Your portfolio, attributed to the calls that explain it.
Compass cross-references your holdings against the universe of research it has already done. Concentration risk and factor exposure get attributed to the actual transcripts and filings — not to opaque scores.
For every position — top management commentary, latest guidance, restated KPIs, prior-quarter trajectory, peer setpoints. Click any line to see the source.
Sector and factor exposure built from the underlying research, not from a static classifier. Concentration risk attributed to the names and the statements driving it.
"What if X happens to position Y?" Chat-driven scenario testing grounded in the same citation-bound corpus the rest of Compass runs on.
Risk officers and PMs see the same evidence.
Concentration scores have always been computable. Compass attributes them: which underlying names drove the score, what those names said in the calls that produced the attribution, when that view was updated.
Risk officers stop relying on summary outputs. PMs stop defending positions against opaque flags. Everyone reads the same citations.
- · Holdings-level attribution to source documents.
- · Sector / factor exposure with full citation drill-down.
- · Versioned: every view of the portfolio is replayable to any past date.
- · Exportable to your risk system in your existing schema.
The portfolio is a document corpus. Ask it questions.
Every Compass scenario answer is grounded in the same corpus the rest of the product runs on — your filings, your calls, your prior research notes. Hallucination-bounded by design.
Same audit log. Different views.
Compass entitles risk officers, PMs, and analysts differently by document and by view. Everyone reads the same evidence; not everyone sees the same trades.
- Entitlements-aware. Holdings visibility scoped per role. Risk sees positions + drivers; analysts see coverage names + research.
- Stable identifiers. Every Compass output carries an ID. Replay it later; see the exact corpus state, prompt, and source set.
- Export to your stack. Outputs serialise to your existing risk schema, your CMS, your research portal — pick the export.
- No model training. Portfolio data never trains any model. Contractual, not just implementation.