Vendor Due Diligence — for PE & Search Funds

Your DD analyst used a different
churn definition last quarter.
FinArrow does not.

Three sell-side analysts will give you three NRR figures for the same target. Only one of them was computed the same way as last deal’s. You don’t know which.

Developed at HEC Lausanne · M.Sc. Finance thesis, 2026
Deterministic engine · AI‑assisted, never AI‑decided
Section 01 · The Risk

What you
can’t see
can be priced.

On a €50M SaaS acquisition, a 4-point error in NRR is worth roughly €2M of misallocated equity value at typical multiples. FinArrow runs the calculation deterministically, so the error is visible — not hidden.

The error doesn’t announce itself. It arrives as a footnote in the IC memo, six months after the wire.

01 / 03
Definition drift

NRR computed three ways across three sell-side decks for the same target. Cohort window varies. Churn convention varies. The IC sees one number.

FinArrow fixes a 12-month cohort window, caps GDR at 100%, and logs every assumption.

02 / 03
Cohort smoothing

Your analyst rounded the partial month. The retention curve looks flat. The actual cohort fell off the cliff in month seven.

Point-in-time monthly sums — never a running total. Trailing partial months are dropped, not smoothed.

03 / 03
One-off fees as ARR

Onboarding, setup, and professional-services fees billed monthly slip into the recurring line. The ARR looks larger than the subscription base that actually renews — often a double-digit overstatement.

FinArrow separates recurring revenue from one-off and setup fees at ingest, and flags any period where the one-off share crosses a set threshold.

Section 02 · The Report

Same input.
Same output.
Every deal.

Send us your billing data — a CSV or XLSX from any system — for a secure, white‑glove review. Column mapping is AI‑assisted and runs locally; customer identifiers are anonymised before any model sees them. FinArrow normalises FX, expands annual contracts, and computes every metric deterministically, and a human reviews every figure before delivery. The AI‑written narrative is labelled separately and never overrides a deterministic finding — the final call is yours.

−  The old way

How DD usually arrives

  • An Excel workbook assembled by the sell-side, with formulas pasted across 14 sheets.
  • Three analysts. Three definitions of cohort. None of them documented in the dataroom.
  • A week of reconciliation before your team trusts the NRR figure enough to put it in the IC.
  • Refunds entered positive. ARR overstated. Discovered — if at all — in week three.
  • A narrative built on figures no one can reproduce the morning after the deal closes.
Time to trust: days of reconciliation
→  FinArrow

How FinArrow delivers it

  • One engine. Cohort NRR, GDR, ARR bridge, logo retention, gross churn — computed from one Python pipeline.
  • One config. Every threshold is a named, owner-confirmed setting. No silent defaults. No hidden floors.
  • One report. A nine-page HTML/PDF with assumptions logged on the last page — including the ones you didn’t set.
  • Every flag, auditable. The 5 statistical anomaly checks export every flagged transaction to an Excel workbook — open it, filter it, verify any one yourself. The engine flags; you decide.
  • Reproducible. Same input, same config ↠ same output. The number is auditable next quarter.
Time to trust: the first read

The buyer is the victim of a broken DD process, not the architect of it. FinArrow exists so that the next IC memo cites a number you can defend in the same language six months from now.

Section 03 · Methodology

The shape of a deterministic engine.

No imputation. No silent defaults. No rounding of cohort windows. The four figures below describe the entire surface area of what FinArrow promises — and what it refuses to do.

6s
Engine compute time
The deterministic calculation itself — ingest to a nine-page report with a 50,000-path Monte Carlo. Every report is reviewed before delivery.
17
Deterministic invariants
Codified in the engine’s determinism specification.
5
Statistical anomaly checks
Each tied to a single owner-set threshold.
0
Silent imputations
Two exceptions, both opt-in and logged.

Every default applied during a run is written to the assumptions log on the report’s final page. If you didn’t set it, the report tells you what was set on your behalf — and the exact setting that produced it.

Audit your next deal

Run it on your next target.
Read the same report your
analyst would have produced.

A 30‑minute methodology walkthrough. No sales script. We reply within 24 hours.