Measuring observable market pressure.

Price methodology

We report the median and interquartile range (P25–P75), never an inferred price. Bulk pricing remains separate from single-account pricing. Original currency and normalized USD values are retained. Asking price is not necessarily realized transaction price.

Normalization

Every disclosed price is stored in its published currency and converted to USD using a dated reference rate table, so conversions are reproducible and can be re-derived later. A per-account unit price is computed by dividing the normalized total by the observed quantity, which keeps bulk lots comparable with single-account offers without merging them. When no reference rate exists for a currency, the observation is recorded without a normalized value rather than guessing one.

Automatic index updates

Adding, editing, verifying or removing an observation immediately recomputes that company’s component scores, FAM Index, confidence, observation counts and price percentiles from the verified, non-duplicate evidence, using the weights stored in the active methodology version. Companies with fewer than three verified observations report insufficient data instead of a score, and a snapshot is labeled demo only when every observation behind it is demo.

Projections

The Outlook page reports projected price ranges. These are model output, not observations, and they are never written back into the evidence record. Model p0.1 fits a single trend to the natural logarithm of disclosed single-account unit prices, estimating that trend only from platforms with disclosed prices at two or more dates, so price differences between platforms are never read as a trend. Each platform keeps its own observed level. Published intervals are 80% projection intervals combining the pooled residual spread, the level uncertainty of that platform and the standard error of the trend, with a deliberate floor on the residual spread so a handful of prices cannot look precise. A platform with no disclosed price is reported as insufficient data rather than estimated from its peers, and a platform with a single price is flagged as very low confidence. Pressure projections use dated FAM Index snapshots written at each rescoring; no history is back-filled to lengthen a series. The model is deterministic and versioned: the same evidence always yields the same range.

Historical price curve

The global curve is the real observed history first and a model second. Disclosed single-account prices are grouped into calendar years and reported as the median of that year; a year with fewer than two disclosed prices is reported as insufficient data and left as a gap. Nothing is interpolated, carried forward or averaged across years. A quarterly resolution is available for denser periods, on the same rules with a higher minimum.

Each platform also carries its own curve, drawn only from its own dated disclosed prices — never from other platforms or from the fitted market trend. Where a platform has a single disclosed price, that point is shown alone and no trend is claimed. On the outlook chart the solid line is this observed yearly history; the fitted projection (model p0.3) starts at today and runs forward only, so the past is never redrawn as a smoothed model line.

Each platform with a disclosed price also carries its own projected curve: its level comes only from its own dated prices, while the annual drift is the pooled trend estimated from platforms holding two or more dated prices — price differences between platforms are never read as a trend. Where a platform has one disclosed price only, the level is that single figure and the drift is borrowed, which is stated on the chart. Every curve is refitted whenever evidence is filed or verified, so platform curves move with the record rather than from a fixed baseline. Projections stay model output: they are never written into observations, metrics, or the feed.

Threat Board

The Threat Board reports three derived scores per platform under scoring version t0.1, recalculated daily at 06:45 UTC and stored as a dated snapshot so any day’s figure can be re-derived. Motive strength counts the separately sourced abuse objectives documented for that platform (14 points each), the variety of observed activity types (7 points each, capped at 21) and the number of independent sources (3 points each, capped at 15), capped at 100. Price risk scales the platform’s disclosed single-account median against the highest median observed across the watchlist (65 points), adds depth of priced evidence (7 points per disclosed price, capped at 20) and freshness of the latest price (up to 15 points, decaying over 180-day steps). A platform with no disclosed price has no price risk score; nothing is imputed. Threat level weights the FAM Index at 0.5, motive strength at 0.3 and price risk at 0.2, with the weights renormalised over whichever components have evidence, so a missing price is not scored as zero risk. A platform with fewer than three verified observations is not scored. All inputs are published beside each row, and none of these scores measure fraud on the underlying platform.

Scenarios

The Scenarios page compares hypothetical futures. Scenario model s0.1 takes the fitted projection as its only anchor and re-draws the median path under a different assumed annual price drift, or a different annual change in the FAM Index. Nothing else changes: platform price levels stay as observed, no new evidence is created, and scenario output is never written into the observation record, the metrics or the feed. Scenario lines carry no uncertainty interval of their own because they are assumptions rather than estimates; the shaded area shown behind them is the baseline projection interval, widened with the square root of the horizon. Both scenario parameters are displayed on the page, so any line can be re-derived by hand. Each posture also lists the documented abuse motives it acts on, drawn from the cited public sources on the company pages and grouped by whether the motive depends on registration and identity friction or on resale channels and stolen-credential supply. That grouping is an editorial classification of the evidence, not a claim made by any source, and listing a motive under a posture does not assert the posture would defeat it. A rising asking price may indicate that defences are constraining supply and a falling price may indicate the opposite, but the direction is an assumption of the scenario, never a measurement of a platform’s fraud rate.

Confidence

Confidence falls when evidence is sparse, stale, duplicated, ambiguous, or based on secondary reporting. Low-confidence findings remain visibly labeled rather than being presented as certainty.

Versioning & reproducibility

Each metric snapshot references a methodology version and public component scores. Changes to weights or definitions create a new version, preserving the interpretation of historical scores.