Sample buy-side due diligence audit: Oranjedal Payroll B.V.

Saloa does not issue a buy or pass verdict. Every audit ends in one of five statuses: NO_ACTION, ACQUIRE_INFORMATION, MORE_EVIDENCE, HUMAN_DECISION or DEEP_AUDIT_REQUIRED, followed by one required next move.

In the Oranjedal file the status is DEEP_AUDIT_REQUIRED. Three reported figures do not survive recomputation against the underlying records. None of them is a judgement on the quality of the business. Each is a gap between what was reported and what the evidence supports.

Reported ARR: €14.2M. Computed ARR: €13.3M.

An ARR bridge walks from reported annual recurring revenue to the revenue the contract base actually supports. In Oranjedal, 24 contracts are signed but not yet live. Together they total €0.9M. They are included in reported ARR but not in the model.

Signed is not live. A contract that has not started billing is a commitment, not recurring revenue. In buy-side due diligence this is one of the most common points where a headline figure is re-based. Here it is isolated before the indicative offer, not during exclusivity.

What was checked: each ARR line in the seller schedule against the customer record and its go-live status. A deterministic code-check flags every contract counted as recurring without an active billing state.

Reported net revenue retention: 104.2%. Recomputed: 89.9%.

Net revenue retention measures what the existing customer base is worth a year later, after expansion, contraction and churn. It is sensitive to one decision: which customers are left in the denominator.

In Oranjedal the reported figure excludes enterprise churn. Once the churned enterprise accounts are put back into the cohort, NRR falls below 100%. The base contracts instead of expanding.

What was checked: NRR verification rebuilt from the customer records, with churned accounts kept in the starting cohort. A retention figure that only holds when departed customers are excluded is recorded as a CONTRADICTION, not as a fact.

Deviation: €0.76M. Cause: the R&D capitalisation rate rose from 22% to 41%. The change is unaudited.

Capitalising development cost moves expense from the profit and loss account to the balance sheet. The higher the rate, the higher reported EBITDA. That is permitted. It is not neutral. When the rate nearly doubles without audit, part of the reported EBITDA comes from a policy choice rather than from operations.

In an EBITDA adjustment review this deviation is treated as a separate line. Applying the prior-year rate isolates €0.76M of EBITDA that depends on the policy change.

What was checked: capitalised development cost recomputed at both rates and reconciled to the reported EBITDA bridge.

Source tiers T1–T4

Every input is weighted by where it comes from.

• Tier 1: official registries and filings: company registers, statutory filings, regulators, court records.
• Tier 2: independent reputable media and research.
• Tier 3: seller- or subject-supplied material: data room files, models, ledgers, CIM, contracts, management answers. Never FACT on its own.
• Tier 4: aggregators and signals: job boards, review sites, social media, recollection. Never FACT.

The epistemic gate

Every claim gets one of six statuses: FACT, OBSERVATION, INFERENCE, ASSUMPTION, CONTRADICTION or UNKNOWN. A seller figure becomes FACT only when it reconciles with an independent Tier 1 or Tier 2 source. An inference or assumption is never promoted to fact. Missing evidence stays visible as UNKNOWN, together with the party who can resolve it.

Deterministic code-checks

Arithmetic, dates and source weight are checked in code, not in prose. ARR is summed against live contracts. Retention is recomputed from the customer records. EBITDA is reconciled at both capitalisation rates. The result of a code-check does not depend on how the document is worded.

The four-voice dialectical framework

Before the ruling, the evidence is debated by four separated roles:

• Creator: builds the strongest case for the upside.
• Advocate: builds the strongest defence of management’s logic.
• Devil: hunts for anomalies, hidden liabilities and breakpoints.
• Judge: reads the closed debate, weighs it against the source tiers, and issues the status.

Code-checks enforce the boundaries between the roles. The debate reduces single-perspective bias. It does not replace the human decision.

• Signed contracts counted as recurring revenue before go-live.
• Retention metrics calculated on a cohort that excludes churned accounts.
• An accounting policy change that lifts EBITDA, with no audit trail.
• Seller-supplied figures (Tier 3) presented without an independent reconciliation.