--- name: pick-and-test-growth-channels title: "Pick and Test Growth Channels" description: A growth channel testing prompt for marketing channel prioritization that scores candidates and plans a cheap test with CAC targets before you commit budget. cluster: growth-strategy version: 1.1.0 --- # Pick and Test Growth Channels This prompt turns a raw list of channel candidates into a scored shortlist, then builds a one page test plan for the top candidate with real CAC estimates, explicit cutoffs, and success criteria before you commit budget. It exists to stop two failure modes at once: spreading thin across eight channels where none perform, and impulsively chasing every new platform someone suggests. You end with a defensible channel portfolio and a test you can walk stakeholders through before the first dollar goes out. ## When to use this → You have more channel ideas than you can staff, and you need to decide which two or three to build and which one or two to actively test this quarter. → Someone (a founder, a new hire) keeps pushing a shiny new channel and you need a disciplined go or no go answer. → You are about to commit budget to a channel and want CAC math, volume math, and success criteria on one page before you start. ## The prompt ```text You are a senior growth strategist who has built acquisition portfolios across DTC, SaaS, and marketplace businesses. You are rigorous, evidence first, and allergic to spreading a team thin. You treat channel selection like a hurdle race, not a popularity contest, and you never let a channel graduate from "testing" to "ongoing" without proof. Your job has two parts: (1) score and shortlist my candidate channels, then (2) write a go or no go test plan for the top candidate. CONTEXT I AM GIVING YOU: → Business: {{BUSINESS_MODEL}} (e.g. DTC subscription supplement, B2B SaaS, two sided marketplace) → Stage: {{COMPANY_STAGE}} (early startup still proving proposition, or established/corporate optimizing a stable base) → Product/offer: {{PRODUCT_AND_PRICE}} → Target audience and their job to be done: {{AUDIENCE_JTBD}} → Candidate channels to evaluate: {{CANDIDATE_CHANNELS}} → Current channels already running and how they perform: {{CURRENT_CHANNELS}} → Economics: target max CAC or target CAC:LTV ratio {{TARGET_CAC_OR_RATIO}}, current LTV {{LTV}}, monthly test budget {{TEST_BUDGET}} → Team and resource constraints: {{TEAM_AND_RESOURCES}} → Any channel benchmarks I have (CPC, conversion rates, competitor data): {{BENCHMARKS}} FIRST, before scoring, check five gate inputs: the audience's job to be done, whether product market fit is established, target economics (CAC or CAC:LTV ratio), team capacity, and whether I want brand channels included. If any of these five is missing or vague, ask me numbered clarifying questions about only those items, then STOP and wait for my answers. Do not proceed on guesses for these five. For any other missing input, proceed, but state each assumption explicitly at the point where you use it. Then follow this method exactly. STEP 1: Guardrail check. If product market fit is not yet established, warn me: a channel that "failed" pre fit may have failed because of the product, messaging, or audience, not the channel. Do not kill or fully commit to channels yet in that case; recommend cheap tests only. STEP 2: Separate brand from growth. Never score brand channels (PR, SEO, community, awareness social, partnerships for reach) in the same table as growth channels. Scoring a big, risky, low confidence brand idea with a growth model always penalizes it and pushes you toward small safe bets. Split the candidates into two lists and tell me which list you are scoring. STEP 3: Score each growth channel on four factors, 1 to 10, treated as hurdles, not just addends. Hurdle rule: any factor scored 3 or below disqualifies the channel from the test shortlist regardless of its total; move it to "revisit later" and name the failing factor. a. Relevance: how confident are we the audience uses it? Higher when customer research names it, published usage stats support it, or competitors serving this audience are there. b. Ability to compete: can we break through? Lower for saturated channels; higher for niche channels, channels untouched by direct competitors, or still cheap channels. c. Potential reach: how large is the reachable relevant audience? Niche channels often score low here even when we can compete well. d. Fit with resources: time, money, team capability. Lower when the team has zero experience or setup is hard; can be raised if budget to hire expertise exists. Sum the four per channel. Use the totals to COMPARE channels, not to blindly pick the top row. STEP 4: Build a balanced portfolio, do not just take the highest scores. Recommend a MIX: long term brand channels, short and mid term growth channels, plus one or two channels to actively test. Aim for roughly three or four acquisition channels eventually so you are not over reliant on one or two that fluctuate, of which one or two are still being tested. Flag channels whose raw score is low only because the base is too small today (for example word of mouth early on) and mark them "revisit at a later stage" rather than killing them. STEP 5: Pick the single top test candidate and run the go or no go math. Tag every number in this math as GIVEN (from my context), BENCHMARK (from a named external source), or ASSUMED (your estimate). Never present an ASSUMED number as fact. a. Estimate CAC (or ROAS for ads) upfront from prior data, competitor benchmarks, or the numbers I gave. b. Model volume: monthly searches or impressions x click through rate → leads x lead to customer rate → customers per month. A good CAC on tiny volume may not be worth it. c. ALWAYS add management time and fixed costs, not just media. Value an hour of work and add it. Spread that fixed cost across the modeled monthly customers and recompute CAC. Many doomed channels get skipped once one hour of this math is done. d. If LTV is expected to rise, frame the target as a CAC:LTV ratio rather than a static CAC number. STEP 6: Write the one page test plan for that candidate: → Overall goal (why this channel). → Measure of success (X conversions at Y CAC), and the graduation rule that moves it from "testing" to "ongoing channel." → Duration (one, two, or three months) and interim benchmarks (expected result by month two, month three, hard stop date). → Budget and resources. → A backlog of what you will test (multiple angles, formats, hypotheses) so one or two failed tests do not condemn the channel. One test is one learning, not a verdict. → A mid test cutoff rule if testing more than one channel in parallel (for example, at six weeks refocus on the channel showing most potential). → A "far off the mark" rule: if performance is a multiple off target, make BIG lever changes (creative direction, landing page, offer, page speed), not micro copy tweaks, and revisit the original CAC estimate to find which assumption was wrong. OUTPUT FORMAT: 1. Clarifying questions on the five gate inputs if any are missing, then stop there. Otherwise a one line list of assumptions you are making on non gate inputs. 2. Scoring table: channel | relevance | ability to compete | potential reach | fit with resources | total | note. 3. Recommended portfolio: the mix by time horizon, with the one or two test candidates named. 4. Top candidate CAC and volume math, shown step by step including management time. 5. The one page test plan. 6. Self check. SELF CHECK before you finish: → Facts to verify: did you keep brand and growth channels in separate tables? Did you add management time to CAC, not just media? Did you tie the target to my stated CAC or CAC:LTV ratio? Did you apply the hurdle rule (any factor at 3 or below disqualifies) before recommending? → Failure modes to avoid: defaulting to familiar channels (Meta, Google, email) because they are familiar; recommending too many channels; picking only short term channels with no compounding future; treating a raw score as a decision instead of a comparison; declaring a channel dead after one test. Gather every number tagged ASSUMED into one list, and for each, tell me exactly what data would replace it and how it could change the recommendation. ``` ## Prompt versions The standard prompt above works on any model. Use these variants when you want a different tradeoff. ### Frontier model version Built for the most capable models (Claude Opus and beyond). States the goal, constraints, and quality bar up front, then trusts the model to choose its path. ```text You are a senior growth strategist who builds acquisition portfolios like a hurdle race, not a popularity contest. Goal: turn my candidate channels into a scored shortlist and a go or no go test plan for the single best candidate. Deliverable, in this order: a one line verdict naming the top test channel and whether to proceed; then the scoring; then the recommended portfolio; then the CAC and volume math; then the one page test plan. Context: → Business: {{BUSINESS_MODEL}} → Stage: {{COMPANY_STAGE}} → Product and price: {{PRODUCT_AND_PRICE}} → Audience and job to be done: {{AUDIENCE_JTBD}} → Candidate channels: {{CANDIDATE_CHANNELS}} → Current channels and performance: {{CURRENT_CHANNELS}} → Economics: target {{TARGET_CAC_OR_RATIO}}, current LTV {{LTV}}, monthly test budget {{TEST_BUDGET}} → Team and resources: {{TEAM_AND_RESOURCES}} → Benchmarks I have: {{BENCHMARKS}} Non negotiable rules: → Score brand channels (PR, SEO, community, awareness, reach partnerships) in a separate table from growth channels; a growth model always unfairly penalizes brand bets. → Score growth channels on relevance, ability to compete, potential reach, and fit with resources, treated as hurdles: any factor at 3 or below disqualifies the channel and you name the failing factor. Totals compare channels; they do not auto pick the winner. → Recommend a balanced portfolio across time horizons (aim for three or four eventual acquisition channels, one or two still in test), not the highest scoring row. Flag channels that score low only because the base is small today as "revisit later," not dead. → In the CAC math, always add management time and fixed costs spread across modeled monthly customers, not just media. Tag every number GIVEN, BENCHMARK, or ASSUMED and never present an ASSUMED number as fact. If LTV is rising, frame the target as a CAC:LTV ratio. → The test plan states an explicit graduation rule (testing to ongoing), a hard stop date, interim benchmarks, a backlog of angles so one failed test is not a verdict, and a "make big lever changes if far off target" rule. Quality bar: excellent output separates brand from growth with a stated reason, applies the hurdle rule visibly, shows CAC changing once labor is added, gathers every ASSUMED number with the data that would replace it, and hands me a plan I could walk a stakeholder through before spending a dollar. Boundaries: do not invent data or statistics. Do not pad with generic channel advice or default to familiar channels because they are familiar. If the audience job to be done, product market fit status, target economics, team capacity, or whether to include brand channels is missing, ask one focused clarifying question on the missing item instead of guessing. ``` ### Quick version Five lines or fewer, for when speed matters more than rigor. ```text Act as a growth strategist. Score my candidate channels {{CANDIDATE_CHANNELS}} for a {{BUSINESS_MODEL}} selling {{PRODUCT_AND_PRICE}} to {{AUDIENCE_JTBD}}, then name the single best one to test. Score each on relevance, ability to compete, reach, and resource fit (1 to 10); any factor at 3 or below is disqualified. For the top pick, estimate CAC against {{TARGET_CAC_OR_RATIO}} and include management time, not just media spend. Give me a one page test plan with a success metric, a hard stop date, and a graduation rule. Quality bar: recommend two or three channels max, never eight, and flag any number you assumed. ``` ## How to customize → `{{BUSINESS_MODEL}}`: your model and monetization, for example "DTC subscription supplement, monthly and annual plans" or "B2B SaaS, seat based." → `{{COMPANY_STAGE}}`: early startups skew toward testing and linear channels for momentum; established brands skew toward optimizing existing channels and adding loops. Say which. → `{{AUDIENCE_JTBD}}`: describe the job the customer hires you for, not a demographic. A wider job to be done ("a natural way to feel rested") unlocks more channels than a narrow product aware audience. → `{{TARGET_CAC_OR_RATIO}}`: a hard max CAC (for example $500) or a ratio (for example a healthy CAC:LTV of 1:5). Use the ratio when LTV is still climbing with more data. → `{{CURRENT_CHANNELS}}`: include performance so the model can favor channels that complement or share mechanics with what already works. → `{{BENCHMARKS}}`: any CPC, click through, or conversion figures you have. The more real numbers you supply, the less the model has to assume. ## What good output looks like → Brand channels and growth channels are scored in two separate tables, and the output says why they cannot share one. → The recommendation is a balanced portfolio across time horizons, not the single highest scoring row, and it names one or two channels to test rather than eight to run. → Any channel with a factor scored 3 or below is disqualified from the test shortlist with the failing factor named, not quietly carried through by a high total. → The CAC math shows media cost, then adds management hours spread across modeled monthly customers, and the CAC changes visibly once labor is included. → Every number in the math carries a GIVEN, BENCHMARK, or ASSUMED tag, and the assumed ones are gathered in the self check with the data that would replace them. → The test plan states an explicit graduation rule and a hard stop date, plus a "make big changes if far off" rule instead of endless micro tweaks. → Low scoring channels that lose only because the base is too small today are flagged "revisit later," not deleted. ## Related prompts → [Map Growth Loops and Flywheels](./map-growth-loops-and-flywheels.md): once you have chosen channels, place them in a model so you see where each fits and whether it is a linear channel or a reinforcing loop. → [Run a Product Market Fit Survey](./run-a-product-market-fit-survey.md): go here first if fit is not established, so a channel that "fails" is not really failing on product, messaging, or audience. → [Set a North Star Metric and Quarterly OKRs](./set-a-north-star-and-okrs.md): turn the winning channel test into a quarterly lever with its own target and owner.