Methodology

InsideREU aggregates two distinct EU regulator datasets: managers' transactions(MAR Article 19 / PDMR — directors and senior executives buying or selling their own employer's stock) and substantial holdings (Transparency Directive — anyone crossing a 5% / 10% / 25% / 50% / 75% ownership threshold). Every row in our database links back to the original regulator publication.

The coverage matrix below is the live state, refreshed hourly from the same internal view that powers our admin dashboard. We don't edit it, we don't hide rows, and we don't round up. If a country is BRONZE today, we say so today.

Coverage matrix — Filings (PDMR / Article 19)

Coverage matrix — Substantial holdings

Tier definitions

• GOLD— last filing < 5 days old, ≥95% of rows have ISIN, ≥95% have the secondary critical field (value in EUR for filings, new ownership % for holdings). Suitable for backtesting and time-sensitive signal generation.
• SILVER— last filing < 10 days old, ≥80% ISIN, ≥70% secondary field. Suitable for thematic research and longer-horizon analysis. Not recommended for intraday signal use.
• BRONZE— anything else. We do not lead with these in marketing claims; coverage exists but lags or is structurally incomplete (often because the regulator publishes via PDF rather than CSV/JSON, see “Signal-only scrapers” below).

How we ingest

Single nightly cron at 02:30 UTC. ~25 country-specific scrapers run in parallel with a 90-second per-source deadline; the orchestrator upserts each source's output as it returns, so partial completion still leaves earlier rows durable. The whole pipeline stays inside Vercel's 5-minute function ceiling. Each filing is keyed on a deterministic id (regulator-specific reference + insider id + transaction date + price + value), which makes re-scraping idempotent.

Source URLs and raw responses are preserved in the row's source_url and raw_data fields so any downstream researcher can reproduce the parse from the original regulator filing.

Signal-only scrapers

Several countries publish PDMR notifications as PDF attachments rather than structured CSV/JSON: AT (OeKB), ES (CNMV), FR/NL/BE/PT (Euronext consolidated), FI/DK/IS/EE/LV/LT (Nasdaq Nordic news), plus most TD-filings holdings sources. For these we emit one row per regulator filing with the metadata we can extract (insider name, company, transaction date, source URL) but leave shares and value_eur as NULL. The filing existence is a signal even without the dollar amount; full per-row PDF parsing is on the roadmap.

Known caveats

• DE BaFin source-side delay.As of 2026-05-04, BaFin's DealingsInfo CSV is publishing filings 6–7 days after notification. Our scraper is healthy and runs nightly; the lag is at the regulator. We're monitoring; if it doesn't resolve by mid-May we'll add a secondary German source. Until then DE remains SILVER on freshness alone (completeness is unaffected).
• EE / HR scrapers dormant.Estonia and Croatia regulators haven't published a relevant filing in our covered window. Volume in these markets is genuinely tiny, but if the upstream source format changed we'd catch it on the data-health dashboard.
• Holdings new_pct gap.Most TD-holdings sources don't surface the post-trade ownership percentage in structured form, so the ability to filter “everyone who crossed from <5% to ≥10% in March” is currently limited to DE plus partial CH. Expanding this is the next holdings priority.
• Value normalization.Where regulators report transaction value in local currency (CHF, GBP, NOK, etc.) we convert to EUR at ingestion using a fixed approximate rate per currency (for example SEK 0.087, CHF 1.04, DKK 0.134, GBP 1.17), not the ECB rate on the transaction date. That is adequate for ranking and screening but not for precise P&L. Where the price itself is converted (Sweden, Switzerland) the original is preserved in raw_data.

What we don't do (yet)

• Historical backfill. Most country scrapers reach back ~12 months from go-live (2026-04-15). Earlier periods are not in our DB. If a customer needs 5+ year history for backtesting we can source it from a data broker on a per-deal basis.
• Point-in-time corrections.Filings are upserted on the deterministic id. We don't currently version the row, so if a regulator corrects a filing after the fact we overwrite. Versioning is on the roadmap for any customer that needs it for backtest reproducibility.
• Webhook delivery. Filings flow into a Postgres table that we can expose as a read-only credential to enterprise customers. Push-style webhook delivery is on the roadmap and will be built when a customer requires it.

Field reference

Each filings row has the schema documented at /developers/schema. The public API surface is at /developers.

Last refreshed from data-health view at page render. Updates every hour. For real-time admin view see /admin/data-health.