Architecture

Data flows one way, from source to platform

Raw files are kept exactly as downloaded, so every cleaned or adjusted number can be traced back and rebuilt. Only the Curated zone is published.

Five layers

From source to consumer

1Sources

Exchanges (HOSE, HNX, UPCoM), Yahoo Finance, index providers, central banks, IMF, World Bank, GSO, company reports, ESG and alternative data, internal files.

2Ingestion

Scheduled connectors (API, download, SFTP), one script and one catalogue entry per source, retries and time-outs.

3Storage zones

Raw, Clean and Curated: the data moves forward one zone at a time.

4Serving

Parquet and RDS files on SFTP, a REST API, the nxwdata R package, CSV downloads with a data dictionary.

5Consumers

IMS, FUNDMS, INDOBS, CCPR, CCPS, VTP / CCPE and the NXW Monitor.

Storage zones

Raw, Clean, Curated

Raw

As received, never edited, with date, source and checksum.

Clean

Validated, typed, deduplicated, gaps flagged.

Curated

Adjusted, mapped to NXW codes, versioned point in time. The only zone that is published.

Governance across every layer: QC rules · data quality log · audit trail · lineage · access roles · approval of corrections.

Data model

Ten core tables

All keyed by the NXW instrument code and a date. Every row carries its source, load time and validity period, which makes the data point in time.

TableContent
INSTRUMENTISIN, ticker, asset class, exchange, sector
ID_MAPVendor codes, valid from and to
PRICE_EODOHLC, volume, raw and adjusted close
PRICE_INTRADAYPrice and volume (later phase)
CORP_ACTIONType, ratio, dates, adjustment factor
SHARESOutstanding, free float, foreign limit
FXRate by pair and date
INDEX_EXTLevels and constituents by provider
FUNDAMENTALStatements by period and item
MACROSeries, unit, country, release date

Plus two control tables: SOURCE (feed catalogue, owner, schedule, last run) and DQ_LOG (every check, correction and approval).

Need a table?

Tell us which data your app reads today and we will map it to the DMS.

Contact us
UNDER CONSTRUCTIONDMS is being built: pages, data and features are added step by step.