How to do it
Recognise top-down by its first sentence
It begins with a large number from someone else's report and then shrinks it: "the market is $40B, we target the 5% that is SMB, of which we can win 2%". Neither percentage came from anywhere, and the report was measuring a category adjacent to yours.
Build bottom-up from a thing you can count
Registered businesses of a type, sellers on a marketplace, apps in a store category, job postings for a role. The number is smaller and you can name where it came from.
Use top-down only as a sanity check
It has one good use. If bottom-up gives you a figure larger than the published category, one of your inputs is wrong. Top-down is a ceiling to check against, not a foundation to build on.
Keep the inputs visible either way
The value of a sizing exercise is not the total — it is knowing which assumption to attack first when reality disagrees. A number with four visible inputs is a model. The same number alone is a claim.
Worked example
The two methods side by side
The row that decides it is the second. A number whose inputs you cannot check is a number you cannot improve when reality disagrees — and reality always disagrees with the first estimate.
Why top-down keeps getting used anyway
It takes ten minutes and produces a large number, and both of those are attractive at eleven at night before a deadline. It also feels authoritative, because the source is a real firm with a real report.
The problem is not that the report is wrong. It is that the report measures a category defined by an analyst for a different purpose, and the two percentages a founder applies to it come from nowhere. The authority of the source does not transfer to the arithmetic done on top of it.
How to use both properly
Build bottom-up. Then look up a top-down figure and check that your total sits below it. If bottom-up exceeds the published category, one of your inputs is wrong — most often the buyer count, occasionally the price.
If the two land within an order of magnitude, say so in the deck. "Bottom-up gives ₹34 crore; the published category figure implies a ceiling around ₹200 crore" is a stronger slide than either number alone, because it shows the work was checked rather than merely done.
What goes wrong
- Treating agreement between the two methods as proof, when both can share a wrong assumption
- Quoting a report for a category you are adjacent to rather than in
- Applying percentages nobody sourced and presenting the product as research
- Never revisiting the inputs after talking to the first ten customers
Is your number plausible?
Each check catches a different way the arithmetic can be right and the answer still wrong.
- If bottom-up exceeds the published category, one of your inputs is wrong — usually the buyer count or the price
- If the two methods differ by more than about 5x, find which assumption drives the gap before trusting either
- Write down which single input would most change the answer. That is the first thing to go and verify
Related answers
What if I genuinely cannot count my buyers?
That is a finding about the market, not a licence to go top-down. Consumer markets and undefined categories are genuinely hard to count — in those cases size a proxy you can count, such as installs of the tool they currently use, and label it as a proxy.
Are analyst reports ever worth buying?
Rarely for a pre-launch founder. The figure you get is a ceiling you already knew was high, at a price that would fund a month of customer interviews that would tell you something you did not know.
How do I show this in a pitch deck?
One slide, bottom-up, with the three or four inputs visible and sourced, and a single line noting the top-down ceiling. Investors read the inputs, not the total.