The deterministic engine

Normalize

The uploaded P&L is parsed deterministically. The AI classifies which operating-expense lines are add-backs (owner comp, discretionary, non-cash, financing, one-time); the engine adds them back to compute SDE, then subtracts a market-rate manager salary for Adjusted EBITDA. Every amount comes from the P&L — the AI classifies, it never invents a figure.

Anchor

The earnings base and metric select a low / mid / high multiple band from a sourced size-band table (SDE and Adjusted EBITDA, by size). The AI may nudge the mid within the band for the category, but the nudge is clamped inside the sourced band and applied only on the analyst’s click.

Scorecard

Ten weighted Tier A drivers move the multiple within the anchor band. Tier B then applies a symmetric, capped adjustment only for verified Ecomma hold/provenance and operational uplift. Track record, buyer demand, and category trajectory remain neutral here because learning, pricing strategy, and anchor evidence already own them. Missing or unverified evidence is neutral; the engine, not the AI, turns approved scores into the multiple.

Reconcile

Verified, closed, recent, same-category, same-metric, same-basis and similarly sized transactions are assessed for sufficiency. The engine uses an evidence-derived comp weight only when the effective sample is adequate; thin evidence carries zero weight and the scorecard multiple stands. Comparables are optional and never fabricated.

Value & range

Earnings times the reconciled multiple gives the value, with an outer range and a confidence band that widens on thin data and tightens when dense, close comps agree.

Ask & floor

The recommended ask sits a set premium above value; the walk-away floor a set discount below. Both are rounded to a clean figure. These are the numbers taken to market.