Where each number came from

Every input carries a label saying whose number it is: read from a document, computed from your statements, typed by you, inferred, or the industry’s typical value. The label decides how hard it is tested.

3-minute readUpdated October 2026

Every number says whose it is

Every input carries a label, so you never have to wonder whose number a number is: yours, one read from your documents, one the engine worked out, or the industry’s typical value.

Document-verifiedFrom the statementsYour numberMedian, confirmedAdvisor inference · please confirmAdvisor estimateAssumed median

• Document-verifiedRead from a document the company supplied: billing export, carrier invoices, general ledger, contracts.Tested ±5% · counts in full
• From the statementsComputed from the profit-and-loss or the balance sheet.Tested ±5% · counts in full
• Owner-stated (“Your number”)Typed in by the owner.Tested ±5% · counts in full
• Industry median, confirmedThe typical value, which the owner confirmed fits the business.Tested ±5% · counts in full
• Advisor inferenceWorked out from other answers, waiting for the owner to confirm.Tested ±15% · counts half
• Advisor estimateEstimated from size and business type, without the owner’s data.Tested across the typical range · counts nothing
• Assumed medianAn unanswered question, at the industry median.Tested across the typical range · counts nothing

The engine also has a tier for figures checked against a public or independent record; it counts in full.

The label decides two things: how widely the number is tested when the range is built, and whether it counts toward the share of the value resting on your numbers.

The sample, by label

In the finished sample, 54 of 56 inputs are the company’s own figures. Owner-stated answers still carry 85% of the swing, because the riskiest choices (who owns the platform, what happens without the owner) have no document behind them yet. The report says so on the page.

Documents beat typing

Send the billing export, the carrier invoices, the general ledger and the contract schedule, and the engine reads the numbers itself: seats and price per seat from the billing export, carrier dependence from the invoices, owner pay and one-time items from the ledger. Nothing read from a document is typed by hand.

Then it does what a diligence team does: it ties each document to the statements and to your answers, and prints the checks.

• Billing export recurring charges × 12 against recurring revenue on the statements, within 2% for accrual timing.
• Seats in the billing export against the operating statistics, exactly.
• The largest account’s share in the billing export against your questionnaire answer, within a point.
• Twelve months of carrier invoices against cost of service on the statements, to the dollar.
• Every expense line in the general ledger against the statements, to the dollar.
• EBITDA rebuilt from the ledger against EBITDA on the statements, to the dollar.

When a figure moves from owner-stated to document-verified, its value doesn’t change. The next report’s revision bridge still lists the move, at $0, so you can see the evidence firming up.

Rules

• Firm tiers. Owner-stated, from the statements, document-verified, checked against a public record, and a confirmed median are all firm: tested at ±5%, counted in full.
• Inference. Tested at ±15% and counted at half weight until the owner confirms it.
• Estimate and median. Swept across the type’s band and counted at zero.
• The stamp. Every page carries a fingerprint of the input set: every resolved value and its tier. Two reports with the same stamp carry the same inputs with the same labels.

See it on a real report, or see it on yours.Download a sample reportSchedule demo