Our approach
A process, not a black box.
The credible thing to publish is not a claim of magic but a description of method. Here is how a price becomes an analysis.
Ingest
Time-stamped prices for one asset are pulled from every available venue and normalized to a common format and clock. Data-quality checks run first — a signal built on a bad tick is worse than no signal.
Reconstruct
We rebuild the complete price path per source, preserving every genuine change while discarding redundant flat ticks, so the full movement is auditable.
Compare
With sources aligned, we measure lead-lag (who moves first), divergence (who disagrees and by how much), and convergence (how gaps close) — the raw material of price discovery.
Filter
Candidate moves face a battery of checks — spike-and-revert detection, liquidity floors, stale-quote and out-of-band filters, a structural-basis adjustment, and a cross-source consensus test — designed to reject artifacts before they're mistaken for information.
Validate
Each read is scored against the subsequent market outcome. Execution-quality thinking from quantitative finance — implementation shortfall and slippage — informs how we judge whether a detected edge was real.
Engineering
The whole pipeline is automated end-to-end — from ingestion through a database of reconstructed price histories to reporting — with per-item analysis running in seconds. It is built and operated by the person who designed the methodology; the tooling serves the method, not the other way around.