What is Quartyl? The Transfer Pricing OS, Explained

Quartyl is a multi-tenant workspace where a transfer pricing study moves from an uploaded data file to a signed-off, fully documented deliverable — with every decision along the way recorded, reviewable and exportable. The short version of what it is and what it is not:

What it is

Three layers, one workflow:

1. The benchmarking engine. You upload the benchmark dump — the comparable companies’ multi-year financials, as an Excel workbook (.xls/.xlsx/.xlsm) exported from your own source — and describe the tested party alongside it. Quartyl runs the analysis pipeline — quantitative screening, qualitative and AI screening, and the statistical analysis — and produces the arm’s length range with a documented reason for every comparable included or excluded. The screening is a governed step: the system’s dispositions are recorded with their evidence, and a reviewer’s accept/reject override is a decision with recorded rationale, not a spreadsheet cell.
2. The workflow. A study is not a file, it is a state machine: Draft → Study Initialized → Processing → In Review → Report Generated → Document Generated → Signed Off → Archived, with explicit rejection and failure paths. Each state has named roles who may act on it (roles and permissions), and every transition is written to the study’s audit trail.
3. The evidence layer. Every screenable decision — the AI’s reasoning, the reviewer’s override, and the ledger events recorded afterwards — is captured in the evidence ledger with its source. When a transfer pricing officer asks “why is this comparable in (or out) of the pool?”, the answer is a record, not a memory.

On top of those, the workspace adds the things a practice needs around the study itself: team and firm administration, industry diagnostics, and — where the firm’s plan includes the feature — predictive risk scoring per study and across the portfolio, the evidence repository, and white-label branding of the workspace shell (the sidebar, the sign-in overlay and the claim page). The generated reports carry the platform’s own fixed brand identity.

The study lifecycle at a glance

State	Who drives it	What happens
Draft	Analyst	Study configured: details, tested party profile, uploaded file, benchmarking parameters
Study Initialized	Analyst	Configuration is complete and the run is queued
Processing	The system	The 8-step analysis pipeline runs; progress is visible step by step
In Review	Manager	The comparables and results are reviewed; approved or sent back with reasons
Report Generated	The system	The Excel workbook (the primary deliverable) is built on approval
Document Generated	Partner, Firm Admin or Superadmin (on demand)	The seven-chapter Word master report plus the branded PDF, built from the recorded results with the engagement details
Signed Off	Partner, Firm Admin or Superadmin	Final approval
Archived	Partner, Firm Admin or Superadmin	The study is closed out and immutable

The final document can be requested from any state from In Review onwards, as long as the run has produced results — it does not have to wait for sign-off, and generating it never locks the study.

Rejected and Failed are working states, not dead ends: a rejected study goes back to In Review after the analyst addresses the reasons; a failed study is re-run.

Who uses it

Role	Day to day in Quartyl
Analyst	Creates studies, uploads data, addresses rejections, re-runs failed studies
Manager	Reviews studies in In Review, overrides comparables with a rationale, approves or sends back
Partner	Reviews and approves, signs off, archives, requests the final document
Firm Admin	Runs the firm: team and roles, firm settings and security, billing and usage; acts across study states as the firm’s operator
Superadmin	Platform level (plans, firms); study access is owner-scoped to the records they created

What makes it different

• The record is the product. The Accept-Reject matrix, the override rationale, the AI’s citations and the access logs are first-class, exportable artefacts — the same artefacts a TPO examination turns on.
• AI is a screened step with receipts, not a black box. Screening recommendations come with the evidence that produced them, and a human makes (and owns) the final disposition per company.
• The report engine is deterministic. The seven-chapter Word master report and the branded PDF are built from one report context assembled from the study’s recorded analysis — method selection, characterization and adjustments come from a rules-based decision engine, not from prose written at generation time.
• Multi-tenant by construction. Every study, every record, every log lives inside your firm’s tenancy; plan features gate capabilities per firm.
• No learning, nothing hidden. The statistics are descriptive (counts, moments, percentiles, correlations) and the risk and reliability signals are documented weighted rules — not a trained model whose reasoning you would have to take on trust.

What it is not

• Not a database. The population it screens is the dump you upload: no registry record is joined onto a comparable, and it does not reach out to a commercial benchmark source for you. The one step that looks outward is optional — Web Research, behind a plan feature — and it reads a handful of public pages about companies already in your dump, reporting what those pages say as labelled findings with the URL behind each. It is not a substitute for your benchmarking source.
• Not a judgment machine. It makes the recorded judgment faster and more defensible; the accountable decision — tested party, method, disposition — remains a human’s, with a human’s name on it.

The next page is roles and permissions — what each of those roles can actually do; claiming your account and sign-in with MFA cover getting in; and your first study takes you from empty workspace to a benchmarked result.