Coverage, what a finding means, and the limits: answered directly, including where the answer is no.
European wholesale power: 38 bidding zones across 24 countries (21 EU member states, plus Great Britain, Norway and Switzerland). Day-ahead and the three intraday auctions come from exchange and regulatory feeds, continuous intraday across 23 market areas, and Great Britain from its own market feeds. Gas, oil, carbon and non-European power markets are not covered.
No. Every finding is an indicator for human review, never a determination. A price the reconstruction cannot explain most often has an explanation the model cannot observe: an unpublished outage, a network constraint, a hedge being unwound. Findings carry the model's calibrated out-of-sample error, and a gap inside that error is reported as inside it.
Accuracy is reported per zone rather than as a single headline figure, because it varies enormously. Reconstruction is weakest in hydro-dominated zones, where marginal price reflects a storage scheduling decision rather than a fuel cost; in scarcity hours, where price is set by willingness to pay; and in congested hours, because allocated cross-zonal capacity is not yet ingested. Calibration is walk-forward, so the reported error is out-of-sample.
Not today. It surveils the published market, not your book. Post-trade controls over a firm's own order and trade feed (orders per second, price and volume collars, T+1 verification) are on the roadmap and are not built. That means order-book typologies such as layering, spoofing and wash trades are out of scope.
Because they are not published. Orders are aggregated into 13 clusters, and zones outside those clusters, among them Switzerland, Czech Republic, Spain, Hungary, Portugal, Slovenia and Slovakia, have no public order-book view at all. Great Britain has none either: it clears through its own day-ahead auction and a GB-exclusive half-hourly product for which aggregated curves are not published.
It depends on the stream, and the constraint is upstream rather than ours. Aggregated bid curves are served for the current and next delivery day only (ask for the day before and the API returns 401), so history exists only from the point daily capture began. The regulatory feeds serve much wider windows and can be backfilled.
Nothing. Every product, including auction surveillance, bid curves, the auction simulator, balancing surveillance and the UMM assessment, is free. The project is sustained by the motivation and voluntary contribution of the people who use it rather than a paid tier; supporting it is welcome and never a condition of access.
Yes. Signing up creates an account, not access. An admin reviews every new sign-up and grants it the specific pages it needs before the console opens, the same review that keeps a multi-tenant tool from opening itself to anyone who finds the sign-up form.
No. It produces evidence that informs whether a report is warranted. Submitting a suspicious transaction or order report, and deciding to, remain the firm's process.
The application database is managed Postgres in an EU region, and raw upstream payloads are held in Cloudflare R2 object storage. Organisation-scoped data is isolated by Postgres row-level security. Market data is shared reference data, because it is the same public facts for every tenant.