All insights
    19 Apr 2026·TAM·8 min read

    Stress-testing TAM bottom-up without a McKinsey budget

    Top-down TAM numbers from analyst slides survive the IC and then die in the boardroom. Here is how we build a bottom-up estimate in a week using public company filings, job postings, and ICP density.

    Top-down TAM rarely survives the third board meeting. Bottom-up TAM rarely survives the IC, because it takes too long to build. The compromise most teams reach is a top-down number with bottom-up theatre.

    A defensible bottom-up TAM for a lower-middle-market deal can be built in about a week. The inputs are unglamorous: the ICP definition, count of qualifying entities by geography, observed conversion benchmarks from analogues, and a realistic ARPU band.

    The output is not one number. It is a TAM, a serviceable available market, and a five-year reachable market with explicit assumptions tagged as observed, inferred, or user-supplied. The point is not precision. The point is that the assumptions are now arguable. Arguable assumptions are also schedulable ones: each becomes a named test with a date once the team gets to executing the 100 day plan.

    Step one: define the ICP tightly

    Start from observed customers, not the stated ICP. Pick the two or three segments that account for most of closed-won revenue and describe each by industry, size band and geography. A loose ICP inflates every number that follows.

    Step two: count qualifying entities

    Company registries, industry association lists, public filings and job-posting data can count qualifying buyers in most B2B segments within a few days. Job postings are especially useful: companies hiring for the role that buys the product are the most likely buyers.

    Step three: apply realistic conversion and ARPU

    Take conversion benchmarks from analogous companies rather than from the target's pitch. Apply an ARPU band built from the target's actual contract values, not list prices. The resulting range is usually a fraction of the top-down figure, and that gap is itself a finding.

    Step four: tag every assumption

    Mark each input as observed, inferred or user-supplied. When the IC challenges the number, the discussion moves to specific inputs rather than to whether the model can be trusted at all.

    What we'd do this week

    Rebuild the target's TAM using only the closed-won ICP and a job-posting count for the buying role. If the result is less than half of the management figure, raise it at the next IC and adjust the growth case before the bid is final.