gtm-teardown

GTM teardown: understand a company before GTM work

This skill turns a company name or URL into a complete picture of the business through a fixed chain of tables. Every stage outputs a table before the next stage starts. The tables are the thinking: one answer per cell, one source per answer, one rank per channel. Gaps show up as MISSING instead of hiding inside confident paragraphs.

The chain answers four questions. Rows: the sections. Columns: the question each answers, and the stages that answer it.

Section	Question	Stages
A. The business	What does it do, for whom, why is it hired, where does it fit?	2, 3, 4, 5
C. Competitors and comparables	Who else does the job, and who sells a similar way to the same buyer?	6
D. Competitor GTM	Where do they actually invest, and what is working for them?	7
B. Channels	Which channels come first, ranked by impact and likelihood of success?	8
Summary	The whole business in 21 fields, comparable across clients	9

B runs after C and D on purpose. A channel ranking built only on fit is an opinion; competitor evidence turns it into a bet.

Reference files. Read each one at the stage that names it, not before.

File	What it holds	Read at
references/channel-library.md	20 channels with works-when, fails-when, evidence and leading metric, plus the one eligibility rule	Stage 8
references/gtm-proxies.md	How to measure sales headcount, ads, SEO, partners and the rest, with search strings, tool routes and reading rules	Stage 7
references/worked-example.md	A hypothetical company with every table filled	When unsure of depth or format
assets/client-framework-page.html	Fillable one-page version of the framework	Only if the user asks for a fillable page or template

---

Table rules (apply to every table)

1. State rows and columns. One line above each table: Rows: ... Columns: .... Nobody should have to guess what a row is.
2. Fill every cell. Unknown is MISSING. Not applicable is n/a plus the reason. Never guess a customer name, a number, a headcount or a price. Every MISSING becomes a line in the Stage 9 open questions table.
3. Tag every fact with its source (tags below). A fact without a tag counts as an assumption.
4. Words, not symbols, in check columns: PASS / FAIL, High / Med / Low / None, Now / Build / Support / Later. They survive every renderer and paste cleanly into sheets.
5. Tables first, prose second. At most two connecting lines between tables. Never a paragraph that restates a table.
6. Customer words and numbers beat adjectives. Quote real phrases from reviews or calls in quotation marks. "Saves finance 6 hours a month" beats "streamlines reporting".
7. Don't skip stages the route table requires. Stage 8 without Stage 7 is guesswork.
8. No em dashes anywhere. Use commas, periods or colons.

Source tags. Rows: the tags. Columns: the tag and what it means.

Tag	Meaning
[Site]	The company's own website, docs, pricing page
[Reviews]	G2, Capterra, Trustpilot, app stores, Amazon reviews
[LinkedIn]	LinkedIn company page, people search, Sales Navigator
[Apollo] [ZoomInfo]	Contact and company databases (mostly LinkedIn-derived)
[Ahrefs] [Similarweb]	Traffic and keyword estimates
[AdLib]	Meta, LinkedIn, Google or TikTok ad libraries
[Stack]	BuiltWith or Wappalyzer
[News]	Press, Crunchbase, funding announcements
[Client]	Provided by the client or the user
[Judgment]	Reasoned assessment. Allowed only for scores and verdicts, never for facts
[Illustrative]	Teaching data. Never allowed in real work

---

Stage 0: Route the request

Rows: request types. Columns: stages to run, and the final output.

Request	Stages to run	Final output
Full client understanding (new client, kickoff)	1 to 10	All tables and the master summary
Quick snapshot (call tomorrow)	1, 2, 3, 4 (primary persona only), 9, 10	Master summary with MISSING flagged
Interview or pitch prep for a prospective employer or client	1 to 6, 7 (D1 and D3 only), 8, 9, 10	Summary plus the Now channels with evidence
ICP and personas only	1, 2, 3, 4	Persona matrix and persona test
Market or geography entry check	1, 2, 4, 5, 6 (filtered by country)	Geography scorecard with tiers
Competitor landscape	1, 2, 6	Register and coverage check
Competitor GTM teardown	1, 6, 7	D1 to D4
Which channels should we use, or in what order	1, 2, 3, 4, 7, 8	Ranked channel list, priority matrix, test plan
Update an existing summary with new information	Only the stages the new information touches, then 9	Changed rows marked "Updated"

If the request is ambiguous, run the quick snapshot and offer the full version in one line at the end.

---

Stage 1: Intake

Rows: the inputs every analysis needs. Columns: the input, the value, status, and why it matters.

Input	Value	Status	Why it matters
Company name and URL		Have / MISSING	Everything starts here
Why we are doing this (kickoff, pitch, interview, market entry)		Have / MISSING	Sets depth and which stages carry weight
Client materials (deck, CRM export, call notes, pricing, metrics)		Have / MISSING	First-party truth beats public estimates
Markets in scope, or "find them"		Have / MISSING	Scopes Stage 5 and every count
Competitors the client names		Have / MISSING	Seeds Stage 6
Named customers and real numbers		Have / MISSING	Proof for value and personas
12-month goal (new revenue, pipeline or customers)		Have / MISSING	The yardstick for channel Impact in Stage 8
Tools available in this session		Have / MISSING	Decides which proxies are measurable and which stay MISSING
Depth and deadline		Have / MISSING	Snapshot or full

Research first, ask once. Run the public research for Stages 2 to 7 before asking anything. Then ask for every remaining MISSING input in a single message. Never ask the user for something the website already says.

Tools. Use whatever is connected: web search and fetch for the site, reviews and ad libraries; Apollo, ZoomInfo or Clay for company and people counts; Ahrefs or Similarweb for traffic; BuiltWith or Wappalyzer for stacks. Prefer search endpoints over paid enrichment. If a proxy can't be measured with the tools present, write MISSING (no tool) rather than estimating.

---

Stage 2: A1, what it does and why it is hired

Standards. Rows: the eight questions. Columns: the question and what a good answer looks like.

#	Question	A good answer
A1.1	What does it do?	One sentence a non-expert gets in 10 seconds: "[Company] helps [who] do [what] without [pain]." No category jargon
A1.2	What problems does it solve?	Top 3, ranked by how often customers mention them, each tied to the person who feels it
A1.3	What do customers do today without it?	The status quo named precisely: spreadsheet, agency, in-house hire, a named competitor, or nothing. Usually the real competitor
A1.4	What value does it create?	A number with a unit of time: revenue gained, cost removed, hours saved, risk reduced, per customer
A1.5	What job is it hired for?	"When [situation], I want to [motivation], so I can [outcome]", plus the emotional job and the social job
A1.6	What triggers the purchase?	An observable event: new hire, funding, regulation, expansion, season, life event. Triggers become outbound signals and ad timing
A1.7	How does it make money?	Pricing model, entry price, ACV or AOV, payback, who signs the budget
A1.8	What proof exists?	Named logos, case studies with numbers, review count and rating, retention

Output. Rows: A1.1 to A1.8. Columns: code, answer, source, status (Verified / Assumption / MISSING).

#	Answer	Source	Status

Then the A1 check. Rows: checks. Columns: check, result. Fix any FAIL before Stage 3.

Check	Result
A1.1 has no jargon and names who, what and the pain avoided
A1.3 names a concrete alternative, not "manual processes"
A1.4 contains a number and a time unit
A1.5 follows When / I want to / So I can, with emotional and social jobs
A1.6 is an event you could find in news, job posts or data
At least one real customer phrase is quoted somewhere in A1

---

Stage 3: A2, business model

Rows: the classification tests. Columns: the test, the answer, the source, and what it implies for GTM.

Test	Answer	Source	GTM implication
Who pays? (business, consumer, both, two sides, government)
Who uses it, and is it the same person who pays?
Who else must approve? (nobody, manager, committee)
Size of one purchase (ACV, or AOV and repeat rate for B2C)
Purchase frequency (once, recurring, usage-based)
Does a consumer product sell through businesses?
Model label (B2B, B2C, B2B2C, marketplace, B2G)

Reading rules of thumb, to confirm with client data. Rows: signals. Columns: signal, usual implication.

Signal	Usual implication
ACV under about $1k a year	Self-serve or PLG. Sales touch rarely pays back
ACV $1k to $25k	Inside sales, outbound viable, demo-led
ACV above $25k	Field or enterprise sales, ABM, events
Buyer and user differ	Two messages; PLG only if the user can adopt alone
Three or more approvers	Long cycle, multi-threading, business case needed
Recurring or usage revenue	Can fund paid acquisition, affiliates, referral rewards
Consumer product sold through businesses	Run Stage 4 twice: end consumer and the business channel

---

Stage 4: A3, personas

One persona is one industry plus one title (B2B), or one audience type plus one life situation (B2C). Two titles never share a row, because they don't share a pain. This matches the ICP rule in outbound-framework, so B2B rows carry straight into its ICP matrix. Aim for 3 to 6 personas plus one anti-persona (looks like a fit, never buys).

B2B persona matrix. Rows: one persona each, anti-persona last. Columns: the fields that make a persona findable and reachable.

#	Industry	Company size	Title	Role in deal (buyer, champion, user)	Profile keywords (tool filters)	Pain in their words	Trigger	Disqualifier	Addressable count and market	Where reachable

B2C persona matrix. Rows: one persona each, anti-persona last. Columns: the fields that make a consumer audience targetable.

#	Audience type	Age band	Gender split	Income band	Life stage	Interest and search keywords	Pain in their words	Trigger	Audience size and market	Where reachable

Column rules:

• Profile keywords are words that appear in LinkedIn titles or headlines and work as Apollo or Sales Navigator filters. For B2C, interests, hashtags and search terms that work as Meta, TikTok or Google targeting.
• Pain in their words is a quote from reviews, calls or forums, not a paraphrase.
• Addressable count comes from a tool count with a source tag, or MISSING (no tool).
• Hybrid models fill both matrices.

Persona test. Rows: one persona each. Columns: the five tests and the verdict. Any FAIL: fix or drop the persona before Stage 5.

#	Filterable in a tool	Owns the pain (measured on it)	Trigger is observable	A reachable channel exists	Evidence this persona buys (customer or competitor's customer)	Verdict

---

Stage 5: A4, geography fit

Always check where the business fits before judging channels. A channel that works in one market can be empty, illegal or too expensive in the next.

Scorecard. Rows: candidate countries, starting with where revenue comes from today. Columns: seven checks scored 1 (blocker) to 5 (strong), total out of 35, tier, and the one-line reason.

Country	Volume	Language	Regulation	Payments and price	Competition (5 = open)	Local proof	Sales coverage	Total	Tier	Reason

What each check means:

• Volume: B2B, persona contacts and target companies per country from Stage 4. B2C, search volume and reachable audience.
• Language: product, support and sales material in the local language, or a market that buys in English.
• Regulation: licences, data residency, GDPR, sector rules, outreach rules (B2B cold email is heavily restricted in Germany, for example).
• Payments and price: local payment methods, currency, price against local purchasing power and local competitor prices.
• Competition: number and strength of local players, from Stage 6.
• Local proof: local logos, reviews, case studies, press.
• Sales coverage: time zone overlap, and whether deals need people on the ground.

Tiers: 28 or more is Tier 1 (enter now), 21 to 27 is Tier 2 (next 12 months), below 21 is Tier 3 (no). A Regulation score of 1 makes the country Tier 3 whatever the total.

---

Stage 6: C, competitors and comparables

Rows: the three types. Columns: definition, why it matters, and the minimum to collect.

Type	Definition	Why it matters	Minimum
Direct	Same job, same persona, same market	Price anchor, and the list buyers compare against	3
Indirect or substitute	Same job, different solution (agency, spreadsheet, in-house, doing nothing)	Usually the status quo from A1.3; the message must beat it	2, including the A1.3 status quo
Comparable	Same buyer, same price band or same sales motion, different product or market	Playbooks already proven with this buyer or model	3, at least one from a more mature market

Discovery log. Rows: discovery methods. Columns: whether it was run, and what it found. Run at least five.

Method	Run	Companies found
Client's own list: who they lose to, who customers mention	PASS / FAIL
Review site alternatives and comparison pages (G2, Capterra, Trustpilot)
Searches: "[category] software", "[competitor] alternatives", "[competitor] vs"
Ahrefs organic competitors or Similarweb similar sites
Crunchbase, Tracxn or Dealroom similar companies
App stores, Amazon categories, Shopify or HubSpot marketplaces
AI assistant prompts: "best tool for [job to be done]" (also the GEO baseline)
Other tools sold to the same title (comparables, same buyer)
The same model in a more mature market (comparables, same model)

Register. Rows: one company each. Columns: identity, scale, offer, positioning, and what their customers say.

Company	Type	Why included	HQ and active markets	Stage and funding	Headcount	Pricing, ACV or AOV	Positioning line (their words)	Target persona	Praised for	Complained about

Coverage check. Rows: the minimums above. Columns: required, found, result. Any FAIL: go back to the discovery log.

Minimum	Required	Found	Result

---

Stage 7: D, competitor and comparable GTM

Read references/gtm-proxies.md before starting. Public proxies show where each company spends. Where they keep investing is what works; where they cut is what failed.

D1, sales headcount. The outbound proxy. Rows: companies from the register. Columns: team size, sales mix, trend, and the read.

Company	Total headcount	Sales roles	SDR / BDR	AE	Partner or channel	Sales share	12-month sales trend	Open sales roles	Where sales staff sit	Source	Read

The Read column uses the interpretation rules in the reference file (for example, 3 or more SDRs means outbound is a core channel). LinkedIn counts are ranges, not exact figures; say so when the gap matters.

D2, evidence log. Rows: one observation each. Columns: company, channel, the proxy measured, what was observed, source, and the level it supports.

Company	Channel	Proxy	Observation	Source	Level

D3, investment map. Rows: companies. Columns: the 12 channels. Cells: High, Med, Low, None or MISSING. Every High or Med must trace to at least one D2 row.

Company	Outbound	Content and founder social	Paid social	Paid search	SEO	AI search	Creators	Affiliate and referral	Partners	Marketplaces	Events	PLG

D4, lessons. Rows: the 12 channels. Columns: how many companies invest heavily, what the winners do, what was tried and dropped, and the evidence score carried into Stage 8.

Channel	High count (of assessed)	What the winners do	Tried and dropped	Evidence score

Evidence score rules. Rows: scores. Columns: score, condition.

Score	Condition
5	Three or more companies at High, and still growing it (trend or hiring)
4	Two companies at High
3	One company at High, or three or more at Med
2	Only Low or a single Med
1	None everywhere, or tried and dropped

---

Stage 8: B, channels ranked by impact and likelihood of success

Read references/channel-library.md before starting. It holds the works-when and fails-when conditions for all 20 channels.

Two principles run this stage:

1. Every channel is on the table for every business, with one exception: no cold email when the recipient is a consumer. A B2C business never runs cold email outbound. When a B2C business also sells through businesses (B2B2C, or the supply side of a marketplace), cold email is allowed toward that business side only, never toward consumers.
2. Everything else is prioritization. A channel that suits the business badly is not removed; it ranks low. Impact and likelihood of success decide the order.

B0, eligibility. Rows: the one exclusion rule. Columns: the rule, whether it applies, the fact that decides it, and the effect.

Rule	Applies	Deciding fact (with stage)	Effect
Recipient is a consumer (B2C)	Yes / No	Model label from A2	Yes: cold email excluded. B2B2C or marketplace: cold email allowed toward the business side only, stated with that side's persona

No other channel is ever excluded. Local legal limits (such as cold email restrictions in Germany) lower Likelihood for that market; they don't remove the channel.

B1, reasoning. Rows: every eligible channel (20, or 19 for B2C). Columns: why it could work for this business, why it may not, each citing the stage that proves it.

Channel	Why it could work (cite stage)	Why it may not (cite stage)

B2, scoring. Rows: every eligible channel. Columns: Impact and its basis, the three inputs to Likelihood, Likelihood, Priority, and two sequencing scores.

Channel	Impact	Impact basis	Fit	Evidence	Readiness	Likelihood	Priority	Speed to signal	Cost to test

Scoring rubric. Rows: the scores. Columns: what 1 means, what 5 means. Scores 2 to 4 sit in between.

Score	1 means	5 means
Impact	Could add under 5% of the 12-month goal	Could add 50% or more, or become the main growth engine
Fit	Persona, price or geography contradict the channel (Stages 3 to 5)	Matches how this persona buys, at this price, in the Tier 1 markets
Evidence	Copied from D4, using the channel-to-column mapping in references/channel-library.md. Channels with no D3 column are scored from D2 with the same rules	Copied from D4
Readiness	The client lacks what the channel needs: proof, asset, integration, content capacity, team or budget	Everything needed to launch is in place next week
Speed to signal	Six months or more before you can judge it	Readable in 2 to 4 weeks
Cost to test	Needs a large budget, new hires or a build first	Small budget, existing team

How the numbers combine:

• Impact bands: 1 is under 5% of the goal, 2 is 5 to 10%, 3 is 10 to 25%, 4 is 25 to 50%, 5 is 50% or more. If the 12-month goal is MISSING, score Impact relative to the other channels: the largest realistic contributor anchors 5.
• Impact basis is one line of sizing: reachable volume in Tier 1 markets (Stages 4 and 5), times value per customer (A1.7), times how far the channel scales with effort or spend. Any conversion rate used is an assumption, tagged [Judgment].
• Likelihood is the average of Fit, Evidence and Readiness, rounded to the nearest whole number. If Fit is 1 or 2, Likelihood can't exceed Fit + 1: competitor evidence doesn't rescue a channel the persona doesn't use.
• Priority is Impact times Likelihood, from 1 to 25.
• Speed and Cost don't enter Priority. They order channels with equal Priority and decide which Support channels are worth running.

B3, prioritization. Ranked list. Rows: every eligible channel, in rank order. Columns: rank, channel, priority, impact, likelihood, quadrant, and the next step.

Rank	Channel	Priority	Impact	Likelihood	Quadrant	Next step

Sort by Priority, then Likelihood (surer bets first), then Speed to signal, then Cost to test.

Quadrants. Rows: the four quadrants. Columns: impact band, likelihood band, what it means, what to do.

Quadrant	Impact	Likelihood	Means	What to do
Now	4 or 5	4 or 5	Big and likely	Test now. Run at most four at once, in rank order; the rest queue behind them
Build	4 or 5	1 to 3	Big, not yet likely	Name the lowest Likelihood input and what would raise it (proof, asset, integration, hire), with a revisit date
Support	1 to 3	4 or 5	Likely, small	Run alongside Now channels when Cost to test is 4 or 5. They compound, they rarely lead
Later	1 to 3	1 to 3	Small and unlikely	Parked, not excluded. Revisit at the next quarterly review

Priority matrix. Rows: impact band. Columns: likelihood band. Cells: the channels in that quadrant, in rank order.

Impact	Likelihood 4 to 5	Likelihood 1 to 3
4 to 5	Now:	Build:
1 to 3	Support:	Later:

Build list. Rows: each Build channel. Columns: the input holding Likelihood down, what would raise it, who owns that, when to revisit.

Channel	Lowest input (Fit, Evidence or Readiness)	What would raise it	Owner	Revisit

B4, test plan. Rows: each Now channel being run (four at most). Columns: the first test, its size, the metric that reads it, and when to stop.

Channel	Test design (persona, market, offer)	Budget and duration	Leading metric	Target	Kill criteria

---

Stage 9: Master summary

The one-page view of the business. The same 21 fields for every client, so clients compare side by side. For several clients, add one answer column per client.

Rows: the 21 fields. Columns: field number, field, answer, source stage, confidence (High / Med / Low).

#	Field	Answer	Source stage	Confidence
1	Client and URL		1
2	What it does, in one sentence		A1.1
3	Top problem solved		A1.2
4	Status quo it replaces		A1.3
5	Value created, as a number		A1.4
6	Job to be done		A1.5
7	Purchase trigger		A1.6
8	Model		A2
9	Buyer, user and approvers		A2
10	ACV or AOV, and sales cycle		A1.7, A2
11	Primary persona		A3
12	Anti-persona		A3
13	Tier 1 markets		A4
14	Now channels, in rank order		B3
15	Excluded channels (B2C: cold email) and lowest-ranked channels, and why		B0, B3
16	Top 3 direct competitors		C
17	Top 3 comparables		C
18	Competitor sales intensity (median sales roles and SDRs)		D1
19	Dominant competitor GTM motion		D3
20	Biggest GTM lesson from competitors		D4
21	Biggest unknown		All

Open questions. Rows: every MISSING and every Assumption that changes a verdict. Columns: the question, why it matters, who can answer.

#	Question	Why it matters (which verdict it could change)	Who can answer (client, customers, data)

Handoffs, in one line at the end: B2B persona rows feed the ICP matrix in outbound-framework; the summary, ranked list and test plan feed the GTM plan.

---

Stage 10: QA before delivering

Rows: the checks. Columns: result, and the fix if it failed. Deliver only when every row is PASS, or the FAIL is listed in open questions.

Check	Result	Fix if FAIL
Every table has a Rows / Columns line
Every fact carries a source tag; no [Illustrative] in real work
No invented names, numbers, headcounts or prices
A1 check all PASS
Every persona passed the persona test
Geography scored, Tier 1 named, Regulation blockers applied
Competitor coverage minimums met
Every High or Med in D3 traces to a D2 observation
B2C: cold email excluded (B2B2C and marketplaces: business side only); no other channel excluded
Every eligible channel has Impact with a basis, Fit, Evidence and Readiness
Evidence scores in B2 match D4
Likelihood and Priority follow the formulas, including the Fit cap
Every eligible channel appears in the ranked list and the matrix
At most four Now channels running, each with a B4 test and kill criteria
Every Build channel names what would raise its Likelihood
Summary fields all filled or MISSING, open questions listed
No em dashes, no prose paragraphs restating tables

---

Output order for a full analysis

1. Stage 1 intake (questions only after the research, in one message)
2. A1 table and A1 check
3. A2 table
4. Persona matrix and persona test
5. Geography scorecard
6. Discovery log, register, coverage check
7. D1, D2, D3, D4
8. B0 eligibility, B1 reasoning, B2 scoring, B3 ranked list, matrix and Build list, B4 test plan
9. Master summary and open questions
10. QA table

Default format is markdown tables in the reply. If the user asks for a document, page or file, render the same tables in that format without changing their structure. Very long analyses can be delivered in two turns: A and C first, then D, B and the summary.