What is in a case file
One taxpayer, one folder, two sides. What the preparer was given sits on one side. The return, and the reasons behind it, sit on the other.
The folder
This is the Illinois sample. The files come with the sample set, which is free on request.
given/
What a preparer, or a model, is allowed to see.
- narrative.mdThe taxpayer’s year, written out: who they are, what happened and what they handed over.
- documents/w2-lakeshore-cloudworks.pdfForm W-2 from Lakeshore Cloudworks, Inc.
- documents/1099b-calumet-harbor-securities.pdfForm 1099-B from Calumet Harbor Securities, LLC.
- documents/3921-lakeshore-cloudworks.pdfForm 3921 from Lakeshore Cloudworks, Inc.
key/
The answer side. Keep it away from the model.
- return.pdfThe prepared return, printed by the tax software.
- lines.csvEvery line with a figure, as data, with where the figure comes from.
- workpaper.mdThe decisions: the question, what the return does, why, the authority and what it is worth.
case.json sits beside the two folders. It names the case, lists every file with its checksum, and carries your ID for the case next to ours.
What each part is for
- The narrative
- What the preparer was told and handed, written out. Each passage has a number and a source: the taxpayer said it, or a named document shows it. When a model gets a line wrong, you can see which passage it missed.
- Source documents
- PDFs laid out like the real forms and statements: W-2s, 1099s, K-1s, closing statements, lender letters. The sample set has one to five per taxpayer.
- The prepared return
- One PDF with the federal return and every state and city return the facts require, printed by the tax software. Returns in the sample set run from 10 to 61 pages.
- Every line as data
- A spreadsheet row for each line that carries a figure, saying where the figure comes from: a box on a document, another line, a formula over other lines, or an amount set by law. The Illinois sample has 126 rows, and a script recomputes the 71 that are formulas. A grader can compare a model’s return to it without reading a PDF.
- The workpaper
- Each decision on the return, written out: the question, what the return does, why, the authority, and what the decision is worth in dollars. That last part tells a grader how much each mistake matters.
- Two sides
- Everything under given is safe to show a model. Everything under key is the answer. The split is in the folder names, so a pipeline cannot mix them up by accident.
The format is described in full on the case file format page.
The whole engagement
A real client does not hand over a tidy folder of tax forms. Some teams want the job the way an accounting firm sees it, so that the model has to find the facts before it can use them.
- A firm and a preparer. Both invented, with an engagement letter the client has signed.
- The email thread. The client and the preparer writing back and forth, with attachments, over the weeks the return takes.
- Last year’s return. So carryovers and changes from the prior year have to be noticed.
- The client’s own records. Bookkeeping reports, a profit and loss spreadsheet, invoices, receipts, tax bills and donation letters, alongside the tax forms.
- The return and the workpaper. On the key side, as in every case file.
Our ID and yours
Every case gets one random ID from us when it is created. It never changes and is never reused.
Your own ID for the case is stored against ours. Deliveries arrive named the way your pipeline expects, and either ID finds the same case.
- Our ID
- 7b46e35c-b49f-48a1-b935-da612987d6e2
Shortened to 7b46e35c in file names and on this site. - Another ID
- IL01
The folder name for this case in the 2025 sample set.
Three things we make
- Prepared returns
- Complete federal and state returns for fictional taxpayers, prepared in professional tax software and printed as the software prints them. Where the facts call for a city return or a second state, those are included.
- Answer keys for evaluation. Targets for training.
- Case files
- A prepared return with its narrative, its documents, every line as data and the workpaper, in the folder shown above.
- Lets a grader score line by line, and tell a judgment call from an error.
- Custom cases
- Fictional taxpayers and the papers they would hand over, written so that a chosen set of forms has to interact. The 20-state sample set was written this way. For a harder test we build the whole engagement, described below.
- Fills the gaps in your coverage.
The 20-state sample set was made before the case file format. Each of its folders holds a taxpayer profile, the source documents and the prepared return. Illinois has been rebuilt as a full case file.
Invented taxpayers, real experience
We adhere entirely to Section 7216 and never reuse any client data.
- What is invented. Every name, address, identification number, employer, account and figure in a case is made up for that case.
- What is real. The experience. We may use kinds of situations we have seen before, such as a prenuptial agreement in Washington that allocates assets, income and liabilities between spouses, or a taxpayer who lives in another country and works remotely through an LLC in the United States.
- Coincidence. Cases are written to look real, so a made-up name or number can match a real one by chance.
- Your files stay yours. Files you send us are used for your project and nothing else.