Most GTM ideas in service companies fail quietly: the campaign runs for six weeks, gets a few polite replies, and nobody can say whether the angle, the list or the follow-through was the problem. This skill forces the idea into a testable bet before launch, finds the weak points, and leaves the team with a test that can fail on a date everyone agreed to.
Answer in the user's language. Be direct. Agreeing with the user is allowed only when the evidence supports it.
Inputs
Take the idea in whatever shape it arrives. Ask one question, and only if the answer changes everything:
- the target buyer is unknown, or
- the goal is unknown (pipeline, a first case in a new vertical, retention, awareness).
For everything else, state your assumption in one line and proceed. Useful context if the user has it: stage of the company, current pipeline numbers, who would execute and how many hours a week they have.
Modes
| Called from | Mode | Output |
|---|---|---|
| User directly | Full | All sections below |
| gtm-positioning | Angle test | Steps 1 to 4 plus verdict, applied to the positioning angle |
| gtm-action-plan | Plan test | Steps 2, 3, 4 and 7 run across the plan's top 3 to 5 initiatives, plus verdict per initiative |
Step 1. Rewrite the idea as a bet
One sentence, filled in completely:
We believe [segment] will [observable behaviour] when we [action], and we will see it in [metric] within [window].
If any bracket cannot be filled, that gap is the first finding. An idea with no observable behaviour ("build awareness among CTOs") cannot fail, so it cannot be tested yet.
Step 2. What must be true
List the conditions the bet depends on. Check at least these five areas and rate each condition evidence (data or a closed deal shows it), belief (plausible, not shown) or unknown:
- Buyer: the segment has this pain now, and a reachable person owns it
- Reach: you can find and contact enough accounts. A hypothesis tested on fewer than 300 contacts is not tested; small niches (a few thousand firms worldwide) run out fast
- Proof: you hold a case, reference or sample that makes the claim believable to this segment
- Capacity: a named person runs it with real weekly hours, and delivery can absorb the work if it converts
- Economics: deal size and margin justify the cost of acquisition and the sales cycle length you see in your own CRM
The two conditions rated belief or unknown that would hurt most if false become the focus of the test in Step 5.
Step 3. Blind spots
Pick the three that apply most to this idea and explain each in two or three sentences tied to the specifics. Common ones in service companies:
- Founder bottleneck: the plan works only while the founder personally sells or writes every message
- Bench pressure: the idea is driven by idle staff, so the segment is chosen for capacity and ignores demand
- Referral comfort: past growth came from referrals, and the team has no habit of measuring cold channels
- Sameness: the message matches what every other outsourcing firm sends ("dedicated team, flexible, cost-effective")
- Adoption gap: the consultant or founder builds the system and the client-side team never runs it weekly
- Proof mismatch: cases come from one vertical, the campaign targets another
- Timing: the trigger used is stale, or the buying cycle (budget season, project award dates) is ignored
Then write the strongest case against the idea in one paragraph, the version a sceptical CEO would make, and one uncomfortable question the user would rather not answer.
Step 4. Kill criteria
Agree before launch, not after. For each criterion: metric, threshold, date, and the decision taken if it is missed.
If [metric] < [threshold] by [date], we [stop / change the list / change the angle / change the channel].
Use leading metrics that show up in two weeks (positive reply rate, accepted connection rate, booked calls per 100 contacts) and one lagging metric checked later (qualified opportunities, proposal sent). Benchmarks come from the user's own past campaigns first. If you quote an external benchmark, name the source.
Step 5. Cheapest test, two weeks or less
Design the smallest experiment that can prove the riskiest condition from Step 2 false:
- What ships: the exact asset (list of N accounts, 3-touch sequence, 10 manual calls, one landing page, five discovery interviews)
- Sample: enough to read a signal; say why this number is enough
- Owner and hours: a named role and the time it takes
- Cost: tools, data credits, paid time
- Pass / fail line: taken from Step 4
If the cheapest honest test takes longer than two weeks, say so and split it into a first slice that fits.
Step 6. The bigger version
If the test passes, what is worth building? Describe it in concrete terms: second segment or geo, a productised offer with a fixed scope and price, a channel combination, automation of the manual steps, a hire. Add one contrarian variant (the opposite move a competitor would not make) only if it is genuinely viable.
Step 7. Execution plan
| # | Action | Owner (role) | Due date | Metric | Done when |
|---|
Dependencies go in their own rows (data, copy, tool access, approvals) with status: ready, easy, blocker. No line without an owner. "Team" is not an owner.
Verdict
Close with one of four labels and three sentences of reasoning:
- Go: the evidence is there, launch the plan
- Go with changes: launch after the listed fixes
- Test first: run Step 5 before committing anything more
- Park: the bet is weak now; say what would have to change to reopen it
Then: the single thing to do tomorrow morning.
Output template
# Stress test: [idea in five words] · [date] · mode: [full / angle / plan]
Bet: We believe ... (Step 1)
Assumptions I made: [list]
## What must be true
| Condition | Area | Rating | Why |
## Blind spots
[three, specific]
Strongest case against: [paragraph]
Uncomfortable question: [one]
## Kill criteria
[lines in the Step 4 format]
## Two-week test
What ships, sample, owner and hours, cost, pass/fail line
## If it works
Bigger version, plus a contrarian variant when viable
## Plan
[Step 7 table]
## Verdict: [label]
[three sentences] · Tomorrow: [one action]
Quality gate
- The bet sentence has every bracket filled, or the gap is named as finding #1
- At least one condition is rated
belieforunknown(if all areevidence, check again) - Blind spots refer to this idea, not generic advice
- Kill criteria have numbers and dates
- The test fits in two weeks and can actually fail
- Every plan line has an owner role and a date
- Any benchmark quoted has a source
Hand-offs
- Turning a passed test into a full hypothesis card: if installed,
hypothesis-builder(packoutbound-engine-skills) orhypothesis-scoring(packgtm-skills) - Back to the calling step:
gtm-positioningorgtm-action-plan(this pack) - Risk register for a larger plan:
risk-assessment(this pack)
Credits
Idea adapted from a public outbound-skills collection; rewritten for B2B service companies.