HomeGuides › Top-down vs bottom-up

Top-down vs bottom-up market sizing: which to trust

Method · ~8 min read · public-methodology

Ask two analysts to size the same market and you'll often get numbers that differ by 3×. Usually it's because one worked top-down and the other bottom-up — and neither said so. Both methods are legitimate. Both are wrong on their own. The point of using them together is that where they disagree tells you exactly which input to distrust.

Top-down: start big, take a slice

Top-down starts from a broad published market and carves your sector out of it.

TAM (top-down) = parent market × capture share

Warehouse robotics, illustratively: a published warehouse-automation market of $30B, of which robotics is a 30% slice, gives $9.0B. Fast, and it anchors to a number a committee has probably already seen. The danger: you inherit whatever assumptions sit inside that $30B, and if a report brokered it optimistically, your slice inherits the optimism. Worse, if the "capture share" creeps above ~60%, you've quietly redefined the sub-sector as the whole parent market.

Bottom-up: count the buyers

Bottom-up builds from countable units — buyers, seats, facilities — times what each pays.

TAM (bottom-up) = units × price per unit × attach rate

Same market: 72,000 automatable facilities × $150K/yr × 28% attach ≈ $3.0B. Every factor is a countable, citable number, which makes it the most honest method — and the easiest to lowball, because a tightly-drawn buyer universe silently shrinks the whole answer.

 Top-downBottom-up
Anchors toA published parent marketCountable units × price
Fails byInheriting an inflated parent, or an implausible capture shareDrawing the buyer universe too tightly
Bias tendsHighLow
Best forA fast ceiling / sanity anchorThe defensible headline number

The cross-check that catches the error

Because the two methods bias in opposite directions, running both isn't redundant — it's a trap for bad inputs. Our engine performs an explicit price-coherence check: it back-solves the price-per-adopter implied by the top-down number and compares it to your bottom-up price assumption.

implied price (top-down) = top-down TAM ÷ (units × attach rate) coherent when: 0.1× ≤ implied price ÷ your price ≤ 10×

In the sample, top-down implies roughly $1.19× your bottom-up price — comfortably coherent. If that ratio blew out past 10×, it would mean one of the two methods has a broken input (usually price, attach rate, or capture share), and the report refuses to quote a single number until it's fixed.

Rule of thumb: if top-down and bottom-up disagree by more than ~2×, don't average them and move on — that gap is the finding. Trace it to the single input driving the spread and re-source that one number.

So which do you trust?

Neither, alone. You trust the agreement. Bottom-up gives the most defensible headline; top-down gives the ceiling and the sanity anchor; a third method — value-theory — breaks ties. Reducing all three to a median with a stated spread is the whole point of triangulation.

Honest caveat. Figures here are illustrative and rounded for teaching. In a commissioned report the parent market, unit count, price, and attach rate each carry a dated public citation, and the capture-share check runs on the real numbers, not these.

See both methods run on your sector

The instant estimate computes top-down and bottom-up side by side and flags the agreement band for your own inputs.

Free instant estimate → Commission a cited report