Fynto drafts each year's report from fresh evidence, then stops and asks wherever the draft contradicts last year's position. Your team's calls are kept, with the reasoning, and never silently rolled forward.
Works with the benchmark study or database you already have. Runs in your cloud account or on your hardware.
A transfer pricing file is rewritten every year about mostly the same entities and transactions. The report records the conclusion. It does not record the alternatives your team considered, or why they chose this one. When the person who made the call moves on, the reasoning goes with them.
The new draft reflects this year's evidence. It also quietly changes a characterization or a method, and nobody notices until someone happens to compare it against last year's file. Usually that someone is an examiner.
Last year's positions carry into this year's report. A position that stopped being true in March is still in the file in December, because nothing prompted anyone to revisit it. The error compounds exactly when staff turn over.
Every year's draft is written from that year's evidence. Separately, Fynto keeps a record of every call your team made last year, with the reasoning. Where the draft and the record disagree, the draft stops and asks. Where they agree, last year's call is offered back as one click, and the click is recorded. Last year's position is brought back as a question when the draft disagrees, an offer when it doesn't. Never as an answer.
This is what a reviewer sees when they open a file. Nothing is decided that they did not click.
Upload the agreements, financials, and organization charts for the year. Fynto proposes the controlled transactions, the functional profile of each entity, the characterization, and the method, each one pointing at the document it came from.
The draft does not look at last year's file. That independence is what makes the comparison in the next step mean something.
A real change in position becomes a gate that blocks the report. The gate shows last year's position, who set it, and the reason they gave. The reviewer confirms the change, reverts to last year, or leaves it open with a note.
No model sits between the reviewer's click and what gets recorded. Each gate is one decision. There is no “confirm all.”
When the draft and the record agree, last year's value and reasoning sit under the current one with a Reuse button. The reviewer takes it or doesn't. Either way, the choice is written down with their name on it.
Items left open last year come back first. “Leave open” never turns into “forgotten.”
Try it
Make the call a reviewer would make. Then see what Fynto writes down, and what the next reviewer sees a year later.
Illustrative data. Nothing you type here leaves your browser.
Each one is recorded the moment a reviewer makes it, with their name, the date, the reason, and the alternative they turned down.
How each controlled transaction was characterized: a service, a license, a loan, a sale. And why the draft's proposal was accepted or overruled.
Which pricing method applies to each transaction, which ones were rejected, and the facts the choice rested on.
Each entity's functions, assets, and risks, the characterization that followed, and the narrative the reviewer wrote or edited about it.
Every candidate company in the benchmarking study, whether it was accepted or rejected, and the reason. A rejected comparable comes back next year with its reason attached.
Upload the prior-year report you already have, whether your team wrote it or your advisor delivered it. Fynto reads its transactions, functional analysis, and conclusions into this year's file as proposals for your team to review. Each confirmation becomes the first entry in the record. The drift gates start working in your second cycle, and the reasoning is captured from the first.
Curious how the record is built? Read the engineering post →
Keep the comparables database you already license and the screening approach you already trust. Export the search, or hand over the accept/reject matrix your team already worked. If your advisor delivers the benchmarking study, ask for the accept/reject workbook behind it and upload that. Fynto records every accept and reject, computes the arm's-length range, and carries the result through method selection into the report.
Uploaded comparables stay in your workspace. The range is computed from your rows, never quoted from a PDF.
The report says what the position was. It rarely says what else was considered, what the evidence looked like at the time, or who made the call. By the time the question arrives, that person may have left the company.
With Fynto, every position in the file has a record behind it. Re-decisions do not overwrite earlier ones. They sit on top, in order, so the full history of a position is there to hand over.
Fynto is a San Francisco company. We are onboarding a small number of design partners: in-house tax teams and advisory firms that sign off on transfer pricing files every cycle and want the reasoning behind each position to outlast the people who wrote it, and the advisor who drafted it.
[Design-partner quote]
[Founders' transfer pricing and engineering credentials]
Most teams run Fynto inside their own cloud account. Teams with stricter obligations run it on their own hardware, with AI analysis routed to a model endpoint they control or switched off.
|
|
POPULAR
|
|
|
|---|---|---|---|
| Who manages infrastructure | Fynto | Your team | Your team |
| Where your data lives | Our cloud | Your cloud account | Your data center |
| Data stays in your boundary | |||
| AI analysis | Full | Full | Optional |
| Enterprise controls | Standard | Full | Maximum |
| Best for | Smaller teams | Most teams | Teams with strict data obligations |
Any document rewritten every cycle about the same subjects, where a named person has to defend a closed set of positions, can run on the same record. Transfer pricing is where it is live today.
Talk to us about a design-partner engagementNo. The AI drafts proposals from the year's evidence and points at where each one came from. Every position in the report is confirmed, changed, or reverted by a named reviewer. Where the draft contradicts last year, the report is blocked until a person decides. There is no “confirm all” for those gates, and the recalled value is never pre-applied.
It records your team's judgment. Your screening approach, your characterizations, and your method choices are the ones that get written down and brought back. Fynto makes the senior review a queue of specific questions instead of a re-read of the whole file, and makes a reviewer's reasoning outlast their tenure.
Whatever is in the prior-year report you upload. Fynto reads its transactions, functional analysis, and conclusions into the new file as proposals. Each one your team confirms becomes the first entry in the record for that subject. From the second cycle on, the draft is compared against those entries and the drift gates fire.
No. If your advisor delivers the benchmarking study, upload the accept/reject workbook behind it. If you license a database, export the search from it, or hand over the matrix your team already worked. Either way, Fynto records each accept and reject as a decision, computes the interquartile range from your rows, and carries it into the report. A study PDF rides along as evidence; the range is never quoted from it.
They stay with you. The record lives in your Fynto workspace, not in your advisor's working papers. A new advisor, or a new member of your own team, opens the file and sees each position, who set it, and the reason they gave at the time.
In your cloud account or on your hardware, for the two deployment options most teams choose. Your documents, comparables, and the decision record stay inside your boundary. The AI analysis can be pointed at a model endpoint you control or switched off entirely for on-premises deployments.
For any position in the file: what it was, what the draft had proposed instead, who decided, when, the reason they wrote at the time, which documents they were looking at, and every re-decision since, in order. Open items are disclosed, not dropped.
A 30-minute walkthrough on our sample file. If you would rather see your own, we can sign an NDA first and use last year's report and this year's documents.