DenyTranDFW/SEC-Financial-Statements-And-Notes-Dataset

sec_fsnds

A ready-to-run SQL Server database of the SEC Financial Statement and Notes Data Sets (FSNDS), packaged with a Docker image, installer script, stored procedures, and recommended indexes.

This is the full historical FSNDS corpus (all quarterly dumps the SEC has published) loaded into a single, normalized SQL Server database. The data file is ~300 GB on disk.

---

What's in this package

sec_fsnds_package/
├── README.md                       -- this file
├── Dockerfile                      -- SQL Server 2022 + Full-Text Search
├── docker-compose.yaml             -- parameterized compose file
├── sec_fsnds_setup.cmd             -- Windows installer (interactive)
└── sql/
    ├── schema/
    │   └── sec_fsnds_schema.md     -- table / column / index reference
    ├── indexes/
    │   └── sec_fsnds_recommended_indexes.sql
    ├── sps/
    │   ├── usp_fsnds_get_statement.sql
    │   ├── usp_fsnds_get_statement_series_combined.sql
    │   ├── usp_fsnds_get_statement_series_quarterly.sql
    │   ├── usp_sec_find_filings_by_cik.sql
    │   ├── usp_sec_find_filings_by_name.sql
    │   ├── usp_sec_find_tag.sql
    │   ├── usp_sec_get_filing_metadata.sql
    │   ├── usp_sec_get_full_filing.sql
    │   ├── usp_sec_get_segment_values.sql
    │   ├── usp_sec_get_tag_history.sql
    │   ├── usp_sec_get_tag_usage.sql
    │   ├── usp_sec_get_text_facts.sql
    │   ├── usp_sec_list_peers.sql
    │   └── usp_sec_peer_analysis.sql
    └── validation/
        └── usp_sec_validate_statement_integrity.sql

---

Requirements

• Docker Desktop (Windows / macOS / Linux) or Docker Engine + Compose v2
• ~350 GB free disk space on the drive that will hold the database files
• At least 4 GB RAM available to the container (8+ GB recommended)
• curl available on the host (Windows 10 / 11 ship with it in System32)
• A SQL Server client: sqlcmd, Azure Data Studio, or SSMS

---

Quick start (Windows)

1. Create an empty folder somewhere, e.g. C:\sec_fsnds.

2. Copy sec_fsnds_setup.cmd into that folder and run it.

3. Answer the prompts:

   • SA password for the database
   • Host port (default 1434; change it if 1434 is in use)
   • Container name (default mssql-sec-fsnds)
   • Host data directory (must exist and have 350 GB free)
   • SQL Server edition (Developer, Express, Standard, Enterprise)

4. The script downloads Dockerfile, docker-compose.yaml, sec_fsnds.mdf, and sec_fsnds_log.ldf from Hugging Face. The .mdf download is the large one and supports resume via curl -C - if interrupted.

5. When downloads finish, run:

   docker compose up -d --build
   

   from the same folder.

6. Connect with sqlcmd or your favorite client:

   sqlcmd -S localhost,1434 -U sa -P "<your password>" -d sec_fsnds -Q "SELECT TOP 5 adsh, cik, name FROM dbo.sub ORDER BY filed DESC"
   

---

Quick start (macOS / Linux)

The .cmd installer is Windows-only, but the docker-compose.yaml is portable. Steps:

1. Download Dockerfile, docker-compose.yaml, sec_fsnds.mdf, sec_fsnds_log.ldf from the Hugging Face repo (same URLs the installer uses, see sec_fsnds_setup.cmd).

2. Create a .env file in the same folder:

   ACCEPT_EULA=Y
   MSSQL_SA_PASSWORD=<pick a strong password>
   MSSQL_PID=Developer
   HOST_PORT=1434
   CONTAINER_NAME=mssql-sec-fsnds
   HOST_DATA_DIR=/absolute/path/to/data/folder
   

3. Put the .mdf and .ldf files in HOST_DATA_DIR.

4. docker compose up -d --build

---

After first start

On first start, SQL Server attaches sec_fsnds.mdf and sec_fsnds_log.ldf automatically via the pre-populated master database inside the container image. Give it a minute to warm up, then verify:

SELECT name, state_desc FROM sys.databases WHERE name = 'sec_fsnds';

You should see sec_fsnds in state ONLINE.

Install the stored procedures

From the package folder:

sqlcmd -S localhost,1434 -U sa -P "<pw>" -d sec_fsnds -i sql/sps/usp_fsnds_get_statement.sql
sqlcmd -S localhost,1434 -U sa -P "<pw>" -d sec_fsnds -i sql/sps/usp_fsnds_get_statement_series_combined.sql
sqlcmd -S localhost,1434 -U sa -P "<pw>" -d sec_fsnds -i sql/sps/usp_fsnds_get_statement_series_quarterly.sql
sqlcmd -S localhost,1434 -U sa -P "<pw>" -d sec_fsnds -i sql/sps/usp_sec_find_filings_by_cik.sql
sqlcmd -S localhost,1434 -U sa -P "<pw>" -d sec_fsnds -i sql/sps/usp_sec_find_filings_by_name.sql
sqlcmd -S localhost,1434 -U sa -P "<pw>" -d sec_fsnds -i sql/sps/usp_sec_find_tag.sql
sqlcmd -S localhost,1434 -U sa -P "<pw>" -d sec_fsnds -i sql/sps/usp_sec_get_filing_metadata.sql
sqlcmd -S localhost,1434 -U sa -P "<pw>" -d sec_fsnds -i sql/sps/usp_sec_get_full_filing.sql
sqlcmd -S localhost,1434 -U sa -P "<pw>" -d sec_fsnds -i sql/sps/usp_sec_get_segment_values.sql
sqlcmd -S localhost,1434 -U sa -P "<pw>" -d sec_fsnds -i sql/sps/usp_sec_get_tag_history.sql
sqlcmd -S localhost,1434 -U sa -P "<pw>" -d sec_fsnds -i sql/sps/usp_sec_get_tag_usage.sql
sqlcmd -S localhost,1434 -U sa -P "<pw>" -d sec_fsnds -i sql/sps/usp_sec_get_text_facts.sql
sqlcmd -S localhost,1434 -U sa -P "<pw>" -d sec_fsnds -i sql/sps/usp_sec_list_peers.sql
sqlcmd -S localhost,1434 -U sa -P "<pw>" -d sec_fsnds -i sql/sps/usp_sec_peer_analysis.sql
sqlcmd -S localhost,1434 -U sa -P "<pw>" -d sec_fsnds -i sql/validation/usp_sec_validate_statement_integrity.sql

Each file is idempotent (IF OBJECT_ID ... DROP, then CREATE).

Install recommended indexes (optional but recommended)

sqlcmd -S localhost,1434 -U sa -P "<pw>" -d sec_fsnds -i sql/indexes/sec_fsnds_recommended_indexes.sql

Read the header comment in that file first. Two of the indexes are large (several GB) and take 10-45 minutes to build. Everything in the sql/sps/ set works without them, just slower for certain queries.

---

What the stored procedures do

Statement retrieval

SP	What it returns
usp_fsnds_get_statement	Single BS/IS/CF for one filing, optional comparative
usp_fsnds_get_statement_series_combined	Wide pivot across 10-Ks and 10-Qs, raw as-reported
usp_fsnds_get_statement_series_quarterly	Wide pivot with Q4 derived (10-K minus Q1+Q2+Q3)

Filing discovery

SP	What it returns
usp_sec_find_filings_by_cik	All filings by CIK, filter by form / date
usp_sec_find_filings_by_name	Companies whose name matches a LIKE pattern
usp_sec_get_filing_metadata	sub row + pre/num/txt row counts for one filing
usp_sec_get_full_filing	Every line of every statement in a filing, stacked

Tag exploration

SP	What it returns
usp_sec_find_tag	Tag name search, optionally including tag text
usp_sec_get_tag_usage	Global usage stats for a tag
usp_sec_get_tag_history	Time series of one tag for one company

Peer analysis

SP	What it returns
usp_sec_list_peers	Other CIKs in the same SIC code
usp_sec_peer_analysis	One tag across all peers for a given fiscal period

Segments & text

SP	What it returns
usp_sec_get_segment_values	Dimensional (dimn > 0) numeric rows for a filing
usp_sec_get_text_facts	Narrative txt rows for a filing (needs IX_txt_sub)

Validation

SP	What it returns
usp_sec_validate_statement_integrity	BS / CF / IS sanity check over a CIK slice

---

Important notes

dim table is stripped. sec_fsnds.dim holds only dimID — the axis/member text (e.g. us-gaap:StatementBusinessSegmentsAxis / msft:ProductivityAndBusinessProcessesMember) is not stored. usp_sec_get_segment_values will tell you which rows are segmented and their dimID, but cannot decode what each dimID actually means. To add that, load the original SEC dim.tsv into a separate dim_text table keyed by dimID.

Collation is case-sensitive. Assets and assets are different tag names. Parameters passed to SPs respect case.

txt is huge. 48M rows / ~100 GB with only PK(txtID). Without IX_txt_sub from the recommended-indexes script, usp_sec_get_text_facts will do a full table scan on every call.

---

License / attribution

The underlying data is public-domain SEC submissions. The SEC publishes the quarterly FSNDS dumps at https://www.sec.gov/dera/data/financial-statement-and-notes-data-set.

The schema design, stored procedures, Dockerfile, installer, and this README are provided as-is, no warranty. Use at your own risk.

Follow SEC fair-access guidelines when redistributing or querying this data in volume.