Strategy & Research · Agent workflow
TAM, SAM and SOM with every assumption and source on the page
ZeroTwo drafts the model in Google Sheets and the source log in Notion: a top-down and a bottom-up estimate, each assumption written down, each input linked to where it came from, and a range instead of a single number. It starts from public data and your own licensed reports, and a person reviews the result before a board or an investor sees it.
TAM
accounts × annual contract value
SAM
TAM × share you can actually serve
SOM
SAM × share you can win in the period
Two ways to build the same number
A credible sizing is a structured argument built on public evidence, not a guess. One method alone hides its weak assumption. Two methods that disagree show you where to look.
Start wide, narrow down
Top-down
- Find a published size for the broad category
- Cut it to the segment, region and buyer you serve
- Apply the share you can reach, then win
Main risk: The category definition is rarely your category, so the starting number can be too wide.
Start from the unit
Bottom-up
- Count the buyers who fit your definition
- Multiply by what one buyer pays per year
- Apply the share you can reach, then win
Main risk: The buyer count and the price are assumptions you have to defend one by one.
Reconcile before you report
Put the two totals side by side and write down what would have to be true for them to match. The explanation is the most useful part of the analysis, and it is the part a reviewer will ask about.
Where the numbers come from
Public data only unless you add your own licensed reports. The agent logs every input, and the stronger sources come first. For broader research beyond sizing, see the AI market research tool.
Company filings
Revenue and segment disclosure from listed players, where a company reports it.
Government statistics
Census and statistical-agency tables for counts of firms, people and spend.
Industry associations
Published surveys and member counts. Check how the sample was built.
Publicly quoted analyst figures
Excerpts and press releases only. Paywalled reports are out of reach unless you supply them.
Proxies
Adjacent markets, job-posting volume, funding data. Useful when direct data is missing, and labelled as soft.
The arithmetic, with toy numbers
These inputs are invented round numbers to show how the formulas chain and how one assumption flows through to SOM. They are not market data and not a result from ZeroTwo.
| Step | Formula | Low | Base | High |
|---|---|---|---|---|
| TAM | 1,000 accounts × $10,000 | $10.0M | $10.0M | $10.0M |
| SAM | TAM × reachable share (30% / 40% / 50%) | $3.0M | $4.0M | $5.0M |
| SOM | SAM × 5% won over the period | $150K | $200K | $250K |
Reconciliation
A toy top-down route, $60M category spend × 20% segment fit, gives $12.0M. The bottom-up TAM above is $10.0M. The $2.0M gap is the thing to explain.
Sensitivity
Moving only the reachable share from 40% to 30% or 50% moves SOM from $200K to $150K or $250K. One assumption, a 25% swing either way.
Why it matters
A single point estimate hides this. A range with the driving assumption named lets a reader decide how much weight the number can carry.
The workbook to ask for, tab by tab
Name the structure in your prompt so the output is inspectable. Formulas stay live, so a reviewer can change an assumption and watch the total move.
- Assumptions
Every input with its unit, definition, value and who decided it. The tab a reviewer reads first.
- Sources
One row per data point: link, publisher, date, geography and a confidence label.
- Top-down
The category figure and each narrowing step, as formulas you can trace.
- Bottom-up
Buyer count, price per buyer and the same reach and win shares, as formulas.
- Reconciliation
The two totals side by side and the gap between them, with the assumption suspected of causing it.
- Sensitivity
Low, base and high for the inputs that move the answer most.
How ZeroTwo conducts market sizing research
Run on demand. Time depends on the market, how much data exists and the model you choose.
Gather public inputs
Filings, statistical tables, association reports and publicly quoted estimates. Each input is logged with a link, a date and a unit.
Build both models
A top-down and a bottom-up estimate in Google Sheets, with assumptions kept on their own tab.
Flag the proxies
Where direct data does not exist, the input is labelled a proxy and the reason is written next to it.
Run the sensitivity
Vary the assumptions that move the answer most, such as growth rate, reach and price, and report a range instead of a single point.
- You review
Write the memo, then hand it to a person
Methodology, citations, confidence per input and open questions for the analyst, delivered to Notion. A person reviews it before it is used.
What the agent drafts and what stays with you
The agent drafts
- Searches public sources and logs each input with link, date and unit
- Builds the top-down and bottom-up models as formulas
- Labels proxy inputs and says where direct data is missing
- Runs low, base and high cases on the assumptions you name
- Drafts the methodology and citation memo
You decide
- Open the cited links and confirm each number appears there
- Check the units, year and geography on every input
- Confirm the segment definition matches the buyers you mean
- Explain the gap between top-down and bottom-up instead of averaging it
- Replace proxies with licensed or primary data where you can
- Have an analyst sign off before a board or investor sees a figure
A prompt to start from
Swap in your market and buyer. Naming the output structure is what makes the result checkable.
Size the North American AI-powered customer support market. Buyers are VP Support and CX leaders at mid-market SaaS companies. Build top-down and bottom-up models in Google Sheets, log every source in Notion with link, date and unit, label proxy inputs, show the gap between the two methods and add a low, base and high sensitivity case.
Delivers to Google Sheets and Notion when both are connected
Questions about AI-assisted market sizing
How credible is the data?
The agent prefers primary and official sources, such as company filings and government statistics, over secondary summaries, and marks weaker inputs as lower confidence. Credibility still depends on what is publicly available for your market, so open the links yourself. The source log exists to make that quick.
Does it produce both top-down and bottom-up estimates?
Ask for both, plus the reconciliation. When the two methods disagree, the gap is information: it points to the assumption worth investigating, so treat it as something to explain rather than average away.
Can it use paywalled reports from firms such as Gartner or Forrester?
It works from public data by default. Paywalled reports are not available to it unless you supply them, and publicly quoted excerpts are partial. If you hold a licence for a report, attach it, ask the agent to cite it separately and keep to the licence terms.
What happens when the market has little data?
Use proxy indicators such as adjacent market sizes, growth in analogous markets, job-posting volume and funding data. Ask the agent to label each proxy and say why it was chosen. The estimate for such a market is wider, and the sensitivity range should show that.
Can it refresh an earlier market sizing?
Give it the previous sheet and ask it to re-check each input against its cited source while keeping the structure. Review every changed input and definition afterwards, because a restated source can quietly change the answer.
Can I put the output in a pitch deck or board pack?
Use it as an analyst draft. Estimates built from public data are starting points. Before a board or investor sees a figure, a person should check the sources, the definitions and the assumptions, and present the range with its caveats. We do not claim the output is investor-ready.
Related research workflows
This page owns the numerical sizing deliverable. See all agent use cases.
- Competitor Intelligence MonitorTracks competitor pricing, launches and shifts over time, which is a different job from sizing a market once.Open
- Funding & M&A MonitorFollows funding rounds, acquisitions and IPO filings, useful as a source of proxy signals.Open
- Best AI for market researchIf you are choosing a product for broader research, start with the comparison.Open
Size one market and audit the sources yourself
100 free credits to start on the Free plan. Work agent access starts on Plus at $14.99/mo. See pricing.