Trading Comps Builder

⚠️ **Finance warning / Avertissement finance** : informational only, not investment advice. A comps range frames what the market pays for similar businesses, it does not certify what one business is worth: peers can be mispriced together, multiples move, there is a real risk of loss in any investment priced off them, and no range here is a promise of return. Curated by Skill Harbor: the trading comps builder skill at the root of andreworia/claude-finance-skills. Use it when building the comps page of a valuation deck, framing a minority valuation reference, or pressure testing a public market reference range. The governing principle sets the tone: a peer is a company whose multiple a sophisticated investor would actually use to price the target, and everything else is a distractor, so the discipline sits in what gets cut, since a page with 25 names usually hides 6 peers behind 19 distractors. A candidate belongs only if it passes three tests: business model match on revenue model, customer profile and unit economics; financial profile match on growth, margins and capital intensity, or differences priced explicitly into the discussion; and enough trading liquidity and disclosure to produce a defensible multiple. Geography is a tag, not a filter, with any geographic multiple delta acknowledged. The workflow builds a long list of 25 to 50 names from competitor disclosure, analyst peer sets, proxy peer groups, sector indices and precedent deal sets before any cutting, scores every candidate against the target in a table (business model must score high, financial profile at least medium), then pulls metrics on consistent conventions that are named as the most common comps error when mixed: one pricing date, treasury method diluted shares, the latest balance sheet for net debt, clearly labelled trailing or forward periods, one consensus source for the whole set, and one reporting currency at the pricing date spot rate. Off cycle peers are calendarized to a common basis with the mechanics shown in an appendix, EBITDA is adjusted consistently across target and peers with a default treatment table (stock compensation subtracted as a real cost, recurring one time items left in, lease treatment uniform, two adjusted columns shown so nothing is silently adopted), and outliers, multiples 50 percent above or below the median, are handled explicitly by keeping and explaining them, moving them to a near peer page, or excluding them for a stated structural reason, never by silently dropping them. The page presents multiples beside the operating metrics that should explain them, uses medians and interquartile ranges rather than means, and lands with a written recommendation on where the target should price inside the range and why. From the andreworia/claude-finance-skills repository (MIT). Honest caveats: market data, consensus estimates and peer financials are yours to source on one consistent basis and to verify; stale fiscal periods and mixed conventions will quietly corrupt the set if you let them. Skill Harbor never reviews the code, review it yourself before use.