Run in order. Prompts 1-4 design the replication test before you spend a dollar; prompts 5-7 calculate what the test actually costs and whether the cash buffer from M9 can absorb it; prompts 8-10 write the kill, scale and hold rules before launch so the decision at week 8 is made by criteria, not by the founder’s emotional investment in the market they just spent eight weeks selling into. Don’t move to prompt 5 until prompts 1-4 have produced a shortlist of three candidate second markets with scores, one chosen market with a written justification, and a replication test plan with binary success criteria.

1. The second-market shortlist — three candidates scored, not “wherever I have a friend”. “My segment from M2 is [paste]. My offer from M5/M6/M7 is [paste]. The channel that produced the highest-quality buyer in M8 was [paste]. The system from M9 is currently running in [name the actual market 1 — geography + channel combined, e.g. ‘Sydney Northern Beaches via LinkedIn warm outreach’ or ‘Australian e-commerce via Meta paid + my own email list’]. From that base, build a shortlist of three candidate second markets to test. Each candidate must be either a new geography with the same channel OR the same geography with a new channel — never both at once. Score each candidate against five criteria, 1-5: (a) reachability — can I get in front of qualified buyers without a personal warm intro; (b) translatability — does my positioning hold without rewriting the core message; (c) cost to enter — what does the first 90 days realistically cost; (d) signal quality — when this test runs, will the result tell me something I don’t already know; (e) cash recovery time — if the test works, how long until the test pays itself back. Output as a 3-row, 5-column scoring table with totals. Highest total is the candidate. If two are within 2 points, take the one with the higher signal quality score — the point of M10 is information, not revenue.”
2. The market 1 baseline — the conversion rates the second market has to substantially match. “Pull from M8 / M9 the actual conversion rates that produced the first buyers in market 1. Stages to baseline at minimum: contact-to-reply, reply-to-call, call-to-close, customer-acquisition-cost (CAC), revenue-per-customer (AOV or LTV), and time-from-first-contact-to-close. For each: the rate I measured in market 1, the sample size it’s based on (be honest — 5 buyers is a directional baseline, not a statistical one), and the confidence band (‘this rate is reliable / this rate is suggestive / this rate is essentially anecdotal’). Output as a 6-row table — stage, market 1 rate, sample size, confidence. This is the spec sheet the second market has to substantially match for the replication test to count as a pass. ‘Substantially match’ for a directional sample = within 30% on each rate; for a statistical sample = within 15%. Define the threshold now, before the data lands, so you can’t move it later.”
3. The replication test plan — same funnel, new market, binary success criteria. “Using the chosen second market from prompt 1 and the market 1 baseline from prompt 2, write the replication test plan. Structure: the offer (unchanged from M5/M6/M7 — same price, same positioning, same promise; if you change the offer you are not running a replication test, you are running a new product test and you don’t yet know what failed), the channel (the new channel if channel expansion, the new geography’s adaptation of the same channel if geographic), the funnel mechanic (same as market 1, end-to-end — same outreach pattern, same call structure, same close), the test window (90 days, not 30 — anything shorter is a vibe check, not a test), the sample target (minimum 50 qualified contacts to produce a directional read on conversion; 200 for a confident read), and the binary success criteria (the test passes if conversion rates land within the threshold defined in prompt 2 on at least 4 of 6 stages AND CAC is at or below 1.5x market 1 CAC; the test fails otherwise). Output the test plan as a single one-page brief. The brief is signed and dated before launch — that’s the contract with future-self that prevents moving the goalposts when the data is mid-flight.”
4. The single-variable discipline — what you are explicitly NOT testing in market 2. “List every variable that is changing between market 1 and market 2 in this test, and every variable that is not. The point of the discipline: the more variables that change, the less the result tells you. If geography changes AND channel changes AND offer changes AND price changes, a failed test tells you nothing — you don’t know which variable killed it. Output two columns: ‘changed in market 2’ (target: 1, maximum 2) and ‘held constant from market 1’ (everything else — offer, price, positioning, funnel structure, founder time per contact, qualifying criteria, follow-up cadence). For every variable in the ‘changed’ column, write one sentence on why this variable is the one being tested. For every variable in the ‘constant’ column, write the discipline rule that holds it constant under pressure (e.g. ‘if I am tempted to lower the price in market 2 to get the first close, I do not — I close the test as a fail and learn the price ceiling instead of contaminating the data’). The discipline rules are the test’s integrity.”
5. The expansion cost stack — what 90 days of replication actually costs. “Build the full expansion cost stack for the 90-day replication test designed in prompts 3-4. Six cost categories at minimum: (a) acquisition cost — paid ads, sponsorships, list rentals, partnership fees; (b) market research and entry cost — local legal, tax registration if needed, translation if applicable, market intelligence subscriptions; (c) channel setup cost — new platform fees, new tooling, new integrations; (d) operational overhead — incremental hours per week from founder + team at fully-loaded rate, additional fulfilment cost per unit if applicable; (e) inventory or capacity cost — units pre-committed if physical product, capacity pre-blocked if service; (f) contingency — 25% of the above, not optional. Output as a 6-row table with weekly run-rate, total 90-day cost, and the assumption behind each line. Total at the bottom. The total is the number you compare against the cash buffer from M9. If the total exceeds 60% of available buffer, the test is too big — go back to prompt 3 and reduce sample target or shorten test window before launch. M10 doesn’t pass on a cash overrun.”
6. The cash buffer check — what the business can absorb without the test killing it. “Pull the current cash buffer from the financial state established in M9 (or, if M9 didn’t formalise it, build it now: cash on hand minus 90 days of committed operating spend minus 60 days of payables minus 30 days of customer refund risk = available buffer). Compare against the total expansion cost stack from prompt 5. Three scenarios output: (a) test passes at month 3 — when does the test cohort start paying back, what’s the cumulative cash position month-by-month for the next 12 months, when does the business break even on the expansion; (b) test fails at month 3 — the spend has happened, no return, what’s the cumulative cash position, does the business have enough buffer to absorb the loss without breaking another commitment (payroll, supplier payments, tax, debt); (c) test produces mixed signal at month 3 and the founder is tempted to extend by another 90 days — what does the extension cost, what’s the cash position at month 6 if it still fails. The scenario that matters most for the go/no-go is scenario (b). If scenario (b) breaks the business, the test is too big for the buffer regardless of how good the candidate market scored. Output a 12-month cash bridge per scenario.”
7. The opportunity cost — what doesn’t get built while the founder is testing market 2. “List the work in market 1 — the M9 system, the cadence, the named customers, the content engine, the acquisition channel — that will receive less of the founder’s time during the 90-day expansion test. For each: the realistic hours per week that will be diverted from this work to the test, the risk to market 1 from the founder being part-time on it for 90 days (low / medium / high — based on whether the M9 system is genuinely runnable by someone-not-the-founder yet, which is the M11 prerequisite), the mitigation (who covers what, what doesn’t get covered, what intentionally goes on hold). Output as a 5-row table — market 1 workstream, hours diverted per week, risk level, mitigation, what slips. The opportunity cost is the silent killer of expansion tests — the founder runs the test and wins it, then comes back to a market 1 operation that drifted for three months and now needs rescuing. If the M9 handoff readiness audit (M9 worksheet page 6) showed that more than 50% of market 1 tasks are still founder-only, the founder is not ready for M10 yet. Go back to M11 prerequisites before launching.”
8. The kill rule — the conditions that close the test as a fail before month 3. “Write the kill rule for the replication test before launch. Structure: the leading indicators that say ‘this test is going to fail, close it now and reallocate the remaining budget’ — not ‘the test has failed, post-mortem time’. Examples to draw from: contact-to-reply rate in market 2 less than 50% of market 1 rate after first 30 qualified contacts AND no diagnosable fix (wrong segment, wrong channel access, wrong message); CAC tracking at more than 3x market 1 CAC at month 1 with no improvement trajectory; offer rejection reasons in market 2 reveal a structural mismatch (the segment doesn’t actually exist in market 2 / the offer’s premise doesn’t hold here) rather than a tactical mismatch (the headline needs rewriting). Output: 3-5 kill conditions, each with the metric, the threshold, the time window, and the action (‘close test within 7 days, run post-mortem against prompts 9-10, return remaining budget to buffer’). Written before launch, signed before launch. The kill rule’s job is to make the call when the founder, eight weeks in and emotionally invested, will not.”
9. The scale rule — the conditions that turn a passed test into a deployment. “Write the scale rule before launch. The scale rule is what makes a passed replication test actionable rather than a moral victory. Structure: the conditions, all of which must be true, that justify deploying the model into markets three and beyond within 90 days of the test closing. Examples: conversion rates landed within 30% of market 1 baseline on at least 4 of 6 stages AND CAC at or below 1.5x market 1 CAC AND cash payback period at or below 12 months AND the M11 handoff work has progressed enough that adding market 3 doesn’t put market 1 or market 2 at risk. Output: 4-6 scale conditions with the metric and threshold per condition, plus the deployment plan that triggers if all conditions are met (the next 2 markets, in order, scored on the same shortlist criteria from prompt 1; the cash buffer rebuilt to pre-test level before market 3 launches; the M11 prerequisites confirmed met). A passed test without a scale rule means the founder will either over-extend (deploy to five markets at once because momentum) or under-extend (sit on the proof for six months while the timing window closes). Both are common. Both are avoidable.”
10. The hold rule and the post-test diagnosis — when the test produces mixed signal. “Write the hold rule (the conditions that say ‘extend by 30-60 days because the signal is genuinely incomplete, not because the founder isn’t ready to accept the result’) and the diagnosis framework for any test outcome that isn’t a clean pass or clean fail. Hold rule conditions: sample size landed below the directional threshold (less than 50 qualified contacts in 90 days, meaning the test didn’t get fair stress) AND no kill conditions tripped AND the cash buffer can absorb the extension without breaking scenario (b) from prompt 6. Diagnosis framework for mixed signal: which stages held, which stages collapsed; which collapses are likely tactical (fixable in market 2 with adjustment) versus structural (signalling the model doesn’t travel here); what the founder would have to believe to extend versus what they’d have to believe to close. Output: hold rule conditions as a 3-bullet list, plus a diagnosis template the founder fills in at month 3 of any test that wasn’t a clean pass or clean fail. The diagnosis template is the antidote to ‘just give it another month’ — the most expensive sentence in expansion.”