# The Marketing Plan Engine **A modular AI prompt system that researches your market, sizes up your competitors, picks your channels, and hands you a 90-day plan you can actually run.** Works in ChatGPT, Claude, Gemini, Copilot, or Perplexity. No special tools required. --- ## How to use this in the next 30 minutes 1. **Fill out the Business Brief (Module 0).** Ten minutes. This is the single biggest driver of output quality — a vague brief produces a generic plan, and a generic plan is worthless. 2. **Pick your path.** *Fast path:* paste the Master Prompt + your Brief — one long output, one sitting. *Deep path (recommended):* run Modules 1 → 5 in order in the **same chat**, pasting your Brief once at the start. Each module builds on the last: better research, more detail, fewer hallucinations. 3. **Turn on web search / research mode** in whatever tool you're using. Without it the AI is guessing from memory. With it, it's citing sources. 4. **Run the Reality Check pass** at the end. This is the step most people skip and it's the one that separates a plan you'll use from a plan you'll file. > **Rule of thumb:** Budget 45–90 minutes for the deep path. You are replacing a $15,000–$40,000 consulting engagement. Ninety minutes is a bargain. --- ## Why these prompts work (the technique, for anyone who wants to write their own) Steal these six moves and every prompt you write gets better. | # | Technique | What it does | |---|---|---| | 1 | **Assign a role with a track record** | "You are a marketing strategist who has built go-to-market plans for 200+ service businesses" pulls the model toward expert-register output instead of blog-post register. | | 2 | **Force a clarifying-questions gate** | Making the AI ask *before* it answers prevents it from inventing your business for you. | | 3 | **Specify the output schema** | Tables, scoring rubrics, and named sections. If you don't specify structure, you get an essay. You can't execute an essay. | | 4 | **Demand source citation + confidence labels** | `[VERIFIED]` / `[ESTIMATE]` / `[ASSUMPTION]` tagging is the single best anti-hallucination move available to a non-technical user. | | 5 | **Make the numbers reconcile** | "Budget allocations must sum to the stated budget. Show the math." Models drift on arithmetic unless told to check. | | 6 | **Add a self-critique pass** | "Now critique this plan as a skeptical CFO would" catches the soft spots before your bank account does. | --- ## MODULE 0 — The Business Brief Fill this in once. Paste it at the top of every module. Ten minutes here saves you three hours of bad output. Don't have an answer? Write `UNKNOWN` — the prompts are built to handle gaps and will tell you which ones matter. ``` === BUSINESS BRIEF === COMPANY Business name: What we sell (in one plain sentence, no jargon): Years in business: Team size: Website: Business model: [ B2B / B2C / both ] [ local / regional / national / online-only ] OFFERS & ECONOMICS Top 3 products or services, with current price and rough gross margin: 1. 2. 3. Average transaction value: Do customers buy once or repeatedly? Typical repeat rate / lifetime value: Typical sales cycle length (first contact to money): Current monthly revenue (or annual): Revenue goal for next 12 months: GEOGRAPHY Cities / counties / metro areas we serve today: How far will you realistically travel or ship: Anywhere you'd like to expand but haven't: CUSTOMERS Describe your best 3 customers (who they are, what triggered the purchase, why they chose you): Who do you NOT want as a customer: Where did your last 10 customers come from (referral, Google, social, walk-in, etc.): COMPETITION Competitors you know by name: Who do you lose deals to, and why: What do you do better than anyone else (be specific — "great service" doesn't count): CURRENT MARKETING What you do today and roughly what it costs: What has worked: What has clearly failed: Email list size: Social following by platform: Google Business Profile: [ claimed / not claimed ] — review count and rating: CONSTRAINTS — be honest, this drives everything Monthly marketing budget you can sustain for 12 months: $ Hours per week you or your team can spend on marketing: Skills on hand: [ writing / design / video / ads / none ] Anything you refuse to do (e.g., no TikTok, no cold calling): Capacity ceiling — how many new customers per month can you actually serve: GOALS Primary goal for the next 90 days (one thing): How you'll know it worked (a number): === END BRIEF === ``` > **The constraints section is the most important part.** A plan that ignores your budget, your hours, and your capacity ceiling isn't a plan — it's a wish list. Most AI marketing output fails here because the user never told it the truth about limits. --- ## The Master Prompt Use this when you want the whole plan in one pass. Paste it, then paste your Business Brief underneath. ``` You are a senior marketing strategist and market researcher. You have spent 20 years building go-to-market plans for small and mid-sized businesses — service firms, local retail, trades, professional practices, and small B2B companies. You are known for two things: plans grounded in real data rather than vibes, and plans small teams actually execute because they respect the operator's real budget, hours, and capacity. Your job: produce a complete, research-backed marketing plan from the Business Brief below. === OPERATING RULES === 1. RESEARCH FIRST. Use web search for every factual claim about market size, growth, competitors, pricing, platform costs, and industry benchmarks. Cite the source and the date of the data inline. Prefer primary sources: U.S. Census County Business Patterns and Statistics of U.S. Businesses, Bureau of Labor Statistics (QCEW, OES), FRED, state licensing databases, trade association reports, SEC filings, Google Trends, and the competitors' own websites and public pricing pages. 2. LABEL EVERY NUMBER. Tag each data point: [VERIFIED] - sourced, with link and date observed [ESTIMATE] - calculated by you; show the math and the inputs [ASSUMPTION] - you had to guess; state what would confirm or kill it NOT FOUND - you looked and it isn't public. Say this often; it is the honest answer for most competitor pricing, ad spend, and headcount data. Never present an assumption as a fact. A report with no NOT FOUNDs is a failure, because this data is not all publicly available. NEVER infer from general industry knowledge: competitor pricing, ad spend, follower counts, engagement rates, headcount, revenue, or review sentiment. Those seven are where fabrication happens. 3. RESPECT THE CONSTRAINTS. Every recommendation must fit the stated budget, weekly hours, and skills on hand. All budget figures must sum to the stated budget — show the arithmetic. If the goal is not achievable within the constraints, say so directly in the first section and show what budget or timeline would be required. Do not quietly recommend a $10,000/month program to someone with $1,500. 4. BE SPECIFIC. "Post on social media" is not a recommendation. "Publish 3 before/after carousels per week to Instagram and Facebook at 11am Tue/Thu/Sat, sourced from job photos your crew already takes" is a recommendation. Name the platform, the cadence, the cost, the owner, and the metric. 5. NO FILLER. Skip the introductory throat-clearing, the definitions of basic terms, and the closing summary of what you just said. Lead with the answer. === BEFORE YOU BEGIN === Ask me up to 7 clarifying questions — only the ones that would materially change your recommendations. Then wait for my answers before producing the plan. === DELIVER, IN THIS ORDER === PART 1 — EXECUTIVE SUMMARY (write this last, place it first) - The single biggest growth opportunity you found, and why - The one thing currently costing the most money or momentum - Top 3 recommended moves, in priority order, with expected impact - Honest verdict: is the stated goal reachable on the stated budget? Yes/No/Conditional, and what it would actually take PART 2 — MARKET ANALYSIS - Market definition and NAICS code - Market sizing: TAM / SAM / SOM for the specific geography, with the math shown - Growth rate and the 3–5 forces driving it (economic, demographic, regulatory, technological, cultural) - Seasonality curve: which months are strong, which are weak, and what that means for when to spend - Three to five distinct customer segments. For each: size, buying trigger, what they value most, price sensitivity, where they spend attention, and how hard they are to reach - SEGMENT SCORECARD — rank every segment 1–10 on each of: market size, growth, willingness to pay, ease of reach, competitive intensity (inverted: 10 = wide open), and fit with our current capabilities. Weight them 20/15/20/15/15/15, compute a weighted total, rank, and name the top 2 to attack and 1 to deliberately ignore. Explain the "ignore" choice — focus is the whole point. PART 3 — COMPETITIVE & PRICING INTELLIGENCE - FIRST, audit US: our Google Business Profile (claimed, review count, rating, date of most recent review, photos, completeness), and what our website makes clear in five seconds. This is the cheapest, highest-return finding in the whole analysis. - State your geographic boundary explicitly and confirm each competitor has a verified address inside it — not just a service-area page claiming coverage. - Identify 5–8 real, named competitors (direct, indirect, and substitute). Search for them; do not invent them. Include lead-generation marketplaces (Angi, Thumbtack, Yelp paid placement, and the equivalents in this industry) as substitutes — for a business relying on word of mouth, those often compete for leads more directly than the contractors on the list. - COMPETITOR TABLE: name, positioning statement, target segment, estimated size, published pricing, primary channels, review count and average rating, apparent strengths, exploitable weaknesses - PRICING LANDSCAPE: where their prices actually sit, what's bundled, what's charged separately, any pricing models in use (hourly, fixed, subscription, tiered, value-based) - WHERE WE SIT: are we priced high, low, or at market — and is that deliberate or accidental? Give a specific pricing recommendation with the reasoning — but ONLY if you found published prices for at least 3 competitors on a comparable offer. If you didn't, say so and tell me how to get the data (mystery-shop calls, requesting quotes). If nobody in this market publishes prices, note that publishing ours is itself a differentiator and an AI-search advantage. - POSITIONING MAP: two axes that matter in this market, plot every competitor and us, and name the empty quadrant we could own - MESSAGING GAP: what every competitor claims (the sea of sameness), and the claim nobody has taken that we can credibly own - THEIR MARKETING PLAYBOOK: for each significant competitor — website quality and conversion approach, SEO visibility and what they rank for, paid advertising (evidence: Google Ads Transparency Center, Meta Ad Library), social platforms and posting cadence and what actually gets engagement, email and content programs, review generation strategy, local and community presence, partnerships. Flag what's working for them that we should copy, and what they're wasting money on. PART 4 — CHANNEL STRATEGY - CUSTOMER JOURNEY: for our top segment, map how they actually go from "unaware" to "bought" — the trigger, where they look first, who they ask, what they compare, what closes the decision. Name the specific moments where marketing can intervene. - CHANNEL SCORECARD — evaluate every plausible channel: local SEO and Google Business Profile, organic search and content, AI search visibility (ChatGPT/Gemini/Perplexity — increasingly how buyers find vendors), Google Ads, Meta ads, other paid social, email, SMS, direct mail, referral and partner programs, review generation, community and events, PR and earned media, trade associations, speaking, podcasts/video, outdoor, radio. For each: estimated cost to start, realistic cost per lead for this industry and geography [cite benchmarks], time to first result, weekly hours required, difficulty given our skills, and a 1–10 fit score. Rank them. - RECOMMENDED CHANNEL MIX: 2–3 primary channels, 1–2 secondary, and an explicit "not now, and here's why" list. Justify each pick against the scorecard. - BUDGET ALLOCATION: dollars and percentage per channel per month. Must sum to the stated budget. Show the math. Include a small explicit test budget (10–15%) for learning. - MESSAGING: for the top 2 segments, give the core value proposition, 3 proof points, the primary objection and the rebuttal, and 5 headline options written in plain customer language — not marketing language. - EXPECTED RETURN: given the budget and benchmark conversion rates, model expected leads, customers, and revenue per month. Show the assumption chain. State the breakeven point and when to expect it. PART 5 — THE 90-DAY PLAN Organize as three 30-day phases: Days 1–30 FOUNDATION, Days 31–60 ACTIVATION, Days 61–90 SCALE. - For each phase: the objective, the 5–8 specific tasks, owner, estimated hours, cost, and the success metric with a target number - WEEK-BY-WEEK TABLE for all 13 weeks — what gets done each week, concretely - QUICK WINS: everything achievable in week 1 for under $100 that produces visible movement. Lead with these; early momentum is what keeps a plan alive. - ASSETS NEEDED: every piece of content, creative, page, or system required, and when - MEASUREMENT: the 5–7 metrics that matter, where each is tracked, the baseline, the 90-day target, and the review cadence. Include a simple weekly scorecard layout. - DECISION GATES: at day 30 and day 60, what result means "keep going," what means "adjust," and what means "kill it." Be specific with numbers. - RISKS: the 5 most likely ways this plan fails, with early warning signs and the mitigation for each PART 6 — THE AI EXECUTION LAYER - Map every recurring marketing task in this plan against AI suitability: AI DOES IT (near-autonomous), AI DRAFTS IT (human edits and approves), AI ASSISTS (human leads), HUMAN ONLY (relationship, judgment, or trust-critical). Be honest about the last category — AI-generated relationship building is worse than none. - Estimate hours saved per week and what those hours should be redeployed into - Write 8–12 ready-to-use, copy-paste prompts for the specific recurring tasks in THIS plan, each with placeholders in [BRACKETS] and a note on what to check before publishing - Recommend a simple weekly AI workflow: what to run Monday, mid-week, and Friday PART 7 — REALITY CHECK Now step outside the plan and critique it as a skeptical CFO who has watched marketing budgets disappear before: - Where is this plan most likely to fail? - Which assumptions are load-bearing but unverified? - What is the cheapest test that would validate or kill the riskiest assumption in under 30 days? - What did I ask for that I shouldn't have, and what did I fail to ask about? === FORMAT === Use clear headers, tables wherever comparison is involved, and short paragraphs. Bold the recommendations. Put every number in a table, not in prose. === BUSINESS BRIEF === [PASTE YOUR BUSINESS BRIEF HERE] ``` --- ## MODULE 1 — Market & Customer Research *Goal: stop guessing which customers to chase. Run this first; everything downstream depends on it.* ``` You are a market research analyst who specializes in sizing and segmenting markets for small businesses. You work from primary data sources, you show your math, and you never pass off a guess as a finding. TASK: Research the market for the business described below and tell me which segments deserve my attention and budget — and which ones I should deliberately walk away from. RESEARCH REQUIREMENTS Use web search throughout. Prioritize these sources and cite each with a link and date: - U.S. Census County Business Patterns and Statistics of U.S. Businesses (establishment counts and firm-size distribution by industry and county) - Census Business Builder and American Community Survey (local demographics, income, household counts) - Bureau of Labor Statistics QCEW and OES (employment, wages, industry concentration by county) - FRED (economic context, trend lines) - Industry trade associations and their published state-of-the-industry reports - Google Trends (demand trajectory and seasonality for the relevant search terms) - State licensing and business registration databases (real competitor counts) - Local news, chamber of commerce data, and planning/permit records for growth signals Label every figure [VERIFIED], [ESTIMATE] (show the math), or [ASSUMPTION] (state how to confirm it). DELIVER 1. MARKET DEFINITION - Precise definition of the market we are actually in, and the NAICS code - Adjacent markets we could credibly enter, and the barrier to each 2. MARKET SIZING — show the calculation, don't just state a number - TAM: total spend on this category in the broadest relevant geography - SAM: the portion we could serve given our geography, capacity, and offer - SOM: what a realistic share looks like in 12 months, given our size and budget - Method note: state which sizing approach you used (top-down, bottom-up, or value theory) and why, and which inputs would most change the answer if wrong 3. DEMAND DYNAMICS - Growth rate for this market, with source, over the last 3 years and projected - The 3–5 forces driving demand up or down: economic, demographic, regulatory, technological, cultural. Rate each as tailwind or headwind and how durable it is. - Seasonality: month-by-month demand index (100 = average). Name the peak months, the dead months, and the lead time needed to capture the peak — this determines when money should be spent, which most small businesses get backwards. - Demand signals to watch: the specific leading indicators that would tell me demand is shifting before my revenue does 4. SEGMENTATION — identify 4–6 genuinely distinct segments For each segment, give me: - Name and one-line description of who they are - Estimated size in the geography (count of households or businesses), with source - The buying trigger: the specific event that makes them start looking - Their decision criteria, ranked — what actually determines who they pick - Price sensitivity: high / medium / low, with reasoning - Where they spend attention: specific platforms, publications, groups, physical places - Who influences the decision besides the buyer - The objection that stops them from buying - Estimated cost and difficulty to reach them - Current competitive intensity in this segment 5. SEGMENT SCORECARD Score each segment 1–10 on: Size, Growth, Willingness to Pay, Ease of Reach, Competitive Openness (10 = uncontested), Fit With Our Capabilities. Weights: 20% / 15% / 20% / 15% / 15% / 15%. Compute weighted totals. Rank. Present as a table with the weighted total in the final column. 6. THE CALL - The 1–2 segments to build the plan around, and specifically why they beat the others - The 1 segment to deliberately deprioritize, and what it costs us to walk away - One non-obvious segment or use case the business owner has probably overlooked, with the evidence that made you flag it - The 3 riskiest assumptions in this analysis and the cheapest way to test each 7. IDEAL CUSTOMER PROFILE — for the top segment, written as a one-page profile a marketer could hand to a copywriter: demographics/firmographics, a day in their life relevant to this purchase, the language they use to describe the problem (their words, not industry words), what they've already tried, and what would make them say yes today. === BUSINESS BRIEF === [PASTE YOUR BUSINESS BRIEF HERE] ``` **Follow-up prompt worth running:** `Now find me three data sources or local signals that would contradict your conclusion. What would have to be true for the segment you ranked #4 to actually be the best one?` --- ## MODULE 2 — Competitor, Pricing & Marketing Intelligence *Goal: know exactly what you're up against, what they charge, how they get customers, and where the opening is.* > **This is the module most likely to produce confident-sounding fiction.** Competitor pricing, ad spend, and follower counts are mostly *not* publicly available, and an AI asked for a tidy table will fill it in. The rules below are what stop that. Don't remove them to get a prettier document. ``` You are a competitive intelligence analyst. You are rigorous about the difference between what you found and what you inferred, and you are comfortable reporting that something could not be found. TASK: Build a competitive and pricing picture for the business below, then tell me where the opening is. === THE EVIDENCE RULE (this governs everything below) === Every company you name must be a real business you located through search. Give me the URL. The same standard applies to EVERY NUMBER in this report — prices, review counts, ratings, follower counts, headcount, ad spend, ad duration. Format every sourced number like this: $125 tune-up (competitorsite.com/maintenance, observed 2026-09-14) Any item you cannot source that way must read: NOT FOUND — [one-line reason] or, where a source exists but is closed to you: BLOCKED — [source] — [what this means I can't assess] A report with fifteen NOT FOUNDs and twenty sourced facts is a SUCCESS. A report with no NOT FOUNDs is a FAILURE, because this data is not all public. NEVER infer from general industry knowledge: competitor pricing, advertising spend, ad duration, follower counts, engagement rates, employee headcount, revenue, or review sentiment. These are the seven fields most likely to be fabricated. If you don't have the source, say NOT FOUND. === STEP 0: STATE YOUR BOUNDARIES BEFORE YOU RESEARCH === Declare these three things explicitly, then proceed on them: (a) GEOGRAPHY — name the specific cities, counties, or ZIP codes you are treating as in-market. Confirm each competitor has a verified physical address inside it, not just a service-area page claiming coverage. (b) COMPARABILITY — include only competitors serving the same customer at roughly comparable size. Flag separately any much larger, commercial-focused, or multi-line firm, and say why it still matters. (c) SUBSTITUTES — national lead-generation platforms and marketplaces (Angi, Thumbtack, HomeAdvisor, Yelp's paid placement, Amazon, Upwork, whichever apply) compete for the same customers. Treat them as substitutes to analyze, not as sources to cite. If I have not told you which competitor we lose to most, ASK ME before building the battlecard. Do not pick one yourself. === WHERE TO LOOK === TIER 1 — usually accessible, this is where the real findings come from: - Competitor websites: services, published pricing, specials and coupon pages, guarantees, warranties, financing partners, certifications, positioning language, calls to action, how they capture leads - Local "best of" listicles and directory pages (useful for discovery — but verify each name independently rather than reprinting one aggregator's editorial list as research) - Indeed and company careers pages: hiring signals, team size hints, what they're investing in - State licensing and business registration records - Trade association member directories - Google Trends, local news, chamber listings TIER 2 — frequently blocked to automated access. Try. If blocked, say so explicitly and move on rather than guessing: - Google Maps local pack, Google Business Profiles, Yelp, BBB, Angi - LinkedIn (headcount, hiring) - Google Ads Transparency Center, Meta Ad Library - Social profiles and engagement metrics If you cannot reach a Tier 2 source, tell me what I would need to look up manually and how long it would take. That is more useful than an invented number. === DELIVER === 1. OUR OWN BASELINE — do this FIRST, before any competitor work Look up our business. Report: Google Business Profile claimed or not, review count, average rating, date of most recent review, number of photos, whether services, hours, and service area are filled in, whether we respond to reviews, and what our website makes clear within five seconds. Source and date-stamp each. This is the cheapest, highest-return thing in the entire analysis and it gets skipped constantly. 2. COMPETITIVE SET — 6-10 real competitors, each with URL and verified local address - DIRECT: same offer, same customer, same geography - INDIRECT: different offer, same customer need - SUBSTITUTE: how customers solve this without buying from anyone — DIY, doing nothing, in-house, or a lead-gen marketplace. Usually the largest competitor and almost always overlooked. - EMERGING: newer or tech-enabled entrants changing expectations. If you find no evidence of a specific emerging entrant, write NOT FOUND. Do not invent one to fill the category. 3. COMPETITOR PROFILE TABLE Company | URL | Verified address | Positioning (their exact words) | Target segment | Published pricing | Guarantees / warranties | Certifications | Primary visible channels | Reviews (count / rating / source / date observed) | Strengths | Exploitable weaknesses Use NOT FOUND liberally. It is the honest answer for most of these cells. 4. PRICING — report what is published, not what you assume Expect most prices to be unpublished. In many industries, companies deliberately don't post them. (a) Which competitors publish ANY price, and which say "call for a quote" (b) What they DO publish: service plan rates, trip or diagnostic fees, discount percentages, minimums, financing partners, "starting at" figures — with URLs (c) How many published nothing at all (d) DO NOT issue a pricing recommendation unless you found published prices for at least 3 competitors for a comparable offer. If you didn't, say so, and instead tell me exactly what's missing and how to get it — mystery-shop calls, requesting quotes, booking an entry-level service. (e) THE OPACITY OPPORTUNITY: if nobody publishes prices, publishing yours is itself a differentiator and an AI-search advantage. Assess whether that fits us. (f) PRICING MODELS visibly in use: hourly, flat-rate, tiered, subscription, value-based, per-unit — and which appears to be gaining ground, with evidence. 5. HOW THEY ACTUALLY MARKET — for each significant competitor Report only what you can verify, marking each BLOCKED or NOT FOUND where you can't: - Website: clarity of offer, conversion mechanics, mobile experience, speed - Organic search: topics and content volume, whether they invest in SEO at all - AI search visibility: run the queries a real customer would ask an AI assistant — "best [service] in [city]", "how much does [service] cost in [city]", "who should I hire for [problem]". Report which businesses get named and what content appears to earn the mention. Note the mechanism: who publishes the answer-shaped content, the comparison pages, the pricing transparency. - Paid advertising: attempt the ad libraries. If blocked, say BLOCKED and instead look for indirect evidence — landing pages built for ads, tracking parameters in their links, coupon codes, "as seen on" claims. - Social: platforms they're on. Report follower counts and cadence only if you actually observed them. - Email and content: lead magnets, newsletter signup, gated content, webinars - Reviews: volume, rating, whether they respond, and any visible review-request mechanism - Local and community: sponsorships, events, associations, partnerships, PR - Referral and partner programs, if visible 6. SYNTHESIS - THE SEA OF SAMENESS: the claims every competitor makes, quoted verbatim. These are the words we must not use. - THE UNCLAIMED GROUND: credible positioning nobody has taken, and why it's open - POSITIONING MAP: two axes that genuinely drive choice here — not generic "price vs quality" unless that's truly it. Plot everyone including us. Name the empty space and judge whether it's empty because it's valuable or because it's worthless. - COMPETITIVE ADVANTAGE AUDIT: our claimed advantages, tested honestly. Is each real, defensible, and something a customer would pay for? Kill the ones that fail. - THE GAP THAT MATTERS MOST: given our baseline in section 1 versus their numbers, name the single largest measurable gap and what closing it would take. - WHAT TO COPY: 3 things they do well we should adopt this quarter - WHAT TO AVOID: 3 things that appear to be producing nothing 7. RECOMMENDATIONS, COSTED Every recommendation must be priced and must fit inside our stated monthly marketing budget. Show the allocation and the total. Rank by expected return per dollar given our average transaction value. Explicitly name what we should NOT do this year because the budget won't reach it. 8. BATTLECARD — for the competitor we lose to most (ask me if I haven't said): how they'll be positioned against us, the three things they'll say, our rebuttal to each, and the one question we can ask a prospect that exposes their weakness. === FINAL STEP === Before you finish, list every number in this report you could NOT source, and confirm you marked it NOT FOUND rather than estimating it. === BUSINESS BRIEF === [PASTE YOUR BUSINESS BRIEF HERE] ``` **Follow-up prompt worth running:** `Which of your findings came from a single source? For each, tell me what an independent second source would be and whether you checked it.` --- ## MODULE 3 — Channel Strategy & Go-To-Market *Goal: decide where the money and hours go, and be able to defend the decision.* ``` You are a go-to-market strategist who has allocated marketing budgets from $500/month to $500,000/month. You are ruthless about matching channel choice to budget, timeline, and the operator's actual available hours. You have seen more small businesses fail from doing eight channels badly than from doing two channels well. TASK: Design the channel strategy and budget allocation for the business below, using the market and competitor findings from earlier in this conversation. RESEARCH REQUIREMENTS Search for current cost benchmarks specific to this industry and geography — cost per click, cost per lead, cost per acquisition, typical conversion rates, platform minimums. Cite sources and dates. Where industry-specific data doesn't exist, use the closest analog and say which analog you used. Note that published benchmarks are averages across wildly different operators; treat them as a starting range, not a promise. DELIVER 1. THE BUYING JOURNEY For the primary segment, map the real path from unaware to purchased: - The trigger event that starts the search - Where they look first (be specific — "Google" is not specific; "Google Maps local pack on a phone, at the moment of the problem" is) - Who they ask for a recommendation - What they compare, and on what criteria - The moment of hesitation, and what resolves it - What happens after purchase that drives repeat and referral At each stage: how long it takes, what the customer is feeling, and the single highest-leverage marketing intervention available at that moment. 2. CHANNEL SCORECARD Evaluate every channel below for this specific business. Score, rank, and be opinionated — a scorecard where everything scores 7 is useless. Channels to evaluate: - Google Business Profile and local SEO - Organic search / content / blog - AI search visibility (appearing in ChatGPT, Gemini, Perplexity, Copilot answers) - Google Ads (Search, Performance Max, Local Services Ads) - Meta ads (Facebook / Instagram) - Other paid social (LinkedIn, TikTok, Pinterest, Nextdoor, YouTube) — only those that fit - Email marketing - SMS - Direct mail / EDDM - Referral program (customer) - Strategic partnerships and referral relationships (business) - Review generation - Community presence, events, sponsorships - Trade associations and industry groups - Speaking and workshops - PR and earned media - Video and short-form content - Podcast (guesting is usually better ROI than hosting) - Outdoor, radio, print, vehicle wraps - Marketplaces and directories relevant to this industry For each: Cost to start | Ongoing monthly cost | Realistic cost per lead [cited] | Time to first result | Weekly hours required | Skill required vs. skill we have | How well it reaches our #1 segment | Durability (does it compound or stop when the money stops) | FIT SCORE 1–10 Present as a ranked table. Then write one paragraph on the top 5 explaining the score. 3. THE MIX - PRIMARY (2–3 channels): where the majority of budget and hours go. For each, why it wins for this business, what "doing it properly" requires, and what the first 90 days look like. - SECONDARY (1–2): supporting channels, deliberately smaller. - TEST (1): one cheap experiment with a defined budget cap and a defined kill criterion. - NOT NOW: channels we are explicitly declining, each with the reason. This list matters as much as the first one — it's what stops budget leaking. - Note which channels compound (SEO, email list, reviews, referral relationships) versus which stop the day you stop paying (most ads). A plan that's all rented attention and no owned assets is a treadmill. 4. BUDGET ALLOCATION - Monthly dollars and percentage per channel. MUST sum exactly to the stated budget. Show the arithmetic and state the total. - Reserve 10–15% as an explicit test-and-learn line - Show the allocation shifting across months 1, 2, and 3 as results come in - Flag any channel where the budget is below the minimum viable spend — running a channel underfunded is worse than not running it. Say so plainly. - One-time setup costs, listed separately from ongoing spend 5. MESSAGING AND OFFER For the top 2 segments: - Core value proposition in one sentence, in the customer's language - Three proof points that make it believable (specific, verifiable, not adjectives) - The primary objection and the exact rebuttal - Five headline options, written plainly — no clichés, no "unlock," no "elevate" - The entry offer: what gets a cautious first-time buyer to take one small step - Risk reversal: the guarantee or reassurance that removes the fear of choosing wrong - Channel-specific adaptations: how the message changes on a Google ad vs. a Facebook post vs. an email subject line vs. a yard sign 6. THE FUNNEL MATH Build a model with explicit assumptions: Budget → impressions/reach → clicks or contacts → leads → qualified leads → customers → revenue - State each conversion rate assumption and cite a benchmark where possible - Compute cost per lead and cost per acquisition by channel - Compare CAC against the customer lifetime value from the Brief. Give the LTV:CAC ratio and say whether it works. Below 3:1 is a warning; below 1:1 is a fire. - Time to breakeven on the marketing spend - Run three scenarios: conservative, expected, optimistic. Show what leads to each. - SANITY CHECK: does the expected customer volume exceed our stated capacity ceiling? If so, say so — over-marketing past capacity destroys reputation faster than no marketing at all. 7. INFRASTRUCTURE GAPS What has to exist before these channels can work: tracking and analytics, a way to capture and route leads, a follow-up system, a place to send traffic that converts, review collection, a CRM or at minimum a spreadsheet. List what's missing, what each costs, and what's genuinely required versus nice to have. Recommend specific tools at the budget level stated. === BUSINESS BRIEF === [PASTE YOUR BUSINESS BRIEF HERE] ``` **Follow-up prompt worth running:** `Cut my budget in half. What would you change, and what would you protect at all costs? Now double it — what's the first thing you'd add and why isn't it already in the plan?` --- ## MODULE 4 — The 90-Day Marketing Plan *Goal: turn strategy into a calendar with names, numbers, and dates on it.* ``` You are a marketing operations lead who builds execution plans for small teams. Your plans get finished, because you size the work to the hours that actually exist and you sequence it so early wins fund the patience for the slow stuff. TASK: Build a 90-day execution plan from the strategy developed earlier in this conversation. NON-NEGOTIABLE CONSTRAINTS - Total weekly hours across all tasks must not exceed the hours stated in the Brief. Add up the hours in every week and show the total. If a week exceeds the limit, cut something and say what you cut. - Total spend must not exceed the monthly budget. Show monthly totals. - Every task must name a specific deliverable, not an activity. "Improve SEO" is an activity. "Publish the 'cost of [service] in [city]' page with pricing table and FAQ schema" is a deliverable. - Front-load visible wins. Motivation is the scarcest resource in a small business marketing plan; a plan with nothing to show for the first six weeks dies in week four. DELIVER 1. THE 90-DAY OBJECTIVE One sentence, one number, one date. Then the 3–4 key results that roll up to it. 2. PHASE STRUCTURE DAYS 1–30: FOUNDATION Objective, and why this comes first. Fix what's broken, put measurement in place, claim the free assets, and ship the quick wins. DAYS 31–60: ACTIVATION Objective. Turn on the primary channels properly, start the content and outreach rhythm, launch the offer. DAYS 61–90: SCALE AND SYSTEMATIZE Objective. Double down on what's working, cut what isn't, and convert one-off efforts into repeatable systems. For each phase give me: objective, 5–8 tasks, and for each task — Deliverable | Owner | Est. hours | Cost | Dependency | Success metric with a number 3. WEEK-BY-WEEK TABLE — all 13 weeks Columns: Week | Focus | Tasks (specific) | Hours | Spend | Deliverable due | Metric checked Include the weekly hour total in every row so the load is visible at a glance. 4. QUICK WINS — WEEK ONE Everything that can be done in the first seven days for under $100 total and produces visible movement. Typically includes: claiming and fully completing the Google Business Profile, asking the last 20 happy customers for reviews, fixing the phone number and hours everywhere they appear, adding a clear call to action above the fold, setting up call and form tracking, and emailing the existing list with a real offer. Tailor these to this specific business — don't just repeat the generic list. 5. ASSET CHECKLIST Every asset required across 90 days: landing pages, ad creative, email sequences, photos, video, lead magnets, review request templates, sales follow-up scripts, partnership outreach templates. For each: what it is, who makes it, when it's needed, estimated cost or time, and whether AI can draft it. 6. MEASUREMENT SYSTEM - The 5–7 metrics that matter for THIS plan. Resist vanity metrics; if a number wouldn't change a decision, leave it out. - For each: definition, where it's tracked, how to set it up, current baseline (or "unknown — establish in week 1"), 30/60/90-day targets - A one-page WEEKLY SCORECARD layout: the handful of numbers reviewed every Monday in fifteen minutes - The monthly review agenda: what gets examined, what decisions get made 7. DECISION GATES DAY 30 GATE: the specific numbers that mean CONTINUE, ADJUST, or STOP — for each primary channel. Say what "adjust" concretely means in each case. DAY 60 GATE: same, with the added question of where to reallocate budget. DAY 90 REVIEW: what a successful quarter looks like in numbers, and the three questions that set up the next quarter. 8. RISK REGISTER The 5 most likely failure modes for this specific plan. For each: the early warning sign, the probability, the impact, and the mitigation. Include the ones nobody likes to name — the owner runs out of time, the leads come in but nobody follows up fast enough, the phone rings and nobody answers it, capacity gets overwhelmed, a key person leaves. 9. WHAT HAPPENS ON DAY 91 One paragraph on where this plan should lead, and the one thing to start building in quarter one that pays off in quarter three. FORMAT: tables for everything schedulable. This document should be printable and pinnable to a wall. === BUSINESS BRIEF === [PASTE YOUR BUSINESS BRIEF HERE] ``` **Follow-up prompt worth running:** `I only have half the hours I said I did. Rebuild the plan for that reality — what survives, what gets cut, and what do I lose by cutting it?` --- ## MODULE 5 — The AI Execution Layer *Goal: figure out which marketing work AI should actually do for you, then hand you the prompts that do it.* ### Part A — The suitability audit ``` You are an AI implementation advisor for small businesses. You are candid about where AI helps enormously, where it produces mediocre output that quietly damages a brand, and where using it is actively a mistake. TASK: Using the 90-day plan built earlier in this conversation, tell me exactly which marketing tasks to hand to AI, which to co-write, and which to keep entirely human. DELIVER 1. TASK-BY-TASK AI SUITABILITY MAP Take every recurring task in the plan and classify it: AI DOES IT — near-autonomous, light human review (e.g., repurposing one asset into ten formats, first-draft research summaries, metadata and alt text, transcription, data cleanup, reporting summaries, first-pass keyword and competitor scans) AI DRAFTS IT — AI produces, human edits and approves before it ships (e.g., ad copy variants, email sequences, blog drafts, social captions, landing page copy, FAQ content, review responses, proposal boilerplate) AI ASSISTS — human leads, AI accelerates (e.g., strategy development, offer design, pricing decisions, brand voice, analysis of what results mean) HUMAN ONLY — do not automate (e.g., relationship building, partnership conversations, sales calls, apologies, anything requiring accountability, the actual opinions in thought leadership, customer conversations that build trust) For each task give: the classification, the reason, the time saved, and the specific risk if AI is used badly here. 2. THE HONEST WARNINGS - Which tasks in this plan would be actively damaged by obvious AI output, and why - The tells that make AI content recognizable, and how to avoid them in this industry - Where a customer would lose trust if they knew AI wrote it - What must always be fact-checked by a human before it goes out (claims, prices, dates, guarantees, anything regulated, anything about a named person) 3. TIME AND MONEY - Estimated hours saved per week, by task - What those recovered hours should be redeployed into — the human work that actually grows this business - Recommended AI tool stack at this budget level, with monthly cost. Be specific about what's genuinely needed versus what's marketing hype. 4. THE WEEKLY AI RHYTHM A simple recurring workflow: - MONDAY (30 min): what to run to set up the week - MIDWEEK (30 min): what to run to produce and repurpose - FRIDAY (30 min): what to run to review and learn Name the specific prompt used at each point. 5. GUARDRAILS A short checklist the business owner runs before any AI-assisted content goes public. Five to seven items, specific to this business and industry. === BUSINESS BRIEF + 90-DAY PLAN === [PASTE YOUR BRIEF AND THE PLAN FROM MODULE 4] ``` ### Part B — Generate a custom prompt library for this business ``` Now write me a personalized prompt library for the recurring marketing tasks in this 90-day plan. For each of the 10–15 most frequent tasks, produce a ready-to-use prompt that: - Opens with a role and relevant expertise - Embeds this business's actual positioning, voice, offers, and target segment so I don't have to re-explain the business every time - Uses [BRACKETED PLACEHOLDERS] only for what genuinely changes each run - Specifies the exact output format and length - Includes the constraints that keep output on-brand (words to avoid, claims we can't make, required disclaimers, tone) - Ends with a self-check instruction - Is followed by a one-line note: "Before publishing, verify: ___" Format each as a copy-paste block with a clear title. Assume the person using it is busy and non-technical. ``` ### Part C — The starter library (use these today) Replace the bracketed parts. These work as-is in any AI tool. --- **1. Google Business Profile post** ``` You write Google Business Profile posts for a [TYPE OF BUSINESS] serving [CITY/AREA]. Voice: [plain-spoken / warm / expert / no-nonsense]. We serve [TARGET CUSTOMER]. Write 4 GBP posts about [TOPIC / SERVICE / OFFER / SEASONAL ANGLE]. Each post: 150–250 words, opens with the customer's problem in their own words, one specific and concrete detail that proves we know this work, and a clear call to action. Naturally include the phrase "[SERVICE] in [CITY]" once — once, not repeatedly. No exclamation points. No "elevate," "unlock," "game-changer," or "in today's world." Suggest a photo to pair with each. Then rate your own posts 1-10 on whether they sound like a real local business owner wrote them, and rewrite any that score below 8. Before publishing, verify: prices, dates, and any guarantee mentioned. ``` **2. Local service page that ranks (and gets cited by AI search)** ``` You are an SEO content writer who specializes in local service businesses. Write a service page for [BUSINESS] targeting the search "[PRIMARY KEYWORD]" in [CITY/AREA]. Related searches to cover: [LIST 3-5]. Structure: - H1 with the primary keyword, written for a human - Opening paragraph that answers the searcher's actual question in the first two sentences — AI search engines and impatient humans both reward this - What's included / how it works, with specifics - A real pricing section: an honest range with what drives cost up or down. Most competitors hide this; publishing it wins trust and wins AI citations. - Why choose us: three proof points, specific and verifiable, no adjective stacking - FAQ: 6–8 questions written the way customers actually ask them, each answered in 40–60 words so the answer can be lifted directly into an AI-generated response - Clear next step 1,200–1,500 words. Eighth-grade reading level. Write for someone deciding whether to call, not for a search engine. End with: suggested title tag (under 60 characters), meta description (under 155), and the FAQ questions formatted for FAQ schema. Before publishing, verify: every price, timeline, and credential claim. ``` **3. Google Ads search campaign** ``` You are a Google Ads specialist for local service businesses. Build a search campaign for [BUSINESS], [SERVICE], targeting [GEOGRAPHY]. Monthly budget: $[AMOUNT]. Target customer: [SEGMENT]. Our differentiator: [WHAT]. Deliver: 1. Ad group structure — 3-5 tightly themed groups with the logic explained 2. Keywords per group: exact and phrase match, with estimated CPC ranges for this industry and geography (search for current data and cite it) 3. Negative keyword list — at least 30, focused on the terms that waste money for this specific business (job seekers, DIY, free, cheap, wrong services, wrong locations) 4. 15 responsive search ad headlines (30 char max) and 4 descriptions (90 char max) per ad group. Include the benefit, the differentiator, proof, and the call to action across the set. No superlatives we can't substantiate. 5. Sitelink, callout, and structured snippet extensions 6. Recommended bid strategy for this budget, and why 7. What the landing page must contain for these ads to convert 8. The first three things to check after 14 days, with the number that would mean pause Before publishing, verify: character counts, claims, and that the landing page matches the ad promise. ``` **4. Meta ad concepts** ``` You are a direct-response social advertiser. Create 5 Facebook/Instagram ad concepts for [BUSINESS] promoting [OFFER] to [SEGMENT] in [GEOGRAPHY]. Budget: $[AMOUNT]/month. Each concept needs a different angle — vary across: the problem-agitation angle, the before/after proof angle, the myth-correction angle, the local-community angle, and the direct-offer angle. For each: the angle and why it works on this audience, primary text (125 characters before the fold does the work), headline, description, a specific creative direction (what's literally in the photo or video — describe the shot), the call-to-action button, and what audience targeting to pair it with. Then tell me which one to test first and what result after $[AMOUNT] of spend would mean kill it. Rules: no stock-photo energy, no stacked emojis, no "DM us for details." Write like a person, not a brand. Before publishing, verify: claims, pricing, and Meta's rules for this industry. ``` **5. Email nurture sequence** ``` You are an email copywriter for small businesses. You write emails people actually finish reading. Write a [NUMBER]-email sequence for [BUSINESS] that takes someone from [ENTRY POINT — e.g. downloaded our guide / requested a quote / bought once] to [GOAL]. For each email: send timing, subject line (plus one alternative), preview text, full body copy, the single call to action, and the one job that email is doing. Constraints: - 150–300 words each. Nobody reads more. - One idea, one ask per email - Written to one person, in second person, from a named human — not from "the team" - Lead with something useful before asking for anything - No countdown-timer urgency unless the deadline is real Voice: [DESCRIBE]. Our differentiator: [WHAT]. The objection to overcome: [WHAT]. Then write the sequence map: what triggers it, what removes someone from it, and what happens to people who don't convert at the end. Before publishing, verify: the offer terms, any deadline, and unsubscribe compliance. ``` **6. Monthly content calendar** ``` You are a content strategist for a small business with limited time. Build a 30-day content calendar for [BUSINESS] targeting [SEGMENT] on [PLATFORMS]. We can realistically produce [NUMBER] original pieces per week and we have [SKILLS]. Requirements: - Anchor on [NUMBER] substantial pieces per month; everything else is repurposed from those. Never create from scratch what you can cut down from something bigger. - Mix: roughly 40% educational, 30% proof (work, results, customers), 20% personality and behind-the-scenes, 10% direct offer - Tie to the seasonality identified in the market research - For each item: date, platform, format, hook or headline, the core point, the call to action, the asset needed, and estimated production time - Include a weekly total production time so I can see whether this is realistic Deliver as a table, then add: the batch production plan (how to make a month of content in one sitting) and the 5 evergreen pieces worth making once and reusing all year. ``` **7. Turn one thing into ten** ``` Here is one piece of source material: [PASTE ARTICLE / TRANSCRIPT / NOTES / CASE STUDY]. Repurpose it for [BUSINESS], voice [DESCRIBE], audience [SEGMENT]: - 1 LinkedIn post (under 200 words, strong first line, no hashtag spam) - 3 Instagram/Facebook captions with image or carousel directions - 1 email to our list - 5 short-form video hooks with a 30-second outline each - 1 Google Business Profile post - 3 FAQ entries for the website - 1 short answer (40-60 words) formatted so an AI assistant could quote it directly when someone asks about this topic Keep the specific details and numbers from the source — those are what make it credible. Do not pad with generalities to hit length. ``` **8. Review requests and responses** ``` PART 1: Write 3 review request messages for [BUSINESS] — one text, one email, one to say in person at [MOMENT]. Short, specific, no begging, no incentive offered. Make it easy to say yes and easy to know what to write about. PART 2: Write response templates for: a 5-star review, a 3-star review with a fair criticism, and a 1-star review that is unfair or from a non-customer. Each under 75 words, signed by a named person, never defensive, and written knowing that the response is for the hundred future customers reading it, not for the reviewer. PART 3: Design a simple system to ask every customer at the right moment without anyone having to remember. Name the trigger, the channel, and the owner. ``` **9. Landing page copy** ``` You are a conversion copywriter. Write a landing page for [BUSINESS] where the traffic comes from [SOURCE] and the single goal is [ACTION]. Audience: [SEGMENT]. Their biggest hesitation: [OBJECTION]. Sections: headline and subhead (give 5 headline options), the problem stated in their words, the solution, how it works in 3 steps, proof (specify exactly what proof to gather and where to place it), objection handling, risk reversal, and the call to action repeated at the top, middle, and bottom. Rules: one goal, no navigation menu, the offer visible without scrolling, plain language at an eighth-grade level, and every claim either provable or cut. Then list what to A/B test first and why. Before publishing, verify: every claim, price, and guarantee. ``` **10. Partnership and referral outreach** ``` You are a business development strategist. Identify 10 types of businesses in [GEOGRAPHY] that serve the same customer as [BUSINESS] at a different moment, without competing with us. For each: why the partnership makes sense, what's in it for them (lead with this), and the specific referral mechanism. Then write a first-contact email — under 120 words, no attachment, specific to their business, asking for one short conversation rather than pitching a program. It should read like one local owner writing to another, because it is. Add: a simple referral agreement outline, how to track it, and how to thank them in a way that keeps referrals coming. ``` **11. Mine your customers for messaging** ``` Here are [reviews / interview notes / sales call transcripts / support emails]: [PASTE]. Analyze them as a customer researcher and extract: 1. VOICE OF CUSTOMER: the exact words and phrases they use for the problem, the solution, and the outcome. Quote verbatim — these become our headlines. 2. The buying triggers that appear repeatedly 3. The decision criteria they mention, ranked by frequency 4. Objections and hesitations, with the ones that recur 5. What they praise most — this is our real differentiator, which may not match what we think it is 6. What they complain about, including gently 7. Five headlines and three email subject lines built entirely from their language 8. The gap between how we describe ourselves and how they describe us Flag anything you inferred rather than found directly in the source. ``` **12. Monthly performance review** ``` You are a marketing analyst. Here is last month's data: [PASTE — spend, leads, customers, revenue, and any channel metrics you have]. Here were the targets: [PASTE]. Deliver: 1. What happened, in plain language, in five sentences 2. Performance versus target by channel, in a table, with variance 3. What's working — and the evidence it's causal, not coincidence 4. What's not working, and whether it needs more time, a fix, or a funeral 5. Cost per lead and cost per acquisition by channel, trended against prior months 6. The single highest-leverage change for next month, with expected impact 7. Budget reallocation recommendation, with the math 8. The three numbers to watch most closely next month 9. What this data can't tell us, and what we'd need to track to answer it Be direct about failure. A flattering report that hides a dying channel costs me money. ``` **13. Lead follow-up sequence** ``` Write the follow-up system for [BUSINESS] for a lead that comes in via [SOURCE] and doesn't convert immediately. Cover the first 5 minutes, the first hour, day 1, day 3, day 7, day 14, day 30, and the long-term nurture. For each: channel, message, and the job it's doing. Include: the exact words for the first phone call opening, a voicemail script under 20 seconds, a text message, and the "I'm going to close your file" email that reliably gets replies. Note the speed-to-lead research — response within five minutes dramatically outperforms response within an hour — and design the system so a small team can actually hit it. ``` **14. Get found by AI assistants** ``` You are an AI search visibility strategist. Buyers increasingly ask ChatGPT, Gemini, Perplexity, and Copilot for recommendations instead of searching. Assess and improve [BUSINESS]'s visibility in those answers for [GEOGRAPHY] and [SERVICE]. 1. List the 15 questions a potential customer would actually ask an AI assistant that should surface us 2. For each, assess whether our current content could be cited as the answer 3. Identify the content gaps — specifically, the direct questions we don't answer anywhere 4. Write 10 question-and-answer pairs, each answer 40–60 words, factual, specific, and self-contained enough to be quoted without context 5. Recommend the structural changes that help: clear question headings, schema markup, published pricing, named locations and service areas, a genuine About page with real people, and consistent business information across the web 6. List the third-party sources AI assistants tend to draw on for this industry (directories, review sites, association listings, local publications) and where we should make sure we appear Be clear about what is established practice versus what is still emerging. ``` --- ## The Reality Check Pass Run this after any module. It is the highest-value 60 seconds in this entire document. ``` Stop. Before I act on any of this, audit your own work. 1. SOURCES: List every factual claim you made that you did NOT verify with a source. For each, tell me how confident you are and what it would cost me to be wrong. 2. FABRICATION CHECK: Did you name any company, statistic, price, or study you did not actually find? Say so now. I would rather have a shorter document than a wrong one. 3. GENERIC CHECK: Which parts of this plan would be identical if I were a different business in a different industry? Those parts are filler. Rewrite them to be specific to me, or delete them. 4. CONSTRAINT CHECK: Recalculate. Does the budget sum to what I gave you? Do the weekly hours fit what I said I have? Show the arithmetic. If it doesn't fit, fix it. 5. THE CFO TEST: Argue against this plan. Where is it most likely to fail? What's the load-bearing assumption that, if wrong, takes the whole thing down? 6. THE CHEAPEST TEST: What is the single fastest, cheapest experiment that would tell me whether the core strategy is right — before I spend the full budget? 7. WHAT I MISSED: What should I have told you that I didn't? What question should I have asked you that I didn't? Then give me a revised version incorporating your own critique. ``` --- ## Common failure modes (and the fix) | What goes wrong | Why | The fix | |---|---|---| | Output is generic and could apply to any business | Brief was thin | Go back to Module 0. Specificity in, specificity out. | | Competitors don't exist | No web search enabled, or model filled a gap from memory | Turn on research mode. Demand URLs. Run the fabrication check. | | Recommends a budget you don't have | Constraints buried or omitted | Restate the budget and hours explicitly at the top, and demand the arithmetic | | Reads like a textbook | No role, no format specified | Use the role framing; demand tables and named deliverables | | Plan looks great, never gets done | Sized for a team you don't have | Run the "half the hours" follow-up. Cut to what survives. | | Numbers don't add up | Models drift on arithmetic | "Show the math and state the total" — every time | | Confident-sounding statistics that are wrong | The most dangerous failure | The `[VERIFIED]/[ESTIMATE]/[ASSUMPTION]/NOT FOUND` tagging rule, plus spot-checking two or three numbers yourself | | Competitor pricing table is fully populated | Most competitors don't publish prices — the AI filled the gaps | Demand `NOT FOUND` cells. A pricing table with no gaps is usually fiction. | | Named competitors don't exist, or aren't local | No search enabled, or a service-area page mistaken for a local address | Require a URL and a verified physical address for every competitor | --- ## Tool notes | Tool | Notes | |---|---| | **ChatGPT** | Turn on search. For the deep path, use a reasoning model and its research mode for Modules 1–2 — it will take several minutes and produce a properly sourced document. Use Projects to keep the Brief attached across modules. | | **Claude** | Strongest on long, structured documents and following multi-part format instructions. Enable web search. Use Projects to store the Brief so it applies to every conversation. | | **Gemini** | Deep Research mode suits Modules 1 and 2 well. Good integration if you're already in Google Workspace. | | **Perplexity** | Best for Module 2. It cites by default, which makes competitor and pricing verification fast. Weaker on long planning documents — use it for research, take the findings elsewhere for the plan. | | **Copilot** | Fine, and convenient if your business runs on Microsoft. Be more insistent about format and length. | **Universal:** whatever you use, do the deep path in one continuous conversation so each module can build on the last. Starting fresh for each module throws away the context that makes the later modules good. --- ## The one-page version If you remember nothing else from this session: 1. **Write the brief honestly** — especially the budget, the hours, and the capacity ceiling. 2. **Turn on web search.** Otherwise you're getting a plausible-sounding memory, not research. 3. **Make it label its confidence.** `[VERIFIED] / [ESTIMATE] / [ASSUMPTION]`. 4. **Demand the math.** Budgets must sum. Hours must fit. 5. **Force a choice.** A plan that recommends nine channels has recommended nothing. 6. **Run the Reality Check.** Let it critique itself before reality does. 7. **Ship the week-one quick wins before anything else.** Momentum is the fuel. --- *Built for [YOUR NAME / COMPANY]. Adapt freely — that's what it's for.*