TAM, SAM & Market Sizing

The framework — strategy first

Market sizing — the foundation under every revenue plan.

Why market sizing precedes everything else.

THE QUESTION YOUR IDEAL CUSTOMER PROFILE CAN’T ANSWER

An Ideal Customer Profile (ICP) tells you who you sell to. It does not tell you whether enough of them exist to support the revenue plan. Most early-stage marketing functions skip the market-sizing layer because it feels like finance work — and then six quarters in, the board asks why pipeline coverage is short and the answer is structural: the SAM was never big enough to underwrite the plan in the first place.

Market sizing asks — and answers — the three questions that anchor every downstream decision. Get them right and the rest of the playbook compounds. Get them wrong and every channel is over-budgeted against a target population that doesn’t exist.

Three rules that determine whether a market is worth the company’s effort. First: the market has to exist — there are real buyers with real budgets actively spending in this category. Not could-spend, not should-spend, actually-spend. Second: you have to be able to reach them — given your motion, geography, language coverage, regulatory posture, and integration footprint, the buyers are physically and operationally accessible. Third: your capture rate, multiplied by the reachable market, has to be larger than the revenue plan — with comfortable headroom. If the math doesn’t close, the plan needs to change or the addressable market needs to expand.

The three layers — TAM, SAM, SOM.

THE CANONICAL THREE-LAYER MARKET-SIZING FRAMEWORK

Each layer narrows the universe by removing a class of buyers who don’t belong in the next layer’s population. The discipline is naming the filters explicitly — not picking numbers because they look good in the deck.

LAYER	WHAT IT MEANS	WHO IT’S FOR	THE DOLLAR SIZE
TAM — Total Addressable Market	Every account in the world that could ever buy your category — ignoring whether you can reach them, sell them, or serve them today.	Board decks, fundraising narratives, category-creation arguments, long-horizon planning.	The biggest number in the deck. Useful for narrative; almost never useful for operating decisions.
SAM — Serviceable Addressable Market	The slice of TAM you can actually serve today — given your motion, geography, language, regulatory clearance, and integration footprint.	Annual planning, channel mix, headcount math, the budget conversation with the CFO.	The honest number. The one the marketing function is actually competing inside of.
SOM — Serviceable Obtainable Market	The slice of SAM you can realistically capture in the next 3–5 years — given your share-of-voice, win rate, capacity, and competitive pressure.	Revenue plan, quota math, pipeline-coverage targets, fundraising horizon.	The most-important-but-least-discussed number. If your revenue plan is bigger than SOM, the plan is structurally unachievable regardless of execution quality.

The ratio test most operators don’t run: healthy B2B SaaS companies have a SAM that’s 20–50% of TAM and a SOM that’s 5–15% of SAM. If your SAM is 90% of TAM, you’ve probably defined SAM too loosely — you’re counting buyers you can’t actually serve. If your SOM is 40% of SAM in three years, you’ve probably set expectations too high — that’s a market-dominant capture rate, not a startup capture rate. Both errors get caught by leadership eventually; better to catch them upstream in the model.

How to size TAM — top-down meets bottom-up meets triangulation.

Single-source TAM numbers are wrong. Top-down (industry reports) and bottom-up (account count × Annual Contract Value, or ACV) almost always disagree, and the disagreement IS the signal — it tells you which assumptions are weakest. Run both, then triangulate.

Top-down sizing — the analyst lens

Top-down sizing starts with the published industry sizing from an analyst (Gartner, Forrester, IDC) or category report, then applies your category-share assumption to extract your TAM slice. The pattern:

Pick the category sizing. “The global workforce-management software market is $X billion in 2026, growing at Y% CAGR.”
Apply your sub-category share. “Frontline-operations workforce management is approximately Z% of the category, so $X × Z%.”
Apply your geography filter. “North America is 40% of that sub-category by spend, so the geographic TAM is the previous number × 40%.”

Top-down is fast and credibility-rich (the analyst report is the citation), but it’s rough — analyst sub-category definitions rarely match your product’s exact addressable surface.

Bottom-up sizing — the buyer-by-buyer lens

Bottom-up sizing builds TAM by counting the accounts that could buy and multiplying by an average ACV ceiling:

Count the accounts. Source the universe of accounts matching your category-relevance criteria from a firmographic database (ZoomInfo, Apollo, Crunchbase, Clearbit, public sources, government registries for regulated industries). Be specific: “US multi-unit restaurant chains with 50+ locations = 4,200 accounts.”
Estimate the per-account ACV ceiling. What would a fully-penetrated account spend with you? Not what they spend today — what they would spend if every relevant product line was deployed. “Avg ACV ceiling per 50+ location chain = $185K/year.”
Multiply. 4,200 × $185K = $777M TAM.

Bottom-up is buyer-grounded and defensible to the CFO (each input is auditable), but it’s slow and reliant on the firmographic data being accurate.

The triangulation discipline

THE 30% RULE

Top-down and bottom-up TAM estimates should be within 30% of each other. If they’re closer than 30%, you have a defensible number. If they’re further apart than 30%, one of your assumptions is wrong — almost always either the sub-category share (top-down) or the ACV ceiling (bottom-up). Dig until you can explain the gap.

A worked example. Top-down says the addressable workforce-management market for multi-unit restaurants in North America is $1.1B. Bottom-up says 4,200 accounts × $185K = $777M. The gap is 30%. The top-down number includes a slice of accounts (large enterprise chains, 500+ locations) that the bottom-up filter excluded. Reconcile: either bring those large enterprise chains into the bottom-up count, or remove them from the top-down attribution. The reconciled number ends up at $850M with 8% uncertainty — tight enough to defend.

How to size SAM — the capability filters.

SAM is TAM filtered by what your company can actually serve today. Each filter removes a slice of accounts:

FILTER	WHAT IT REMOVES	HOW TO ESTIMATE THE IMPACT
Geography & language	Accounts in markets you don’t sell in, languages you don’t support, time zones your support team can’t cover.	For each geography in TAM, ask: do we have sales coverage? Customer success coverage? Localized product? Cut the accounts where any answer is no.
Motion (Product-Led Growth vs. sales-led)	Accounts that don’t fit your motion — Product-Led Growth (PLG) products can’t serve enterprise procurement well; sales-led products can’t serve self-serve small and medium businesses (SMB) cost-effectively.	Filter to the company-size band that matches your motion. PLG: SMB & mid-market. Sales-led: mid-market & enterprise.
Regulatory posture	Accounts in regulated industries you don’t have compliance for — healthcare without HIPAA, financial services without SOC 2 Type II, public sector without FedRAMP.	Cross-reference your compliance certifications against the regulatory requirements of each vertical. Remove accounts in industries where you can’t legally operate.
Integration footprint	Accounts on platforms your product doesn’t integrate with — the wrong CRM, the wrong POS, the wrong identity provider.	For each must-have integration, count the percentage of accounts in TAM that use the alternative platforms you don’t support. Cut those accounts.
Buyer profile	Accounts where the buying committee doesn’t match your sales motion — a tech-buyer-led purchase pattern in a frontline-ops product won’t close.	Filter to accounts where your typical buyer role exists at the typical seniority. Cut accounts where the role is absent or the seniority is below your decision-maker threshold.

The discipline is documenting each filter and the % impact. The SAM that survives all five filters is your real serviceable market — the population every channel should target.

How to size SOM — the realistic 3-year capture.

SOM is the slice of SAM you can win in the next 3–5 years given your share of voice, win rate against competitors, and capacity to serve. The math is straightforward but the inputs are honest:

THE SOM CALCULATION

SOM = SAM × (annual share-of-voice %) × (win rate against incumbents %) × (years in plan).

Worked example: SAM = $850M. Share-of-voice = 8% (you reach 8% of the SAM with your demand-gen engine annually). Win rate = 35% (when you’re in a deal, you close 35%). Plan horizon = 3 years. SOM = $850M × 8% × 35% × 3 = $71M. That’s the realistic three-year capture.

Three honest checks against the SOM number. Does the revenue plan fit within SOM with at least 30% headroom? If revenue plan is $60M and SOM is $71M, you have 18% headroom — tight. If revenue plan is $40M and SOM is $71M, you have 78% headroom — comfortable. Does the share-of-voice (SOV) assumption match the marketing budget? Doubling SOV usually requires doubling demand spend; if the budget can’t support the SOV assumption, the SOM is fiction. Does the win-rate assumption match the win/loss data? If win/loss shows you closing 22% against incumbents and the SOM uses 35%, the model is over-forecasting by 60%.

The CFO defense pattern.

The market-sizing model is only useful if the CFO believes it. The model that holds up under scrutiny has these traits:

Every input is sourced. Account counts cite the firmographic provider. ACV ceilings cite comparable customer expansion benchmarks. Win rates cite the CRM. Nothing is “the marketing team estimated.”
Both top-down and bottom-up are present. The CFO sees the analyst sizing AND the bottom-up build. They reconcile or they don’t — either way, the model is honest.
Filters are documented. The path from TAM to SAM is explicit: each filter, each percentage impact, each surviving account count.
Sensitivity analysis is shown. What happens to SOM if SOV is half what we assumed? Win rate two-thirds? Sales cycle longer? The model should answer these without you needing to rebuild it on the fly.
The refresh cadence is defined. “We re-run this quarterly against actuals.” A static model from 18 months ago is not defensible.

The four common sizing mistakes.

FAILURE MODES TO WATCH FOR

Counting non-buyers. The TAM count includes accounts that match your firmographic but have zero budget for your category. Common in early-stage when the team is over-eager.
Using consumer data for B2B. Industry reports often blur B2B and B2C sizing. Make sure the report’s sub-category is actually your sub-category.
Ignoring product limits. The model assumes your product can serve every account in SAM. In reality, your product has scale ceilings (per-seat performance, data-volume limits, integration depth). Acknowledge them or the SOM is fiction.
Letting SAM = TAM. If your filters don’t materially shrink TAM down to SAM, you haven’t defined the filters honestly. Real SAM is 20–50% of TAM in most B2B SaaS contexts.

Refresh cadence and ownership.

Market sizing is not a one-time exercise. The model should refresh on a quarterly cadence against three signals: new firmographic data (account counts shift as companies are founded, acquired, or shut down), new ACV data (your average customer expansion shifts the per-account ACV ceiling), and new win/loss data (your competitive win rate shifts the SOM math).

The owner of the model is the head of marketing (or, in larger organizations, the Product Marketing Manager lead). The Chief Financial Officer signs off each quarter. The board sees an updated TAM, SAM, and SOM slide every six months at minimum, with sensitivity analysis if any assumption has shifted by more than 15%.

CAPTURE IN YOUR OPERATOR BRIEF — SECTION 1

The brief-builder cards below collect each input: TAM size + sourcing, SAM size + filters, SOM size + horizon assumptions, market growth rate + key trends. Once Section 1 is populated, every downstream prompt — from Ideal Customer Profile to channel mix to budget — pulls your real market math.

Brief-builder: your TAM

TAM — total addressable market

The full universe of accounts that could buy your category. Save to Brief Section 1.1 — available as [TAM] in every downstream prompt.

TAM in $ (annual)

Top-down sizing source & assumption

Bottom-up sizing — account count × ACV ceiling

Saved to Brief Section 1.1.

Brief-builder: your SAM

SAM — serviceable addressable market

The slice of TAM you can actually serve today, given your motion / geography / regulatory / integration footprint. Save to Brief Section 1.2.

SAM in $ (annual)

The five capability filters and their % impact

SAM as % of TAM (sanity check — should be 20–50%)

Saved to Brief Section 1.2.

Brief-builder: your SOM

SOM — serviceable obtainable market (3-year horizon)

The realistic capture over the next 3–5 years given share-of-voice, win rate, and capacity. Save to Brief Section 1.3.

SOM in $ (3-year cumulative)

Share-of-voice assumption

Win rate against incumbents (from CRM)

Headroom against revenue plan (sanity check — should be 30%+)

Saved to Brief Section 1.3.

Brief-builder: market growth & trends

Market growth rate and key trends

The macro context every campaign and board update inherits from. Save to Brief Section 1.4.

Category CAGR (compound annual growth rate)

Three macro trends shaping demand

Next refresh date (quarterly cadence)

Saved to Brief Section 1.4.

The prompt pack

Five prompts to run the sizing.

For teams running the work themselves — or the human-in-the-loop pass before the agent runs at scale. Every prompt below uses your Brief fields; the bracketed tokens substitute your real values as soon as the Brief is populated.

Prompt 1

TAM top-down sizing from analyst data

Builds the top-down TAM estimate from a published industry report, sub-category share, and geography filter. Returns a defensible number with the reasoning chain documented.

You are a market-sizing analyst for [COMPANY NAME]. Build the top-down TAM estimate for our category. I am providing: - Category sizing from analyst report: $[CATEGORY SIZE], source [ANALYST SOURCE], year [YEAR] - Sub-category that matches our product: [SUB-CATEGORY DEFINITION] - Geography we sell in: [GEOGRAPHY] - Languages we support: [LANGUAGES] For the top-down TAM, walk me through: 1. The total category number from the analyst source. 2. Your estimate of our sub-category share of the category (% with sourcing/reasoning). 3. Your estimate of the geography share of the sub-category (% with sourcing). 4. The math: category $ × sub-category share × geography share = top-down TAM. 5. Confidence level (high/medium/low) and the single weakest assumption. Return as a markdown table I can paste into the board deck, with sources cited for each input.

Prompt 2

TAM bottom-up sizing from account count + ACV ceiling

Builds the bottom-up TAM by counting addressable accounts in a firmographic database and multiplying by a defended ACV ceiling. The number the CFO will reconcile against the top-down estimate.

You are a market-sizing analyst for [COMPANY NAME]. Build the bottom-up TAM estimate for our category. I am providing: - Firmographic filters for category-relevant accounts: [FIRMOGRAPHIC FILTERS] (e.g., industry SIC code, employee count band, geography, integration usage) - Per-account ACV ceiling assumption: $[PER-ACCOUNT ACV CEILING] - Sourcing for the ACV ceiling: [SOURCING — benchmark customer, public comp, internal pricing tier] - Firmographic database source: [DATABASE NAME] For the bottom-up TAM, walk me through: 1. The exact filter set you applied to count addressable accounts. 2. The account count surviving each filter, step by step. 3. The final addressable account count. 4. The per-account ACV ceiling reasoning — what would a fully-penetrated account spend? 5. The math: account count × ACV ceiling = bottom-up TAM. 6. Confidence level (high/medium/low) and the single weakest assumption. Return as a markdown table with the filter chain and per-step account count, then the final ACV math.

Prompt 3

TAM → SAM filter chain

Documents the five capability filters that narrow TAM to SAM, with the % impact of each filter and the surviving account count after each. The audit trail the CFO and the board will both want.

You are a market-sizing analyst for [COMPANY NAME]. Apply the five-filter SAM methodology to our TAM. Our TAM = $[TAM]. Our company profile: - Geography we sell in: [GEOGRAPHY] - Motion: [GTM MOTION] - Regulatory clearances we hold: [REGULATORY POSTURE] - Integrations we ship: [INTEGRATIONS] - Typical buyer profile: [BUYER PROFILE] For each of the five filters (geography, motion, regulatory, integration, buyer profile), give me: 1. The accounts removed by this filter and the reasoning. 2. The % of TAM that the filter removes. 3. The surviving account count after this filter. 4. The cumulative SAM running total in $. Return as a sequential filter-chain table. The final row is our SAM in $. Then sanity-check: is SAM between 20% and 50% of TAM? If not, which filter is misdefined?

Prompt 4

SOM calculation + revenue-plan headroom check

Builds SOM from SAM, share-of-voice, win rate, and plan horizon. Checks whether the revenue plan fits within SOM with safe headroom. The single most-important sanity check on the company plan.

You are a market-sizing analyst for [COMPANY NAME]. Build the SOM for our 3-year plan. Inputs: - SAM in $: $[SAM] - Annual share-of-voice assumption: [SOV %] - Win rate against incumbents (from CRM): [WIN RATE %] - Plan horizon: 3 years - Revenue plan target for the 3-year window: $[REVENUE PLAN] Walk me through: 1. The SOM calculation: SAM × SOV × win rate × years = SOM in $. 2. The revenue plan vs. SOM ratio. 3. The headroom: how much of SOM is unused after the revenue plan? 4. Whether the plan is comfortable (30%+ headroom), tight (10–30% headroom), or over-forecast (under 10% headroom). 5. Sensitivity analysis: what happens to SOM if SOV is half what we assumed? Win rate two-thirds? Both? Return as a markdown table with the SOM math, the headroom check, and a sensitivity table.

Prompt 5

CFO defense memo

Drafts the one-page memo the head of marketing hands the CFO before the quarterly review. Walks through the model, the sources, the sensitivity, and the refresh cadence in language the CFO will sign.

You are the head of marketing at [COMPANY NAME] drafting a defense memo for the CFO. Brief context: - TAM: $[TAM] (top-down $[TOP-DOWN TAM], bottom-up $[BOTTOM-UP TAM]) - SAM: $[SAM] (filter chain documented in Section 1.2) - SOM: $[SOM] over 3 years - Revenue plan: $[REVENUE PLAN] over 3 years - Headroom: [HEADROOM %] - Voice: [VOICE DOS], avoid [FORBIDDEN LANGUAGE] Draft a one-page memo to the CFO covering: 1. The model in two sentences — what was sized and how. 2. The reconciliation between top-down and bottom-up TAM. 3. The five SAM filters and which one removes the most accounts. 4. The SOM calculation and the revenue-plan headroom. 5. The single highest-risk assumption and the planned monitoring of it. 6. The quarterly refresh cadence. Length: one page maximum. Operator-direct voice. No hedging. Numbers sourced.

The agent

Market Sizing Agent — the agent that keeps the model honest.

For teams operating in AI Operating Model territory, the five prompts above are orchestrated by one agent that watches the inputs (firmographic feeds, CRM win-rate data, marketing-spend ledger) and surfaces drift before the model goes stale.

Market Sizing Agent

Keeps TAM / SAM / SOM defensible to the CFO. Re-runs the model quarterly with named sources, surfaces saturation early, and exposes the single assumption that, if wrong by ±20%, would change the budget recommendation.

Who is this agent

Identity card

NameMarket Sizing Agent

RoleQuantitative market sizing — the math the budget conversation depends on

OwnerVP Marketing (with CFO partnership)

Reports toVP Marketing + CFO

Versionv0.5 (supervised)

SurfaceClaude Project + Postgres (historical sizing snapshots) + Python for bottom-up calculations

Output target/market-sizing/quarterly/Q<n>.md + /market-sizing/saturation-tracker.json

Review cadenceQuarterly refresh; monthly saturation watch; ad-hoc on major market events

Mission

Maintain a TAM / SAM / SOM model the CFO defends in front of the board. Show the math, cite the sources, expose the assumptions, surface the single assumption that — if wrong by ±20% — would change the budget recommendation. Recompute quarterly against fresh data. Watch saturation curves so the marketing function knows when expansion-into-adjacency beats penetration-of-core.

Goals & KPIs the agent moves

Leading indicators — the agent controls these

Sourcing rigor: every TAM / SAM / SOM number cites an analyst report, government data set, or named CRM query100% — zero unsourced numbers

Quarterly refresh shipped within 14 days of quarter close, with sensitivity analysis on the riskiest assumption100% on-time

Lagging indicators — downstream outcomes with review triggers

SOM forecast accuracy vs. actual closed-won (4-quarter rolling). Trigger: drift > 25% in any single quarter pages the CFO + VP Marketing for an assumption audit.Within ±15%

Saturation early warning lead time. Trigger: SAM penetration crosses 25% without ≥ 2 quarters of prior warning pages the VP Marketing for a methodology review.≥ 2 quarters of lead time

What it does

Task list

Daily Pull the firmographic data refresh (LinkedIn Sales Nav, ZoomInfo, Clearbit, or equivalent). Append to the SAM-eligible account universe.
Weekly Cross-reference new CRM closed-won accounts against the SAM definition. Flag any closes outside SAM as ICP-drift signals.
Monthly Compute saturation: SOM-to-date / current-quarter-SAM. Surface if trending toward 25% (the saturation watch threshold).
Monthly Watch analyst report releases (Gartner, Forrester, IDC). Surface any release that materially shifts the TAM number.
Quarterly Run the full TAM / SAM / SOM refresh. Three tiers, math exposed, sources named, sensitivity analysis on the riskiest assumption.
Quarterly Draft the CFO defense memo — one page, math + sources + risks + the single sensitivity check.
Quarterly Compare SOM forecast (prior year) to closed-won actuals. Surface forecast accuracy.
Quarterly Run the beachhead analysis: which sub-segment inside SAM has the highest win-rate × ACV? Recommend beachhead concentration for the next quarter.
Event When a major analyst report drops or a category competitor announces a number, re-run the assumption that’s most affected.
Event When Account Intel Hub portfolio review surfaces ICP drift, work with the ICP Researcher Agent to update SAM filters before next quarterly run.

Schedule grid

Task	Frequency	Duration	Output goes to
Daily firmographic refresh	Daily 03:00	~15 min	SAM universe table
Weekly CRM cross-reference	Weekly Mon 09:00	~20 min	Director MarOps + ICP Researcher
Monthly saturation watch	Monthly 1st	~30 min	VP Marketing + CFO
Monthly analyst report watch	Monthly 15th	~45 min	VP Marketing
Quarterly TAM/SAM/SOM refresh	Quarterly Q+10 days	~6 hours	CFO + VP Marketing + Board
Quarterly CFO defense memo draft	Quarterly Q+12 days	~3 hours	CFO (review + sign-off)
Quarterly forecast accuracy review	Quarterly Q+14 days	~2 hours	CFO + VP Marketing

Triggers

Scheduled (cron-style):

Schedule	What it runs
`0 3 * * *`	Daily firmographic refresh
`0 9 * * 1`	Weekly CRM cross-reference
`0 9 1 * *`	Monthly saturation watch + analyst watch
`0 9 10 1,4,7,10 *`	Quarterly TAM/SAM/SOM refresh

Event-driven:

Event	What it runs
Gartner / Forrester / IDC publishes a category report	Re-evaluate affected TAM assumptions within 5 business days
Closed-won account outside the declared SAM	Flag for ICP-drift discussion at next quarterly review
Saturation tracker crosses 20% (warning band)	Page VP Marketing; trigger beachhead-adjacency analysis
ICP Researcher updates the ICP definition	Re-run SAM filters within 2 weeks; flag impact on SOM math

Who it works with

Inputs

Source	Type	Cadence	Required?
Operator Brief (Sections 1, 2, 7)	Markdown	Read every run	Required — ICP + KPIs anchor the math
Firmographic data provider (ZoomInfo / Clearbit / LinkedIn Sales Nav)	API	Daily	Required
Analyst report subscriptions (Gartner / Forrester / IDC)	PDFs + web	Monthly watch	Required for category-level TAM
CRM closed-won + opportunity data	API	Real-time	Required for SOM ground-truth
Revenue Attribution Engine output	Markdown / JSON	Weekly	Required for win-rate ground-truth
Account Intel Hub portfolio patterns	Markdown	Monthly	Required for ICP-drift signal
Government data (NAICS / SIC / Census Business Patterns)	CSV / API	Annual refresh	Required for bottom-up TAM

Outputs

Output	Format	Target path	Audience
Quarterly TAM/SAM/SOM model + memo	Markdown + spreadsheet	/market-sizing/quarterly/Q<n>.md	CFO + VP Marketing + Board
Saturation tracker	JSON + chart	/market-sizing/saturation-tracker.json	VP Marketing + CFO
Monthly saturation digest	Markdown	/market-sizing/saturation/YYYY-MM.md	VP Marketing
Beachhead recommendation	Markdown	/market-sizing/beachhead/Q<n>.md	VP Marketing + Sales Director
ICP-drift alerts (when closed-won outside SAM)	Slack DM	Slack DM to ICP Researcher + VP Marketing	ICP Researcher Agent + VP Marketing

↑ Upstream — agents/sources that feed this one

Operator Brief (human-maintained). ICP definition + KPI targets — the math is anchored to what the Brief declares.
ICP Researcher Agent. Owns the ICP definition that filters TAM → SAM.
Revenue Attribution Engine. Win-rate ground-truth for SOM math.
Account Intel Hub. Portfolio patterns that may signal ICP drift.
Market Intelligence Agent. Competitor share-of-market intel that informs SOM.

↓ Downstream — agents/humans that consume its output

CFO (human). Defends the number to the board; signs off on quarterly memo.
VP Marketing (human). Uses the model to defend the budget envelope.
Budget Allocation Agent. Uses SOM-to-pipeline-conversion ratio to validate budget asks.
Brief Sync Agent. Receives ICP-drift signals to propagate back to Brief Section 2.
Pipeline Math Agent. Uses SOM math to validate pipeline gap calculations.

Human escalation paths

Trigger condition	Escalate to	Within
Quarterly forecast accuracy > ±25% off actuals	CFO + VP Marketing + Board	Before next quarterly Board review
Saturation tracker crosses 25% before warning was raised	VP Marketing + CFO	Immediate (escalation failure)
Major analyst report invalidates a primary assumption	VP Marketing + CFO	< 5 business days
3+ closed-won outside declared SAM in a quarter	VP Marketing + ICP Researcher + CFO	Same week (ICP-drift signal)
Quarterly memo missed Q+12-day deadline	VP Marketing + CFO	Immediate (board prep dependency)

How to build it

System prompt

You are the Market Sizing Agent for [COMPANY]. YOUR JOB Maintain a TAM / SAM / SOM model the CFO defends in front of the board. Show the math. Cite the sources. Expose the assumptions. Recompute quarterly. Watch saturation. Never produce an unsourced number. INPUTS (always read in this order) 1. /operator-brief.md - ICP + KPIs anchor the math 2. /firmographics/sam-universe.csv - daily firmographic refresh 3. /crm/opportunities.json - closed-won ground truth 4. /attribution/per-channel.json - win-rate ground truth 5. /market-sizing/prior-quarter.md - prior model for delta analysis OUTPUTS - /market-sizing/quarterly/Q<n>.md (quarterly TAM/SAM/SOM + CFO memo) - /market-sizing/saturation-tracker.json (continuous) - /market-sizing/beachhead/Q<n>.md (quarterly) RULES 1. Every TAM / SAM / SOM number cites: analyst report + year + page, OR gov data source + field, OR our own CRM with date. 2. Show the math, not just the result. TAM = (NAICS count) x (avg employee count) x (penetration rate) x (ACV). 3. SAM filters explicit: vertical fit, employee count, geography, regulatory. 4. SOM = SAM x realistic 3-year win rate, with the win-rate basis cited (competitor share-of-market or our historical win rate). 5. End every memo with: the single assumption that, if wrong by +/-20%, would change the budget recommendation - and how to validate it. 6. Saturation warning at 20%; saturation alert at 25%. 7. Never extrapolate beyond the data. If the analyst report covers the wrong geography, say so and surface the gap. ESCALATION - Forecast accuracy >25% off: CFO + Board before next review. - Saturation crossed 25% before warning: page VPM + CFO immediately.

Tools & integrations

Platform / tool	Used for	Required?
Claude Project + Postgres	Historical sizing snapshots + reasoning surface	Required
Firmographic data API (ZoomInfo / Clearbit / LinkedIn Sales Nav)	SAM universe construction	Required
Salesforce / HubSpot API	CRM closed-won + opportunity data	Required
Analyst report library access (Gartner / Forrester / IDC)	Category-level TAM	Required if covered by analysts
Python + pandas	Bottom-up TAM math + saturation analysis	Required
Slack API	Saturation alerts + ICP-drift signals	Required
Spreadsheet (Google Sheets / Excel)	CFO-facing model file	Required — the CFO wants a sheet

Guardrails — what it must not do

Never produce an unsourced number. Every figure cites a specific source.
Never extrapolate analyst data beyond its declared scope (geography, segment, time window).
Never inflate TAM to make the SOM math look better — the CFO will catch it.
Never compress sensitivity analysis. The single-assumption check is the most-asked board question.
Honor data licensing — analyst reports are licensed; don’t share verbatim with parties outside the license.
Never modify the ICP definition. ICP changes go through the ICP Researcher Agent.
Never claim a category-creation TAM (“everyone who has this problem”) — SAM must be defensible.

Evals + hallucination defense

Evals — output quality checks:

Sourcing audit. Quarterly: random sample 10 cited numbers from the memo. Every one must trace to its declared source. Target 100% — zero tolerance for unsourced numbers.
Forecast-vs-actual accuracy. Annual: prior-year SOM forecast vs. actual closed-won. Target within ±15%.
Saturation early-warning lead time. When saturation alert fires, audit: did the warning come ≥ 2 quarters before actual saturation? Target 100%.
CFO memo defensibility. Quarterly: CFO rates the memo 1–5 on “could you defend this in a board meeting?”. Target average ≥ 4.5.

Hallucination defense — specific checkpoints:

Analyst data must cite report title, year, and page number. No “Gartner says” without specifics.
Gov data must cite NAICS code, dataset name, year. No “according to the Census” without specifics.
Win rates must trace to specific CRM stage transitions over a specific time window.
Penetration rates must be either competitor-disclosed (with citation) or our own historical (with date range).
When data doesn’t cover the question, surface the gap rather than estimate.

Maturity curve + first-run checklist

v0.1 — Manual-assistAgent helps compile the model on-request when the VP Marketing asks. No autonomous monitoring. Useful from day 1 for replacing manual sizing.

v0.5 — SupervisedDaily SAM universe refresh + monthly saturation watch + quarterly refresh. All outputs reviewed by VP Marketing + CFO before board prep. Default ship state.

v1.0 — Semi-autonomousAfter 4 quarters of clean evals + zero forecast misses > 15%, agent auto-publishes the quarterly memo. CFO review still required for board prep. Methodology changes always human-approved.

First-run checklist — 5 steps from spec to running agent:

Stand up the firmographic data integration + SAM universe table.
Author the SAM filter rules with VP Marketing + Sales Director. Each filter cites the source (Brief Section 2 ICP, regulatory list, etc.).
Wire the CRM + Revenue Attribution Engine inputs. Verify weekly closed-won cross-reference.
Run the first quarterly model end-to-end with VP Marketing + CFO. Tune assumptions. Lock the methodology.
Schedule the quarterly cadence. Subscribe CFO + VP Marketing to saturation alerts. Log every run in /market-sizing/agent-log.md.