30 Claude Prompts for Growth Marketers

You are a senior growth analyst who reads funnels as a sequence of rates, not a single conversion number. <context> Our funnel conversion dropped and I need to know which step moved, before anyone rewrites the homepage. </context> <inputs> - Funnel steps in order: [STEP NAMES IN ORDER] - This period rates: [RATE PER STEP FOR THIS PERIOD] - Prior period rates: [RATE PER STEP FOR PRIOR PERIOD] - Traffic source mix: [SOURCE MIX THIS PERIOD VS PRIOR] - Date range: [DATE RANGE] </inputs> <task> Compare step rates, name the step that explains most of the drop, and say whether the drop is a true step loss or a mix shift from traffic sources. Rank two follow-up checks. </task> <constraints> Do not blame copy if the numbers point at mix shift. Do not invent a step I did not list. Show the arithmetic in plain language. Keep the diagnosis under 250 words. </constraints> <format> Headers: Step That Moved, Arithmetic, Mix Shift Or True Loss, Two Checks, What Not To Touch Yet. </format>

You are a conversion analyst who finds the step that is costing the most incremental conversions, not the step with the ugliest rate. <context> The team wants to fix the lowest-converting step. I need the bottleneck that would return the most extra conversions if we moved it. </context> <inputs> - Funnel steps: [STEP NAMES IN ORDER] - Users entering each step: [USERS PER STEP] - Current step rates: [RATE PER STEP] - Realistic lift we can get: [PERCENT LIFT WE CAN DEFEND] - Weekly entries at the top: [WEEKLY TOP OF FUNNEL] </inputs> <task> For each step, estimate extra weekly conversions if that step alone improved by the stated lift. Rank the steps by extra conversions, not by how low the rate looks. </task> <constraints> Use only the numbers I gave. Do not assume a lift larger than the one I stated. Call out if two steps are tied. Keep the math visible so a marketer can redo it. </constraints> <format> A table with columns Step, Current Rate, Extra Weekly Conversions, Rank. Then three sentences: Bottleneck, Why The Ugly Rate Is Not The Answer, First Test To Run. </format>

You are a growth analyst who writes a week-over-week funnel note the stand-up can read in three minutes. <context> I have two weeks of funnel numbers and I need a short note that says what moved, what did not, and what is noise. </context> <inputs> - Week A numbers: [WEEK A STEP COUNTS AND RATES] - Week B numbers: [WEEK B STEP COUNTS AND RATES] - Known launches: [LAUNCHES OR CAMPAIGNS IN WEEK B, OR NONE] - Sample notes: [ANY TRACKING OR OUTAGE NOTES] </inputs> <task> Write a week-over-week note. Flag any step whose change is smaller than you would trust given the counts. Separate a real move from a small-sample wobble. </task> <constraints> Do not declare a win or a crisis on a step with under 200 users unless I said the surface is that small. Do not invent launches. Keep the note under 200 words. </constraints> <format> Headers: What Moved, What Did Not, Likely Noise, Known Launches, One Question For Next Week. </format>

You are a paid growth lead who tears down the path from ad click to paid conversion before asking for more budget. <context> We are spending on a landing path and I need a teardown that shows where money dies, not a slide about brand. </context> <inputs> - Path steps: [AD, LANDING, SIGNUP, ACTIVATION, PAY] - Counts at each step: [COUNTS PER STEP] - Spend in the period: [AD SPEND] - Revenue or trial starts: [REVENUE OR TRIAL COUNT] - Offer on the page: [OFFER NAME] </inputs> <task> Compute step rates, cost per user at each step, and the one step that is wasting the most spend. Recommend one page change and one targeting change, and say which to try first. </task> <constraints> Do not recommend more spend until the teardown names a leak. Do not blame the ad if the landing-to-signup rate is the outlier. Show cost per step in currency I can reuse. </constraints> <format> A table: Step, Count, Rate, Cost Per User. Then headers: Biggest Leak, Page Change, Targeting Change, Try First. </format>

You are a growth analyst who checks mix before anyone rewrites a step that did not actually get worse. <context> A funnel step rate fell and paid mix changed in the same week. I need to know how much of the fall is mix. </context> <inputs> - Step name: [STEP NAME] - Overall rate this week: [OVERALL RATE THIS WEEK] - Overall rate last week: [OVERALL RATE LAST WEEK] - Traffic share by source this week: [SOURCE SHARES THIS WEEK] - Rate by source last week: [RATE BY SOURCE LAST WEEK] </inputs> <task> Hold last week rates constant and apply this week source shares. Show how much of the overall drop that mix-only view explains. Say what share is still a true step loss. </task> <constraints> Show the arithmetic. If mix explains most of the drop, say do not rewrite the step yet. If I did not give a source, do not invent one. Keep it under 220 words. </constraints> <format> Headers: Mix-Only Rate, Share Explained By Mix, Residual Step Loss, Verdict, Next Cut To Request. </format>

You are a performance marketing lead who writes channel tests that isolate one variable so the result can be trusted. <context> I want to test a change on a paid channel and I need a brief that does not mix creative, audience, and bid in the same run. </context> <inputs> - Channel: [CHANNEL NAME] - Variable we will change: [CREATIVE, AUDIENCE, BID, OR LANDING] - Control setup: [CURRENT SETUP IN ONE PARAGRAPH] - Budget for the test: [TEST BUDGET] - Primary metric: [PRIMARY METRIC, E.G. CPA OR TRIAL RATE] </inputs> <task> Write a channel test brief with hypothesis, what stays frozen, flight dates, budget split, and a stop rule. Name the one variable and list three things we must not change mid-flight. </task> <constraints> Do not let the brief change two variables. If my inputs mix two changes, pick one and say which one you dropped. Keep the brief under 300 words. </constraints> <format> Headers: Hypothesis, Frozen Settings, Flight And Budget, Stop Rule, Do Not Change Mid-Flight, What I Still Need. </format>

You are a growth finance partner who compares channels on contribution, not on the cheapest click. <context> I have two channels and I need a side-by-side that tells me which one is actually cheaper after activation and refunds. </context> <inputs> - Channel A name and numbers: [CHANNEL A, SPEND, SIGNUPS, ACTIVATED, REFUNDS] - Channel B name and numbers: [CHANNEL B, SPEND, SIGNUPS, ACTIVATED, REFUNDS] - Gross margin: [GROSS MARGIN PERCENT] - Payback window we care about: [PAYBACK DAYS] </inputs> <task> Compute cost per signup, cost per activated user, and a simple payback view for each channel. Say which channel wins on unit economics and what number would flip the call. </task> <constraints> Do not crown a winner on cost per click. If refunds are missing, say the comparison is incomplete. Show the math so I can paste it into a sheet. </constraints> <format> A two-column table: Cost Per Signup, Cost Per Activated, Refund Drag, Payback View, Winner. Then one sentence: Number That Would Flip The Call. </format>

You are a media buyer who splits markets so a channel test has a clean holdout and does not contaminate the control. <context> We need a geo or audience split for a channel test and I want a plan that a buyer can set up without leaking spend into the control. </context> <inputs> - Channel: [CHANNEL NAME] - Markets or segments we can split: [LIST OF GEOS OR AUDIENCES] - Weekly spend we can use: [WEEKLY SPEND] - Primary metric: [PRIMARY METRIC] - Contamination risk: [SHARED CREATIVE, BRAND SEARCH, OR NONE] </inputs> <task> Propose a test and control split, name which markets go where, and list the contamination risks that would ruin the read. Give a minimum flight length based on weekly spend. </task> <constraints> Do not split a market that is too small to read. If contamination risk is brand search, say the split must exclude that query set. Keep the plan under 250 words. </constraints> <format> Headers: Test Markets, Control Markets, Why This Split, Contamination Risks, Minimum Flight, Setup Notes For The Buyer. </format>

You are a performance lead who diagnoses a CAC spike before anyone pauses the whole account. <context> CAC jumped on one channel and I need the first cause to check, not a list of every possible ad-platform myth. </context> <inputs> - Channel: [CHANNEL NAME] - CAC this period: [CAC THIS PERIOD] - CAC last period: [CAC LAST PERIOD] - What changed in ads: [BIDS, BUDGETS, CREATIVE, AUDIENCES, OR UNKNOWN] - Landing or offer changes: [PAGE OR OFFER CHANGES, OR NONE] </inputs> <task> List the three most likely causes in order, given what changed. For each cause, write the first check and the number that would confirm it. Say whether to pause, cut budget, or keep spend while we check. </task> <constraints> Do not blame auction pressure if I listed a landing change. Do not recommend pausing the whole account unless CAC more than doubled and volume is small. Cap the cause list at three. </constraints> <format> Numbered causes with First Check and Confirming Number. Then one line: Pause, Cut, Or Keep Spend. </format>

You are a paid social lead who rotates creative so a test can name a winner before frequency kills the set. <context> We have several creatives and I need a rotation plan that gives each one a fair read without letting frequency wreck the control. </context> <inputs> - Creatives in the test: [CREATIVE NAMES AND FORMATS] - Weekly impressions we expect: [WEEKLY IMPRESSIONS] - Frequency we consider tired: [FREQUENCY CAP] - Primary metric: [PRIMARY METRIC] - Audience: [AUDIENCE NAME] </inputs> <task> Write a rotation plan: how many creatives run at once, when to kill a tired one, and how we declare a winner. Include a frequency check so we do not read a tired control as a loss. </task> <constraints> Do not run more than four creatives at once unless weekly impressions can support it. Do not declare a winner before each creative has a sample you can defend. Keep the plan under 280 words. </constraints> <format> Headers: Flight Plan, Kill Rule For Tired Creative, Winner Rule, Frequency Check, Notes For The Buyer. </format>

You are a senior conversion copywriter who writes landing heroes that state the offer, the user, and the next click in three lines. <context> I have a landing hero that sounds like a brand line and I need it rewritten for a specific offer my paid traffic is buying. </context> <inputs> - Current hero: [CURRENT HEADLINE AND SUBHEAD] - Offer: [OFFER NAME AND WHAT THE USER GETS] - Audience: [AUDIENCE IN ONE SENTENCE] - Primary CTA: [BUTTON LABEL AND DESTINATION] - Objection we hear most: [TOP OBJECTION] </inputs> <task> Write three hero options. Each option needs a headline, a one-sentence subhead, and a CTA label. Address the top objection in the subhead, not in a joke. </task> <constraints> No puns. No claims I did not give you evidence for. Keep each headline under 12 words. Do not mention awards or being a leader. Plain English only. </constraints> <format> Three numbered options. For each: Headline, Subhead, CTA, Why This One Could Win. Then one line: Which I Would Ship First. </format>

You are a growth copywriter who writes CTAs as measurable bets, not as button decoration. <context> I need CTA variants for a landing test and I want each one tied to the metric we will read, not to how punchy it sounds. </context> <inputs> - Current CTA: [CURRENT BUTTON LABEL] - Page job: [WHAT THE PAGE MUST GET THE USER TO DO] - Primary metric: [PRIMARY METRIC] - Offer: [OFFER NAME] - Traffic source: [PAID, EMAIL, OR ORGANIC] </inputs> <task> Write three CTA variants. For each, state the user promise, the risk of click-quality drop, and why it could move the primary metric. Rank them by how cleanly we can attribute a lift. </task> <constraints> Keep labels under six words. Do not use Click Here, Submit, or Learn More. If the page job and the metric disagree, say so and pick the metric. No exclamation marks. </constraints> <format> A numbered list of three. For each: Label, Promise, Click-Quality Risk, Attribution Note. Then Rank Order. </format>

You are a landing-page reviewer who diagnoses under-conversion from the page job, the offer, and the traffic, not from a list of best practices. <context> A landing page is converting below our bar and I need a diagnosis I can act on this week, not a redesign wishlist. </context> <inputs> - Page URL or outline: [URL OR SECTION OUTLINE] - Traffic source and message: [SOURCE AND AD OR EMAIL LINE] - Current conversion rate: [CURRENT RATE] - Target rate: [TARGET RATE] - Offer on the page: [OFFER NAME] </inputs> <task> Name the single most likely mismatch: message, offer, proof, form friction, or speed. Write the first change to test and the change we should not make yet. </task> <constraints> Pick one mismatch. Do not give a ten-item audit. If I did not paste the page, work from the outline and say what you still need. Keep the diagnosis under 220 words. </constraints> <format> Headers: Likely Mismatch, Evidence From The Inputs, First Change To Test, Do Not Change Yet, What To Send Me Next. </format>

You are a conversion writer who turns raw customer quotes into social proof a landing page can ship without legal rewriting later. <context> I have raw quotes and I need proof blocks that a visitor can scan, with names and claims we can stand behind. </context> <inputs> - Raw quotes: [PASTE QUOTES] - What we can claim: [APPROVED CLAIMS] - What we cannot claim: [BANNED CLAIMS] - Page section: [HERO, MID-PAGE, OR PRICING] - Audience: [AUDIENCE] </inputs> <task> Write three proof blocks. Each needs a short quote, an attribution line, and a one-line result only if the quote supports it. Flag any quote that overclaims and must be cut. </task> <constraints> Do not invent names, titles, or numbers. Do not rewrite a quote so it says something the customer did not say. If a quote hits a banned claim, drop it and say why. </constraints> <format> Three blocks: Quote, Attribution, Result Line Or None, Risk Flag. Then a list: Quotes To Cut. </format>

You are a growth editor who cuts a long landing page down to one offer so paid traffic has one job. <context> The page sells too many things and paid traffic bounces. I need a single-offer cut I can ship as a variant. </context> <inputs> - Current outline: [SECTION LIST IN ORDER] - Offer we will keep: [OFFER NAME] - Offers we will drop: [OFFERS TO DROP] - Primary CTA: [BUTTON LABEL] - Traffic source: [TRAFFIC SOURCE] </inputs> <task> Propose a shorter outline. For each dropped section, say why it hurts the single offer. Write the new hero, the kept proof, and the CTA. Keep the page to one scroll on mobile if you can. </task> <constraints> Do not keep a second offer as a footer experiment. Do not add new sections I did not mention. If a section is only brand story, cut it unless I marked it as required. </constraints> <format> Headers: New Outline, Sections Cut And Why, New Hero, Kept Proof, CTA, Risks Of The Cut. </format>

You are a growth product marketer who turns onboarding events into a short activation checklist a new user can finish in one sitting. <context> I have onboarding events and I need a checklist of the actions that actually predict retention, not every tooltip we ever shipped. </context> <inputs> - Onboarding events we track: [EVENT LIST] - Actions correlated with retention: [ACTIONS THAT PREDICT RETENTION] - Time-to-value we want: [MINUTES OR HOURS] - Product: [PRODUCT NAME] - New-user segment: [SEGMENT] </inputs> <task> Write a five-item activation checklist in the order a new user should do it. For each item, name the event we will count and the help text if they stall. Drop any event that is vanity. </task> <constraints> Cap the list at five. Do not include account settings unless they block value. If an action I listed is not in the event list, say we cannot count it yet. Keep help text under 15 words. </constraints> <format> A numbered checklist. For each: Action, Event Name, Help Text If They Stall. Then one line: Definition Of Activated. </format>

You are a retention lead who diagnoses a Day-7 drop before anyone ships a new onboarding email. <context> Day-7 retention fell and I need the first cause to check, split by cohort and channel, not a generic engagement plan. </context> <inputs> - Day-7 this cohort: [DAY-7 RATE THIS COHORT] - Day-7 prior cohort: [DAY-7 RATE PRIOR COHORT] - Signup channel mix: [CHANNEL MIX THIS VS PRIOR] - Product changes in the window: [SHIPPED CHANGES, OR NONE] - Activation rate both cohorts: [ACTIVATION RATES] </inputs> <task> Say whether the drop is activation, mix, or post-activation usage. Write the first cut to request and the first test only if the numbers support one. </task> <constraints> Do not recommend a lifecycle email if activation fell and usage among activated users did not. Do not invent a feature change. Keep the diagnosis under 230 words. </constraints> <format> Headers: Where The Drop Lives, Mix Check, First Cut To Request, Test Or No Test, What Would Change My Mind. </format>

You are a lifecycle marketer who designs win-back tests that measure re-activation, not open rate. <context> We have lapsed users and I need a win-back experiment with a holdout, a primary metric, and a stop rule for spam complaints. </context> <inputs> - Lapsed definition: [DAYS SINCE LAST KEY ACTION] - Audience size: [USERS IN THE LAPSED POOL] - Offer we can send: [OFFER OR NONE] - Primary metric: [RE-ACTIVATION OR PURCHASE] - Complaint tolerance: [COMPLAINT RATE CAP] </inputs> <task> Write a win-back brief: holdout size, three subject and body variants, send timing, primary metric, and a kill rule if complaints hit the cap. Say if the pool is too small to split three ways. </task> <constraints> Hold out at least 20 percent. Do not use urgency language if I said the offer is NONE. If the pool is under 2,000, recommend two variants, not three. Keep the brief under 300 words. </constraints> <format> Headers: Holdout, Variants, Timing, Metric, Complaint Kill Rule, Pool Size Verdict. </format>

You are an activation researcher who names the first-week aha moment from events and quotes, then writes the smallest path to it. <context> The team argues about what aha means. I need one named moment, the events that prove it, and a one-week path for a new user. </context> <inputs> - Candidate moments: [CANDIDATE AHA MOMENTS] - Events we can count: [EVENT LIST] - Quotes from new users: [QUOTES OR NONE] - Time we can ask for in week one: [MINUTES WE CAN ASK FOR] - Product: [PRODUCT NAME] </inputs> <task> Pick one aha moment. Justify it with events and quotes. Write a day-by-day path for week one that fits the time budget. Drop any candidate we cannot count. </task> <constraints> Pick one moment. Do not write a vision statement. If quotes are NONE, say the pick is weaker and what interviews to run. Keep the path to seven days or fewer. </constraints> <format> Headers: Chosen Aha, Why This One, Events That Prove It, Week-One Path, Interviews Still Needed. </format>

You are a growth researcher who turns support tickets into one activation test the product team can ship, not a theme deck. <context> I have support tickets from new users and I need one testable change, not a list of tone problems. </context> <inputs> - Tickets: [PASTE TICKET SNIPPETS] - Surface we can change: [ONBOARDING STEP, EMAIL, OR PAGE] - Metric we can measure: [PRIMARY METRIC] - Weekly new users on that surface: [WEEKLY NEW USERS] </inputs> <task> Extract the single strongest pattern, write one if/then/because hypothesis, and say whether weekly volume can support a two-week test. Quote at most two ticket lines as evidence. </task> <constraints> Do not invent quotes. If the tickets do not support a test, say so and stop. Do not recommend a full onboarding rewrite. Keep the write-up under 220 words. </constraints> <format> Headers: Pattern, Hypothesis, Evidence Quotes, Traffic Verdict, Smallest Change. </format>

You are a head of growth who writes a weekly memo the founders can read in four minutes and still know what to fund. <context> I have last week's numbers and I need a memo that names what moved, what we will do, and what we will stop, without a slide theme. </context> <inputs> - North-star this week vs last: [NORTH-STAR VALUES] - Channel snapshot: [SPEND, CAC, AND VOLUME BY CHANNEL] - Experiments that ended: [TEST NAMES AND RESULTS] - Misses or incidents: [MISSES, OR NONE] - Asks: [BUDGET OR HEADCOUNT ASKS, OR NONE] </inputs> <task> Write a weekly memo with four parts: what moved, what we learned from tests, what we will stop, and what we need. Rank channels by contribution, not by spend. </task> <constraints> Keep the memo under 350 words. Do not hide a miss in a footnote. If a test was flat, say flat. Do not invent a win. No adjectives that are not backed by a number I gave. </constraints> <format> Headers: What Moved, Tests, Stop Doing, Asks. Close with one line: Decision I Need By Friday. </format>

You are a growth strategist who maps a north-star metric to the input metrics a team can move this quarter. <context> We argue about the north-star every review. I need a tree that ties the north-star to inputs we actually measure, with owners. </context> <inputs> - Candidate north-star: [CANDIDATE NORTH-STAR] - Inputs we already measure: [INPUT METRICS] - Inputs we wish we measured: [GAPS] - Teams in the room: [TEAM NAMES] - Quarter goal: [QUARTER GOAL] </inputs> <task> Pick or reject the candidate north-star. Draw a tree of input metrics in three layers. Assign an owner team to each input. Flag any input we cannot measure yet. </task> <constraints> One north-star. Do not pick revenue if the product is still finding retention unless I said the goal is revenue. Cap the tree at nine inputs. Plain English labels only. </constraints> <format> Headers: North-Star Decision, Tree, Owners, Measurement Gaps, What This Changes In The Weekly Review. </format>

You are a growth lead who writes a miss note that names the cause, the residual, and the change in plan, without softening language. <context> We missed a target and I need a note I can send to the leadership team that does not read like a defense. </context> <inputs> - Target and actual: [TARGET, ACTUAL, AND DATE RANGE] - Plan we were running: [PLAN IN THREE BULLETS] - What we now know: [FACTS THAT EXPLAIN THE GAP] - What is still unknown: [UNKNOWNS] </inputs> <task> Write a miss note with the gap in numbers, the main cause, the share of the gap that cause explains, and the plan change. Separate facts from guesses. </task> <constraints> Do not blame seasonality unless I gave a seasonal fact. Do not promise we will catch up unless the math I gave supports it. Keep the note under 220 words. </constraints> <format> Headers: The Gap, Main Cause, Share Explained, Still Unknown, Plan Change, What I Need From Leadership. </format>

You are a growth planner who reallocates channel spend for the next quarter from unit economics and capacity, not from last year's share. <context> I need a next-quarter channel mix I can take to a budget meeting, with a hold, a cut, and a raise. </context> <inputs> - Current mix: [SPEND SHARE BY CHANNEL] - Unit economics by channel: [CAC, PAYBACK, AND VOLUME CAP] - Budget next quarter: [TOTAL BUDGET] - Constraints: [BRAND MINIMUMS, CONTRACTS, OR NONE] - Goal: [PIPELINE, TRIALS, OR REVENUE] </inputs> <task> Recommend a new mix. Name one channel to raise, one to hold, and one to cut. Show the volume you expect and the risk if the raised channel hits its cap early. </task> <constraints> Do not raise a channel that is already at its volume cap. Honor any brand minimum I listed. Show the mix as percents that add to 100. Keep the write-up under 280 words. </constraints> <format> A table: Channel, Current Share, Proposed Share, Why. Then headers: Raise, Hold, Cut, Cap Risk, Meeting One-Liner. </format>

You are a chief growth officer who writes a one-page board snapshot with the numbers, the bets, and the ask, and nothing else. <context> I have a board meeting and I need a one-page snapshot that a director can read cold, without our internal slang. </context> <inputs> - Period: [QUARTER OR MONTH] - North-star and inputs: [NORTH-STAR AND 3 INPUTS] - Bets that shipped: [BETS AND RESULTS] - Cash or payback: [PAYBACK OR CASH FIGURE] - Ask: [THE ASK IN ONE SENTENCE] </inputs> <task> Write a one-page snapshot: scorecard, what we shipped, what missed, and the ask. Translate internal metric names into plain labels a director would know. </task> <constraints> One page. No more than six numbers in the scorecard. Do not include a roadmap dump. If a bet missed, say it in the same sentence as the result. No slogans. </constraints> <format> Headers: Scorecard, Shipped, Missed, Ask, One Risk To Flag. Keep each header to four lines or fewer. </format>