Skip to main content
The columns and formatting on this page take effect in the Saturday, September 26, 2026 SFTP snapshot. See SFTP Schema Changes — September 26, 2026 for what’s changing.

Which reference do I need?

What this covers

Every column in the SFTP bulk feed, per dataset, in exact feed order. There are six datasets, listed below. The feed delivers six CSV datasets over SFTP, refreshed weekly.
BenefitFlow aggregates data from multiple public and proprietary sources. See Where does BenefitFlow’s data come from?.

How our exports are formatted

Entity Relationships

The six datasets connect through shared IDs. Use these keys to join across files.

Join keys

Key relationships

  • Employer to Policy is one-to-many. One employer can have multiple policies (Health, Dental, Life). The Policy file is the most granular view of coverage.
  • Employer to Broker shows two ways: the Employer file carries denormalized “Primary Broker” fields (the broker with the most business), while the Policy file shows the specific broker on each policy. These can differ.
  • Broker to Office to Contact is hierarchical. A broker firm has many offices, and each contact is associated with one office.
  • Employer to Employer Contact is one-to-many. Employer contacts are people at the employer (HR, benefits, executives), not at the brokerage.

Employer dataset

One row per employer. 147 columns.

Example row: Walmart

Columns without data are left blank. Modeled Premium is BenefitFlow’s estimate of full annual premium by product, so it is not comparable to Total Premiums (the reported, filed figure) and runs much higher for a self-funded employer like Walmart.

Employer Contact dataset

One row per contact. 163 columns.

Example row: a Walmart contact

Contact details are illustrative; the employer context is real.

Policy dataset

One row per policy or plan. 53 columns. The Modeled Premium columns estimate how this policy’s total premium splits across benefit lines. A single-line policy puts its full premium in that line’s column. A multi-line policy is split across its lines, and the split adds up to Total Premiums. The columns are blank when the policy’s lines aren’t among these 12, such as Legal or Wellness/EAP.

Example row: an NEA policy

Broker dataset

One row per broker firm. 14 columns.

Example row: Marsh McLennan

Broker Office dataset

One row per broker office location. 19 columns.

Example row: Marsh McLennan, Dallas

Broker Contact dataset

One row per contact at a broker. 37 columns.

Example row: a broker contact

Contact details are illustrative; the broker context is real.

Data Delivery Details

For connection setup and SFTP client configuration, see the Bulk Data Access Setup Guide.

Implementation Notes

Many fields contain commas within values (Line of Business, addresses). Use a proper CSV parser; do not split on commas.
EIN, NAICS Code, ZIP Code, and NPN ship as plain digits. Import them as text, not numbers, so leading zeros are preserved. EIN is formatted XX-XXXXXXX.
All BenefitFlow IDs (Employer ID, Broker ID, Broker Office ID, Closest Office ID, Primary Broker ID, Primary Broker Closest Office ID) are UUIDs (xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx).
IDs can change between releases. Each time you load an Employer, Broker, or Broker Office file, remap your stored IDs with the latest ID mapping file. This also covers IDs stored before June 2026, when Broker and Office IDs changed from 32-character hex hashes to UUIDs. See ID Mapping Files.
The Policy and Employer files include denormalized broker and office fields for convenience. The Broker and Broker Office files are the canonical source for broker-level data.
When no broker is identified, the broker name shows No Broker and the broker ID is left blank. Filter these out when analyzing broker relationships. (In snapshots before the June 2026 ID change, the Broker ID was instead the sentinel ffffffffffffffffffffffffffffffff; treat that legacy value the same as blank.)
Record Type, Lead Source, and BenefitFlow User hold the same value for every row in a file. They exist for system compatibility and can be ignored during analysis.
Two broker office locations appear on Employer and Policy records. Closest Office is the BenefitFlow-determined office nearest the employer. Reported Office is the office listed on the Form 5500 filing. These often differ.
Last modified on October 20, 2026