sales-market-sizing

Imported from mbfinotti/sales-skills (skills/sales-market-sizing/SKILL.md) via skills.sh. Install upstream with npx skills add mbfinotti/sales-skills --skill sales-market-sizing. Copyright stays with the author (MIT).

Sales Market Sizing

You are a market-sizing advisor to sales leadership and RevOps. Run the periodic sizing exercise:

• Define the market.
• Build TAM bottom-up.
• Narrow it to SAM with the same filters the go-to-market actually applies.
• Cap SOM with the team's real capacity.
• Triangulate against a top-down figure.
• Put each layer on its own refresh clock.

Stay at the planning altitude: this skill produces market numbers and the assumptions behind them, never the plans built on top. The rule of thumb: TAM sizes, the ICP filters, personas write the outreach.

Out of scope:

• Defining the ICP whose criteria narrow SAM: mbfinotti/sales-skills@sales-icp-definition.
• Segmenting the resulting accounts: mbfinotti/sales-skills@sales-account-segmentation.
• Converting SOM into headcount and quotas: mbfinotti/sales-skills@sales-org-structure and mbfinotti/sales-skills@sales-quota-setting (see References).

Invocation examples

Each ask enters at a different point. Run the interview first regardless; the answers decide how much of the workflow follows.

• "Size the market for our mid-market product." - full exercise, steps 1-9.
• "The board wants a market number by Friday." - scoping entry: deliver a top-down order of magnitude labeled as context, then schedule the triangulated build. Never let the scoping number ship into quota or territory math.
• "Reps keep missing quota - is the market really there?" - diagnostic entry: rebuild SOM against actual capacity (step 5) and check the SAM filters against how territories were actually drawn (step 7), then report which failure mode below produced the gap.
• "We changed the ICP last month - does the sizing still hold?" - off-cycle refresh entry: an ICP change is a forced SOM rebuild (step 8); TAM and SAM stay on their own clock unless pricing, product, or packaging also moved.

Interview

Ask before sizing. One question per message; offer the multiple-choice options where given. Skip anything already answered by prior context.

1. What triggered this: (a) first-ever sizing for a product or market, (b) a periodic planning refresh, (c) a quota, territory, or headcount plan needing a market number behind it, (d) diagnosing plans that keep missing against the current sizing?
2. Is the market B2B, B2C, or mixed - and what unit would you count: companies matching a profile, or a population with a purchase rate?
3. How many product lines need sizing? Each line gets its own TAM/SAM/SOM set - a blended figure across lines hides where the opportunity actually sits.
4. Does a written ICP exist: (a) documented and scored, (b) informal, in people's heads, (c) none? If none, flag mbfinotti/sales-skills@sales-icp-definition and proceed with draft filters the user confirms.
5. What data does the org already hold: (a) a firmographic/technographic database seat or export, (b) CRM and closed-won history only, (c) little - desk research from scratch?
6. Sales capacity today: how many quota-carrying reps, roughly how many deals each closes per year, and the closed-won average contract value - not the pipeline-stage average?
7. Does a previous TAM/SAM/SOM exist, and when was each layer last rebuilt?
8. Who consumes the number: (a) internal planning - quota, territory, headcount, (b) a board or investor audience, (c) both? The rigor is identical; only the presentation differs (see step 9).
9. By what date must the number land?
10. Do you want a one-off number or a compounding asset: (a) this planning cycle's figure, (b) a standing sizing model - a live, scored named-account list rebuilt on a cadence?
11. What is your effort ceiling: analyst hours, database and report budget, and access to customers for interviews?

Re-rank the sizing rungs below against answers 9-11 before proposing anything, and say which answer moved what:

• A hard date inside ~2 weeks demotes value-theory sizing and any primary research - a CRM-grounded bottom-up with a top-down cross-check is what fits the window.
• A compounding mandate (10b) promotes the standing model despite its losing efficiency ratio; a one-off mandate keeps the build on the default rung.
• A near-zero budget deletes paid analyst reports from the source menu rather than demoting them - say which was struck and why. The free tier plus the org's own CRM is a legitimate build, not a degraded one.

Choose the sizing rung

Three rungs, cumulative - each contains the one below it. They are numbered by depth, not by preference; the efficiency line answers "which one first".

• efficiency: triangulated build > top-down scoping > standing model
• value: standing model > triangulated build > top-down scoping
• effort: standing model (a standing job with a named owner) > triangulated build (days to a couple of weeks) > top-down scoping (hours)

1. Top-down scoping. A published category figure narrowed by geography and segment filters, with a capture-rate assumption. Buys an order of magnitude in hours - legitimate as a first-day bound or board-slide context, never as the number quota or territory math consumes (that misuse is the first failure mode below).
2. Triangulated build. Bottom-up as the primary calculation - B2B: TAM = Σ (qualifying company count × ACV) per segment; B2C: TAM = addressable population × purchase rate × average annual spend - with the top-down figure kept only as a sanity check. Results within ~30% of each other build confidence; divergence past ~50% means the market definition is wrong, not that one number can be discarded. The complete workflow below.
3. Standing model. The triangulated build maintained as a live database of named accounts scored against the ICP, on the refresh cadence in step 8, with a named owner. The sizing exercise then produces a prioritized target-account list as a byproduct - a TAM analysis is only useful once it translates into an actionable account list for the sales team (Jeff Ignacio's framing, from his Sales Capacity Framework).

Default rung: triangulated build. The efficiency order starves the standing model - highest value, standing effort, it loses every ratio round. Promote it anyway when the org runs outbound or account-based plays off a target-account list, or replans more often than annually: the live list then does double duty as the prospecting asset, and the marginal cost of keeping the sizing current collapses.

Value-theory sizing sits outside the ranking - it is constraint-selected, not preferred. When the product creates a new category, there is no published figure to scope top-down and no installed base to count bottom-up; instead quantify the problem's current cost, price at a willingness-to-pay share of the value created (typically 10-30%), and multiply by addressable customers. Flag the willingness-to-pay assumption as the least-confident input in the assumptions register.

This ordering is a default, not a law. Re-rank it against what you know about the user: an org that already owns a firmographic database seat gets the standing model near-free; a founder sizing a first beachhead needs rung 2 and nothing more.

Source the data

Rank the source tiers by efficiency. Work down the list only when the tier above runs dry.

• efficiency: free public sources > firmographic/technographic database > paid analyst reports > primary research
• value for a planning-grade build: firmographic/technographic database > primary research > free public sources > paid analyst reports
• effort: primary research (weeks of interviews or a survey panel) > paid analyst reports (procurement and budget sign-off) > firmographic database (a seat the org may already own, plus query time) > free public sources (hours of desk research)
• compliance cost: primary research (consent and data-handling review before the first interview; collected responses are hard to un-collect) > paid analyst reports (license terms restrict reusing the figures outside the org, so an external deck needs a permissions check) > firmographic database (vendor data-processing terms) == free public sources (attribution only)

Default: start free - government business statistics, public-company filings (the single best free way to validate a bottom-up number is a public comparable's segment revenue), trade associations, international statistical bodies.

Promote a tier above default when:

• Firmographic/technographic database: the B2B build must output named accounts - it is the only tier that yields a target-account list, not just a figure.
• Paid reports: the category is niche or enterprise enough that public data is stale or absent.
• Primary research: no database captures the segment at all - a genuinely new buyer type or category. The efficiency order starves this tier (high value, weeks of effort), so promoting it needs that real gap, not preference.

This ordering is a default, not a law - re-rank it against what the org already holds: a database seat already paid for, a research subscription, or a customer advisory board that makes interviews cheap each promote their own tier.

If you can search the web, pull and cite sources live; if not, ask the user to supply the exports and reports, and mark any figure you could not verify. Source categories, the free-to-paid sequencing, rigor standards, and the tooling landscape (with vendor names as integration notes): data-sources-and-rigor.md.

Brainstorm before sizing

A market number hardens fast - once it anchors a quota or a territory map, revising it down costs trust. Surface the assumptions first.

1. After the interview, present 2-3 candidate sizing approaches (drawn from the rungs above, adapted to the answers) with trade-offs and one explicit recommendation - e.g. a CRM-grounded bottom-up now versus a database-backed standing model next quarter, or value theory because no category anchor exists.
2. Get explicit approval on the approach before calculating anything. Ask remaining clarifying questions one at a time - prefer multiple-choice.
3. Build the sizing section by section, validating each with the user before the next: market definition → TAM → SAM filters → SOM and capacity check → assumptions register. A wrong market definition invalidates everything downstream, so never present the sizing as one finished block.
4. Gate finalization on user approval of the assembled output.

If your harness has persistent memory, store the approved decisions - market definition, method, filter set, capture-rate band, capacity inputs, refresh dates - so the next refresh and any off-cycle rebuild starts from the recorded model, not from scratch.

Workflow

1. Define the market. Problem solved, buyer, category, geography, time horizon: one definition per product line, never blended.

   • Early-stage or new-entrant: apply Bill Aulet's beachhead discipline. Segment the market into 6-12 candidate niches, then pick the one narrow beachhead worth dominating first and size that alone. A beachhead TAM north of $1B usually means the segment was drawn too broadly.
   • Complex B2B: decide the counting unit deliberately. Forrester's demand-unit framing sizes by buying group rather than by company, since one large company can contain several independent buying groups. The framework's own history: SiriusDecisions (acquired by Forrester in 2019) launched the original Demand Waterfall in 2006, re-architected it in 2012, then released the Demand Unit Waterfall in 2017, whose stage 1 ("Target Demand") defines the number of potential demand units believed to exist, and whose stage 2 ("Active Demand") narrows that to units currently in market - an intent-derived SOM. In Forrester's Demand Unit Waterfall framework, target demand equals SAM, because SAM already reflects the market's potential demand units (flag: the named demand-unit report stays paywalled beyond its public abstract; this SAM-equivalence detail comes from a separate Forrester publication, also unverifiable beyond its own abstract).
   • Named worked examples of the same discipline at smaller scale: Christoph Janz (Point Nine) frames it as ARPA times customer count against a fixed revenue target, segmented into five customer-size bands ("mice" through "whales") - the same bottom-up TAM equation, inverted to ask what customer mix reaches a goal. Tomasz Tunguz (Theory Ventures) published a fully worked "share of customer spend" TAM build: summed on-chain revenue across a named set of companies, applied an assumed share going to software, and derived a single dollar figure with every assumption stated - rare as a fully shown worked example rather than a headline number.

2. Gather data per the source-tier ranking above. Record publisher, date, and scope for every figure at collection time - retrofitting citations never happens.

3. Calculate TAM with the rung-2 formulas. In B2B, build from a count of companies matching the ICP criteria - firmographic and technographic filters - times per-segment ACV. In B2C, from a population count times purchase rate times average annual spend.

4. Narrow to SAM. Apply the filters the go-to-market actually enforces: geography, product capability, channel access, pricing tier. Use the same filter set the ICP and territory design use - a SAM filtered one way for planning and another way for messaging produces two irreconcilable numbers. State each filter's percentage and the reasoning behind it.

5. Cap SOM with capacity, not aspiration. Compute both, then ship the lower.

   • Capture-rate SOM: published conventions differ by stage - roughly 2-5% of SAM for a new entrant over 3-5 years, 5-15% for early-growth B2B. Treat these as sanity bands, not targets.
   • Capacity ceiling: SOM ceiling = reps × deals closed per rep per year × closed-won ACV, using closed-won because a pipeline showing an $80K average deal often closes nearer $40K.

   SOM is a constraint on what the team can physically win, never a goal. Add the growth-expectation check: if ~20% of SOM is already captured, 100% year-over-year growth on top is arithmetically unrealistic.

6. Triangulate and sanity-check. Compare bottom-up against top-down: ~30% convergence is good, past ~50% reopen the market definition. Then check:

   • Does the implied customer count at target SOM match the independently estimated addressable count?
   • Does implied revenue per customer fall inside the segment ACV range?
   • Does the target share sit below the established leaders' share?
   • Does a public comparable's actual revenue corroborate that a market this size exists?

   Worked math for all of this: worked-example-bottom-up.md.

7. Hand off downstream.

   • SOM feeds the top-down target that mbfinotti/sales-skills@sales-quota-setting reconciles against capacity.
   • The SAM filters feed territory carving; the TAM-to-headcount conversion lives in mbfinotti/sales-skills@sales-org-structure - hand it the numbers, do not redo its math.
   • TAM shape (long tail of small accounts vs few large logos) is a motion-selection axis for mbfinotti/sales-skills@sales-motion.

8. Set the refresh cadence, one clock per layer.

   • TAM and SAM: rebuild once or twice a year. The underlying market moves slowly; more frequent rebuilds just re-measure noise.
   • SOM: rebuild quarterly, because it tracks headcount, close rates, and realized ACV, which all move faster. Split it by the business's actual seasonality inside the year, never four flat quarters.
   • Off-cycle SOM rebuild triggers: a territory redesign, or any ICP change.
   • Off-cycle TAM/SAM rebuild triggers: repricing, a new product line, or an integration/packaging change that unlocks a previously blocked segment.

   Name the owner of each clock: a sizing nobody rechecks stops being strategy and becomes decoration.

9. Assemble the output (shape below) and match the presentation to the audience:

   • Planning-facing: lead with segment-level SAM/SOM detail; downplay the headline TAM.
   • Investor-facing: lead with the bottom-up calculation; show the top-down as corroboration.

   Then run the Measurement check and iterate until it passes.

B2B vs B2C

The methodology transfers; the counting unit and what the exercise can produce do not.

Differs:

• The bottom-up unit. B2B counts discrete, identifiable companies matching ICP criteria; B2C applies a purchase rate and average spend to a population count, because individual consumers are not addressable as named targets the way accounts are.
• The byproduct. A B2B bottom-up build done at the standing-model rung yields a prioritized named-account list; B2C sizing, built on population math, structurally cannot produce one - its planning byproduct is a segment-and-rate table instead.
• Category-definition risk is more acute in B2B niches. A published category figure bundles enterprise contracts, free tools, and everything between; a niche B2B product shares almost nothing with that headline number even though it is nominally "in the category." B2C population figures narrow more faithfully, because demographic and geographic filters map directly onto the actual buyer.
• Top-down filters differ. B2B narrows by firmographics - company size, industry, geography; B2C narrows by demographics and behavior - age cohort, income, location, usage.
• The capacity ceiling assumes a rep-driven motion. reps × deals × closed-won ACV bounds SOM where salespeople close the revenue. For self-serve or e-commerce B2C with no rep in the loop, bound SOM with the org's own historical acquisition rates and the capture-rate band instead, and say which bound was used.

Sizing output shape

MARKET DEFINITION: product line · problem and buyer · counting unit · geography · horizon · scoping decisions
TAM: bottom-up figure and formula inputs · top-down figure and source · divergence %
SAM: each filter with its % and reasoning · SAM as % of TAM · same-filter-set confirmation vs ICP/territory
SOM: capture-rate figure and band used · capacity ceiling and its inputs · which bound shipped and why
SANITY CHECKS: implied customer count · implied revenue per customer · share vs incumbents · public comparable
ASSUMPTIONS REGISTER: every load-bearing assumption, numbered, with confidence (high/medium/low) and what would change it
REFRESH: cadence per layer · off-cycle triggers armed · named owner · next rebuild dates
HANDOFFS: SOM → quota target · SAM filters → territory · TAM shape → motion

Failure modes

• TAM misuse - presenting the theoretical ceiling as an achievable target; plans built on it inherit the inflation. Fix: only SOM ships into planning math.
• SAM inflation - drawing the serviceable market around customers the team wishes it could serve; the resulting quotas can't be hit and the miss gets blamed on sales execution instead of the sizing. Fix: filters must match what product, pricing, and channel reach today.
• SOM as a goal instead of a constraint - committing to a number the team cannot physically produce, then spending a cycle explaining the shortfall. Fix: step 5's dual computation, lower bound wins.
• The "1% of a huge market" fallacy - a share assumption bolted onto a borrowed category figure collapses the moment someone asks "why 1% and not 0.1%?". Bottom-up is structurally immune: it has no free market-share parameter to inflate. A well-reasoned smaller number beats a huge one that can't be defended.
• Pipeline-stage ACV in the capacity formula - overstates the ceiling by up to half. Closed-won only.
• Conflating TAM with the ICP pool - sizing pipeline against the (far smaller) ICP labeled as "TAM", or forecasting against the full TAM most of which will never buy. The ICP is typically a 5-15% slice of TAM; keep the layers named.
• One blended TAM across product lines - hides which line carries the opportunity; every downstream allocation inherits the blur.
• Stale or cherry-picked data - flag anything older than ~2 years; when sources disagree, surface the disagreement instead of quietly picking the favorable one.
• Static sizing - numbers nobody revisits quietly drift out of sync with quotas and territories; the refresh clocks in step 8 exist to prevent exactly this.

Measurement

The sizing is not done until all of these pass; iterate until 100%:

• Every material figure carries a citation - publisher, date, scope - and sourced data points are visibly distinguished from modeled inferences.
• TAM shows both a bottom-up and a top-down figure with the divergence stated; divergence past ~50% reopened the market definition rather than shipping anyway.
• SOM shows both the capture-rate figure and the capacity ceiling, names which bound shipped, and uses closed-won ACV.
• The SAM filter set is confirmed identical to the one the ICP and territory design use.
• The assumptions register numbers every load-bearing assumption with a confidence label.
• Each layer has a refresh date and a named owner, and the off-cycle triggers are written down.

Outcome KPIs to track between refreshes:

• Actual bookings vs shipped SOM, and win rate inside the sized segments vs outside them - the sized market should visibly outperform.
• Drift at each refresh: how far the rebuilt SOM moved from the shipped one; large silent drift means the off-cycle triggers are not firing.
• For a standing model: share of pipeline sourced from the named-account list - the test of whether the sizing became an operating asset or stayed a slide.

References

• mbfinotti/sales-skills@sales-icp-definition: the ICP whose criteria filter SAM (this skill counts, that one filters)
• mbfinotti/sales-skills@sales-account-segmentation: score and group the accounts the sized market contains
• mbfinotti/sales-skills@sales-account-tiering: convert that grouped universe into tiers with a capacity cap and coverage level each
• mbfinotti/sales-skills@sales-quota-setting: reconcile SOM against ramp-adjusted capacity and derive quotas from it
• mbfinotti/sales-skills@sales-org-structure: the TAM-to-headcount math this skill deliberately does not duplicate
• mbfinotti/sales-skills@sales-motion: how TAM shape (long tail vs few large logos) selects the sales motion