Writing & MediaFree
Government Database Request With Field List
Send the agency's data dictionary and prior releases, and get a request that names every table, field and export format, with the burden objection answered.
River reads whatever the agency has already published about its own system, then writes the request in that vocabulary: named tables, named fields, a named delimiter, a named encoding, and a stated date range. Anything the reporter's question does not need gets left out on purpose, and every objection the agency has raised before gets answered in the letter rather than in an appeal six months later. The output is a request and a field-level sheet holding the justification for each column.
Every page ranking for this query gives you a sample FOIA letter with a paragraph describing records in prose. Please provide all records relating to inspections, in electronic format if available. That phrasing is what a burden refusal is built on, because a description no one can bound is a search no one can scope, and an agency answering it gets to choose the interpretation that is hardest to produce. None of the templates name a field.
Built for data reporters and researchers asking a public body for its actual database, not a report about it. File it before a request is denied, since the field list is what a refusal has to answer to. Once the file arrives, dataset cleaning and methodology preserves the original and logs every transformation, data story analysis tests whether the finding survives a different reasonable choice, and evidence chronology records how each file was obtained. If the response is documents instead, document set entity extraction turns the PDFs back into cited rows.
The format you ask for decides whether your request is burdensome
Take an invented request to a federal inspections program, built to a real shape. The records in scope are 18,447 inspection records. Asked for as documents, at the program's own average of six pages per report, that is 110,682 pages, which is 22 times the 5,000-page volume the statute itself treats as a marker of unusual circumstances. Asked for as one delimited export of named fields, it is a single file and a single query.
The burden, in other words, is something the request creates or avoids. That is why the statute is worth reading closely. An agency must provide a record in any form or format requested where it is readily reproducible in that form, must make reasonable efforts to maintain records in reproducible formats, and must make reasonable efforts to search electronically. Automated searching is written into the definition of a search, and the only stated carve-out is significant interference with the system itself.
So the request names the schema. The published dictionary documents 214 fields across six tables, and the reporter's question needs 31 of them across four, which is 14.5 percent of the schema and a fact worth stating in the letter. Six of the 31 carry personal identifiers, so the request asks for those as a consistent surrogate key instead, which keeps records linkable without asking for identities the agency would have to withhold anyway.
How it works
Send the documentation
Attach the data dictionary, record layout or prior releases, then say what your question is.
Get the field list
One row per field you actually need, with the justification your question gives for it.
Read the request
A letter in the system's own vocabulary, with each likely objection answered inside it.
Work it in chat
Ask what to drop if they push back, or how to word the narrowing they will propose.
What you get
- A request naming each table, field, delimiter, encoding and date range rather than describing records
- One sheet row per requested field, with why your question needs it and what breaks without it
- The burden objection pre-empted, using the agency's own documentation of its own system
- Prior releases from the same system cited as proof the export is readily reproducible
- Identifier fields requested as a surrogate key, so the privacy objection loses its target
- The record layout, codebook and lookup tables asked for alongside the data itself
- A fallback ladder: what to accept if they refuse the first format, in priority order
Common questions
Why name fields instead of describing the records?
Because a prose description is what a burden refusal attaches to. An agency answering please provide all inspection records gets to pick the most expensive reading. An agency answering a named list of 31 fields from four tables it documents itself has nothing left to interpret, which is the point of writing it that way.
Can I actually insist on a CSV?
You can ask, and the statute is specific. Under 5 U.S.C. 552 an agency shall provide a record in any form or format requested where it is readily reproducible in that form, and shall make reasonable efforts to maintain records in reproducible formats. Naming the format is what triggers that duty.
They said running a query is not a search. Is that right?
The statute defines a search as reviewing records manually or by automated means to locate what is responsive, and separately requires reasonable efforts to search electronically. The stated exception is where those efforts would significantly interfere with the agency's automated information system, which is a narrower claim than the one usually made.
How do I get around the privacy objection?
By not asking for the identifiers. Where your analysis needs records linked rather than named, ask for a consistent surrogate key in place of each identifier field. Six of the 31 fields in the worked example were handled that way, which keeps the longitudinal analysis intact and removes the thing the agency would withhold.
What if they only have it in some proprietary system?
Almost every such system exports. Ask for the format the system natively produces, name two acceptable alternatives, and ask for the record layout and lookup tables alongside. If a public dashboard runs on the same tables, say so in the letter, because that establishes an export path already exists.
Is this only for federal agencies?
No. The field-level approach works against any public records law, and the arithmetic about format and burden is universal. What changes is the statutory language, so name the state or the body and the letter is written to that law rather than to the federal provisions used in the worked example.
What happens if they refuse anyway?
You are in a better position than a prose request leaves you, because the refusal now has to explain why a named field from a documented table cannot be produced. FOIA denial appeal tests each exemption against the records described and rebuilds the request timeline as evidence.
Government Database Request With Field List
Fill in the form and your workspace opens with the work already underway.