Market sizing tam

Market Sizing (TAM / SAM / SOM)

Builds a market size a founder can defend under questioning: bottom-up first (count of buyers × price they'd pay), top-down only as a cross-check, and every single number carrying a receipt — a source URL, a user-provided figure marked as such, or an explicitly labeled assumption. The standard: no naked numbers. A TAM slide with unsourced figures is the fastest way to lose an investor's trust.

When to Use

• User needs TAM/SAM/SOM for a pitch deck, accelerator application, or go/no-go decision.
• User asks "is this market big enough to bother?"
• Stress-testing a market claim someone else made ("the report says $40B — is that real?").
• Not for: analyzing one competitor (competitor-teardown); tracking market signals over time (trend-radar); crunching a CSV of your own sales data (csv-data-analyst).

Quick Reference

Action	Command / Call
Bottom-up TAM	(# of potential buying units) × (realistic annual price) — both sourced
SAM	TAM × reachable fraction (geography, segment, channel) — each filter sourced
SOM	SAM × attainable share in 3 yrs — anchored to a comparable's actual trajectory
Buyer counts	census/labor stats, industry bodies, app-store counts, LinkedIn filters — record URL + date
Cross-check	independent top-down (analyst report) — flag if it diverges >3x from bottom-up
Sensitivity table	python heredoc, step 6 — pessimistic / base / optimistic per assumption

Procedure

1. Precheck — get from the user: the specific buyer definition ("who writes the check"), the price point (their actual or intended pricing), and geography/segment limits. If web tools are available, sources get fetched live; if not, every external number must come from the user with its origin named — otherwise it goes in as assumption with low confidence.
2. Define the unit — one sentence: "A buyer is <who>, paying <price> per <period> for <what>." Every later number hangs off this. Vague unit = fictional TAM.
3. Count buyers (bottom-up) — find the number of buying units from a primary-ish source: government stats, industry association counts, platform self-reporting, LinkedIn/job-board filter counts. Record the figure, URL, publication date, and what it actually counts (which is rarely exactly your buyer — state the mismatch).
4. Price the unit — use the user's real pricing, or comparable products' actual prices (source them). Annualize. TAM = buyers × annual price. Show the multiplication explicitly.
5. Filter to SAM and SOM — SAM: apply each reachability filter (geography, language, segment, channel) as a separate sourced or labeled percentage — no single mystery "×20%". SOM: anchor to evidence, e.g. a comparable company's year-3 revenue or share (sourced), or channel math (audience × conversion). "1% of the market" with no mechanism is banned.
6. Sensitivity table — vary the 2–3 weakest assumptions:

   python3 <<'PY'
   buyers = {"pess": 60_000, "base": 120_000, "opt": 200_000}   # source each in the report
   price  = {"pess": 240, "base": 480, "opt": 720}               # annual $, sourced
   reach  = {"pess": 0.05, "base": 0.12, "opt": 0.20}            # SAM filters combined
   for k in ("pess", "base", "opt"):
       tam = buyers[k] * price[k]
       print(f"{k:>4}: TAM ${tam:,.0f} | SAM ${tam*reach[k]:,.0f}")
   PY
   

7. Cross-check top-down — find one independent analyst/industry figure for the market. If it's within ~3x of bottom-up TAM, note agreement. If not, investigate which definition differs and say which you trust and why. Then deliver via the template.

Output Template

# Market Size — <product/market> — <date>
Buyer unit: <who> paying $<X>/yr for <what>

## Bottom-up
TAM = <N buyers> [source: <url>, <date>, counts: <what it actually counts>]
      × $<price>/yr [source: <url or "user's live pricing">]
      = $<TAM>
SAM = TAM × <filter 1: X% — source/label> × <filter 2: ...> = $<SAM>
SOM = SAM × <mechanism-based share — anchor: <comparable + source>> = $<SOM>

## Sensitivity
| Scenario | Buyers | Price | Reach | TAM | SAM |
|---|---|---|---|---|---|
| Pessimistic / Base / Optimistic rows |

## Cross-check
Top-down: $<X> [source] — <agrees / diverges because <definition difference>>

## Assumption register
| # | Assumption | Value | Receipt (url/user/assumption) | Confidence |

## Verdict
<Big enough for the goal? One paragraph, referencing the pessimistic case.>

Pitfalls

• Top-down laundering — "$50B market, we take 1%" survives zero investor questions. Recovery: bottom-up is the spine; top-down only cross-checks. If only top-down data exists, say so and mark the whole estimate low-confidence.
• A number without a receipt slips in — one unsourced "80M SMBs" poisons the whole model. Recovery: the assumption register is mandatory; before delivering, scan every numeral in the doc and confirm it appears in the register or derives from ones that do.
• Source counts the wrong thing — "number of US restaurants" used for "restaurants that would buy scheduling software". Recovery: for each source, write one line on the mismatch between what's counted and your buyer, and haircut accordingly with the haircut labeled as an assumption.
• SOM by wishful percentage — "we'll capture 5%" with no mechanism. Recovery: SOM must come from channel math or a named comparable's actual trajectory; otherwise report SOM as "not estimable yet" — that's more credible than fiction.
• Stale or circular sources — a 2019 blog citing a 2016 report citing a press release. Recovery: record publication dates; prefer primary sources; if the chain ends in a vendor press release, label it vendor-claimed and lower confidence.

Verification

• Every number appears in the assumption register with a receipt and confidence label
• TAM math shown explicitly (buyers × price), reproducible by hand
• Sensitivity table has all three scenarios with stated assumption ranges
• SOM anchored to a mechanism or comparable, not a naked percentage
• Top-down cross-check present, divergence explained if >3x