Research & PolicyFree
Market Sizing Model Bottom Up and Top Down
Send your pricing and the segment you sell into, get both builds, the ratio between them, and the gap decomposed by cause.
River builds the estimate twice. One build starts from a count of buyers in scope and multiplies by what one of them pays you. The other starts from a category total and applies the filters that narrow it to your slice. Then it divides one by the other and reports the ratio, and it takes the gap apart into the parts it can attribute to a specific cause and the part it cannot. Both builds ship with every input labelled counted, assumed or borrowed.
Search the query and every result tells you to build both ways and triangulate. None of them names the source you would count buyers from, or what that source leaves out. The Census Bureau's business statistics program covers only establishments that have paid employees. It publishes receipts only in years ending in 2 and 7, so a spend-per-firm ratio from it can be four years older than the count you multiplied it against.
Built for the founder writing the market slide, the corporate development analyst defending a segment forecast, and the product manager asked how big this actually is. It sits next to a competitor and market scan for the players already in the category. A listed comparable's filing read against last year's supplies real segment revenue, and the competitive intelligence report is where the estimate ends up in front of a decision. More research tools sit alongside it.
Two builds, 1.93 times apart, and one of them counted branches
Harbormark sells scheduling and compliance software to commercial mechanical contractors. The top-down build starts from a purchased category total of 2.4 billion dollars for North America. It takes 62 per cent for the United States, 14 per cent for the mechanical trades and 45 per cent for the mid-market tier, which lands on 93.7 million. The bottom-up build counts 41,800 establishments in scope and multiplies by a 4,320 dollar median contract, giving 180.6 million. The two are 1.93 times apart.
The first thing that comes out of the gap is not a judgement. A federal establishment count and a federal firm count are different columns in the same table, and software gets bought by the company, not by the branch. The same segment holds 33,900 firms across those 41,800 establishments, 1.23 establishments each, so the establishment build double counted 34.1 million dollars, 18.9 per cent of its own answer. Rebuilt on firms it reads 146.4 million, and the ratio falls to 1.56.
That leaves 52.7 million unexplained, and the honest treatment is to say which direction each remaining cause pushes rather than to close the gap. The purchased total counts only vendors its author tracks, which understates a tier priced below their inclusion floor. The establishment count also excludes businesses with no paid employees. The Census Bureau states that the majority of all business establishments in the United States are nonemployers, and here that removes buyers under the five-employee floor the product needs anyway.
How it works
Send what you sell
Your price points, the segment you serve, and any category total you have been handed.
Build it upward
A buyer count from named public sources, multiplied by what one buyer actually pays you.
Build it downward
The category total narrowed by each filter, with the source and the date of every percentage.
Reconcile the two
The ratio, the causes of the gap that can be named, and the residual left standing.
What you get
- Both builds side by side, with the ratio between them stated rather than averaged away
- Every input tagged counted, assumed or borrowed, so a reader knows which ones to attack
- The gap decomposed into named causes, with the part that cannot be attributed left open
- A firm count and an establishment count kept in separate columns, because buyers are companies
- Sensitivity on the three inputs that move the answer most, each halved and doubled
- A Sheet that recomputes when a price changes, so the forecast is not a screenshot
Common questions
Which of the two numbers do I actually put on the slide?
The bottom-up one, on firms rather than establishments, with the top-down figure beside it and the ratio stated. A single number invites the question of where it came from, and you answer it once instead of in the meeting. Where the two builds are more than roughly two times apart, the range is the finding and the point estimate is not ready.
Where do the buyer counts come from?
Federal business statistics by industry code, state licensing registers where the trade is licensed, and association membership counts, with the coverage boundary of each one written down. Every source excludes something. The run records what each one leaves out and which direction that exclusion pushes your estimate, rather than treating a public table as a census of your buyers.
My two builds are three times apart. Which one is wrong?
Usually neither, and the gap is the useful part. Check the unit first, because an establishment count and a firm count differ by a fifth in some trades. Then check whether the category total covers only vendors its author tracked, and whether your filters multiply percentages that were measured on different populations. A named cause beats a split difference.
Can it use the industry report I already paid for?
Yes, as one input among several, and it will treat the headline figure as borrowed rather than counted. Where the report does not publish the population, the base year or the vendor inclusion rule, the run says so on the sentence that uses the number. A figure you cannot reconstruct is still evidence, just weaker than a count.
Does it forecast growth as well as current size?
It projects from named drivers rather than applying a compound growth rate to the total. Establishment counts, employment in the segment, price changes you control and adoption you have observed each get their own line, so a reader can disagree with one driver without discarding the whole forecast. Pure category growth rates get labelled as borrowed.
What if my category has no public data at all?
Then the bottom-up build carries the estimate and the page says the top-down check was not available, which is a stronger position than a manufactured category total. Proxy populations work when the proxy is named and its relationship to your buyer is stated. A market map is where an undefined category usually needs to start.
Market Sizing Model Bottom Up and Top Down
Fill in the form and your workspace opens with the work already underway.