to sample size, not a “best device” claim. The finder has separate product and evidence modes. Product mode filters verified formats and methods, with alphabetical results. Evidence mode sorts eligible devices by sample coverage for your selected goal. Product research is separate from the evidence release; it does not establish personal or medical suitability. The measurement calculator performs arithmetic on your own inputs and does not apply a community average as a prediction.