Acumix data is engineered from day one to clear a compliance officer's desk. Every panel is aggregated, point-in-time, and sourced from public information — no personal identifiers, no material non-public information, and full rights to license and redistribute. We complete standard alternative-data DDQs and ship a data dictionary, methodology and coverage report with every dataset.
Alternative data only helps if it survives your diligence process. Each Acumix dataset is designed around the questions an institutional buyer's compliance, legal and data-science teams actually ask — before a single row is delivered.
Every delivered field is an aggregate — counts, medians, means, ratios and indices computed across many observations. We never sell row-level records tied to an identifiable person.
We collect only information that is publicly observable to any ordinary user of a public source — published prices, listing counts, ratings, catalog attributes and dates. Nothing confidential, nothing insider-sourced.
The datasets are proprietary and continuously collected. We warrant the right to license and redistribute them to you, and the deliverable consists of factual data points — not copyrighted expressive content.
Every panel ships with a data dictionary, a written methodology, and coverage statistics. You can trace each aggregate back to its source category, geography and collection window — and reproduce it.
The single most important property for a backtest is knowing what you would have known, when. Every Acumix observation carries both the date the fact was true in the world and the date it entered our knowledge base. Records are append-only and never restated — so history you download today is identical to history you downloaded a year ago.
# one row from a point-in-time panel { "entity_id": "DHER.DE", "signal": "delivery_venue_count_idx", "geo": "DE", "value": 104.7, # index, base=100 "observation_date": "2025-11-30", # when the fact was true "knowledge_date": "2025-12-02", # when we knew it "vintage": "append_only", "restated": false } # backtest as-of any historical knowledge_date: # SELECT * WHERE knowledge_date <= as_of_date # → returns exactly what was knowable then.
Because knowledge_date reflects real collection latency (not the observation date), an as-of query never leaks information that wasn't yet available. Your backtest sees the data on the day a live strategy would have.
Entities that later delisted, merged, or stopped trading remain in the panel with their full history. We do not prune the universe to today's survivors, so cross-sectional studies stay unbiased.
Append-only, never-restated storage means a given (observation_date, knowledge_date) pair always returns the same value. Re-run last quarter's research and get byte-identical inputs.
We routinely complete standard alternative-data due-diligence questionnaires — including FISD-style data-sourcing questionnaires — and provide the supporting documentation your data-governance and compliance teams need to onboard a new provider. Everything below ships with every dataset, at no extra cost.
We fill out your alternative-data / FISD-style questionnaire covering sourcing, PII handling, MNPI, licensing, retention and change management.
Every field defined — name, meaning, units, aggregation rule, valid range and refresh cadence — so your analysts can model against it without guesswork.
How each signal is constructed and aggregated, the collection window, known gaps and limitations, and the mapping from raw facts to delivered indices.
Geographies, platforms, entity counts, history depth and cadence per dataset — plus a security master mapping each entity to its ticker.
Written representations and warranties: no PII in the deliverable, no MNPI, public sources only, and the right to license and redistribute.
A representative point-in-time sample so your team can validate schema, coverage and integrity before committing — reviewed under NDA on request.
Tell us the coverage you need and we'll return a completed DDQ, the data dictionary and methodology, coverage statistics, and a point-in-time evaluation sample — under NDA if you prefer.
Request Compliance Pack & Sample →