TestPlan Scribe
Turns any e-commerce product page URL into a client-ready A/B test plan with prioritized hypotheses, variant copy, and sample-size math for small growth agencies and in-house marketers.
One product, one founder, thirty days. Everything below is built to be executed inside the contest window — the official rules still decide what counts.
Decide in 20 seconds
The facts that decide whether this is worth thirty days of yours.
Owner, account manager, or senior growth marketer at a 5–20 person Shopify/e-commerce growth agency, or a solo CRO consultant, in the US/UK/Canada, onboarding a new e-commerce client who expects a conversion audit and experimentation roadmap.
Monetization potential 4.9/10
Tie the price to a measurable money or time outcome and add a recurring or repeat-purchase path instead of a single low-value charge.
Why this could work in 30 days: one founder can ship the MVP in 5 days; the plan targets a first sale on day 12; 11 sales at $99 clears $1,000 of collected revenue; strongest signal is speed to first revenue at 8.4/10.
Score and projections come from the generated report and are not guarantees. The $100K Hackathon is run by Ship or Die — read the official rules before you build.
Your first 24 hours
16 concrete actions. Tick them off as you go — nothing is saved or sent.
Product opportunity and positioning
Target customer
Owner, account manager, or senior growth marketer at a 5–20 person Shopify/e-commerce growth agency, or a solo CRO consultant, in the US/UK/Canada, onboarding a new e-commerce client who expects a conversion audit and experimentation roadmap.
Why now
The artifact is sold to agencies with an immediate client deadline, approved at $99 without procurement, needs no integration or data access, and can be produced from a single public URL.
Problem
The buyer must deliver a credible conversion audit and A/B test backlog fast, but reviewing pages, inventing test ideas, calculating sample sizes, and formatting a client-ready report takes hours per client and looks interchangeable with every other agency's deck.
Solution
The user submits a product page URL, brand name, baseline conversion rate, and traffic estimate; the product captures the page, scores conversion heuristics, generates 12 prioritized hypotheses with variant copy, computes sample size and duration, and exports a branded PDF test plan plus a CSV backlog.
1. Product opportunity and positioning
TestPlan Scribe sells one artifact: a branded, client-ready A/B test plan for a single e-commerce product page, delivered as PDF plus CSV within ten minutes of payment. The buyer is not a large experimentation team. It is the owner, account manager, or senior growth marketer at a 5–20 person Shopify growth agency, or a solo CRO consultant in the US/UK/Canada, who has just won or onboarded a client and needs something credible to present before the kickoff call.
The pain is concrete and deadline-driven. To look competent in that meeting, the buyer has to review the page, invent test ideas, guess at sample sizes, and format a deck. That is roughly half a day per client, and the result still looks like every other agency's audit. TestPlan Scribe collapses that into a four-field form: URL, brand name, baseline conversion rate, traffic estimate. What comes back is page-specific: captured headline, CTAs, forms, offer, product-image alt text, and above-fold screenshot; a heuristic score across clarity, offer strength, friction, trust, and mobile risk; twelve prioritized hypotheses each with control, variant copy, primary metric, and a labeled expected-lift range; then sample size and test duration, a branded PDF, and a CSV backlog ready for Notion, Sheets, or Asana.
Positioning is deliberately narrow. It is not an experimentation platform, not live A/B test execution, not a statistics engine inside the buyer's site, and not consulting. Optimizely's original problem — teams struggle to run rigorous experimentation — is real, but the enterprise-platform answer is unavailable to a nine-person agency. The wedge is the pre-test planning artifact: the thing agencies already sell and currently fake in spreadsheets.
Why the wedge is sellable in 30 days: the buyer is directly reachable by LinkedIn message and email, approves $99 without procurement, needs no integration, and gets value from one public URL with no data connection. The JEV score (63.5/100, band C) confirms this shape — speed to first revenue is the strongest criterion at 8.41, solo build feasibility is 7.58, while monetization potential (4.88) and defensibility (4.99) are the weak points. This plan answers both without broadening the product: monetization via repeat purchase per client page, defensibility via page-specific evidence and a fixed named output format that generic assistants do not produce.
Claim discipline is part of the positioning. Every expected-lift figure is presented as a hypothesis with stated assumptions, never as a promise. The report says what to test and how long it must run; it does not guarantee outcomes.
2. Smallest sellable MVP
The smallest sellable unit is: pay $99, submit one URL plus brand and baseline rate, receive a branded 8–12 page PDF test plan and a CSV backlog by email within ten minutes, with one automated re-run if the page failed to capture. Everything else — accounts, dashboards, seats, integrations, historical archives — is absent by design.
Eight must-have features ship by day 5: URL intake form; headless capture with Firecrawl plus screenshot; rule-based heuristic scoring; AI hypothesis generation of exactly 12 structured test cards; sample-size and duration calculator; branded PDF with agency logo and client name; CSV export; Stripe Checkout with automatic email delivery and a usage ledger capping API cost per order.
Explicitly out of scope for 30 days: live test execution or traffic allocation, Shopify/GA4/Klaviyo/CMS integrations, team accounts and client portals, SSO or security review packages, done-for-you consulting, and any promise of measured lift.
3. Technical implementation plan
Next.js on Vercel, Firecrawl for capture, OpenAI for hypothesis and copy generation, React-PDF for export, Stripe Checkout for payment, Resend for delivery, Supabase for the order and usage ledger. Order flow: buyer pays via Checkout; a Stripe webhook writes the order, marks it paid, and enqueues generation; the worker captures the page, caches the extracted content, runs the heuristic scorer, calls the model with a strict JSON schema for 12 cards, computes sample size from baseline rate, minimum detectable effect (default 15% relative), and daily traffic, then renders PDF and CSV, stores both in object storage, and emails signed links via Resend. If capture is blocked by bot protection or renders blank, the order is flagged and the buyer is offered an alternative URL or a refund before generation cost is spent.
Build order: intake and capture first (day 2), scoring and generation (day 3), math and export (day 4), payment, delivery, and ledger (day 5). The single hardest technical risk is capture reliability on bot-protected or JavaScript-heavy storefronts; the mitigation is a fallback to Firecrawl's rendered mode, a manual URL retry, and a stored screenshot the buyer can replace.
4. Pricing and monetization
One price, one product: $99 one-time per client test plan. No subscription, no seats, no tiers. That is deliberate — an agency owner can expense $99 against a client engagement without approval, and one-time pricing removes the objection of another recurring tool.
Repeat purchase is the growth path, not a new price. The average agency onboards several e-commerce clients per quarter; each new client page is another $99 order with zero additional acquisition cost. The offer sent after delivery is a second page at the same price, not a discount. Refunds are limited by showing a watermarked sample before purchase.
5. First-customer acquisition
One channel: direct LinkedIn and email outreach to Shopify/e-commerce growth agency owners and solo CRO consultants. Build a 300-contact list from LinkedIn searches ("Shopify growth agency", "CRO consultant", "e-commerce conversion"), public agency directories, and Shopify Partner listings. The first message offers a free watermarked test plan for one client page; the first reply asks for feedback, not money. Paid conversion is offered only after the sample lands.
Daily cadence from day 7: 40–60 new contacts, a 48-hour follow-up with the sample attached, and one community post per week in founder/agency spaces where operators ask for CRO feedback, linking only to a free sample request form. The week-one goal is 100 contacts and 5 free samples delivered.
6. 30-day roadmap
Days 1–5 build: accounts and Stripe/TrustMRR connection, capture pipeline, scorer and hypothesis generator, math and PDF/CSV export, then checkout and delivery. Day 6 produces three watermarked samples from public brand pages and a 60-second demo. Days 7–14 run the first 150 outreach contacts, collect feedback, harden report quality, and push for the first paid order around day 12. Days 15–21 run a second wave of 120 contacts, publish two public teardowns, and convert free-sample recipients. Days 22–30 run a third wave, ask every buyer for a second page, fix the top complaint, and close the pipeline with a final-call message before the December 16 cutoff.
7. Revenue and profit scenarios (projections only)
All figures are projections from stated assumptions, not verified revenue. Conservative: 3 customers at $99, $297 gross, near-zero refunds, no ad spend, hackathon profit roughly $285. Base: 10 customers, $990 gross, $39 fees. Upside: 30 customers, $2,970 gross, $60 refunds, $118 fees, $150 paid spend — the only scenario with marketing spend, used to test one sponsored post. Operating costs stay between $25 and $110 (hosting, domain, email, Firecrawl, OpenAI) and are excluded from hackathon profit but included in business profit.
8. Risks and validation experiments
Cold outreach response is the biggest risk; the test is 50 messages in 48 hours with a free-sample offer and a kill threshold of fewer than 3 replies. Generic-feeling output is second; the test is showing one sample to five agency owners and asking whether the hypotheses are page-specific enough to present. Blocked page capture is third; the test is running ten varied storefronts through capture and counting failures. Willingness to pay $99 for a software artifact is fourth; the test is asking three sample recipients to pay, not to praise.
9. First 24 hours
Hours 1–2: create the Vercel project, Supabase instance, Stripe account, and connect TrustMRR. Hours 3–5: build the intake form and Firecrawl capture with screenshot storage. Hours 6–8: build the list of the first 60 agencies with names and profile URLs. Hours 9–12: ship a raw text output of captured page elements to prove the pipeline end to end. Output: a working capture skeleton and a 60-row contact list before the first build day ends.
Smallest sellable MVP
Pay $99, submit one product page URL plus brand name and baseline conversion rate, receive a branded 8–12 page PDF test plan with 12 prioritized hypotheses and a CSV backlog by email within ten minutes.
Must-have features
- URL intake form with brand name, page type, baseline conversion rate, and traffic estimate
- Headless page capture (Firecrawl) extracting headline, CTAs, forms, offer, image alt text, and above-fold screenshot
- Rule-based conversion heuristic scorer across clarity, offer strength, friction, trust, and mobile risk
- AI generator producing exactly 12 structured test cards with hypothesis, control, variant copy, primary metric, and lift range
- Sample-size and duration calculator from baseline rate, minimum detectable effect, and daily traffic
- Branded PDF export with agency logo, client name, date, and prioritization table
- CSV backlog export for Notion, Sheets, or Asana
- Stripe Checkout with automatic email delivery and a usage ledger capping API cost per order
Explicitly out of scope
- No live A/B test execution, traffic allocation, or in-site statistics engine
- No Shopify, GA4, Klaviyo, or CMS integrations
- No team accounts, seats, comments, or client portals
- No enterprise SSO, procurement workflow, or security review package
- No human CRO consulting or done-for-you service promise
- No subscription billing or seat-based pricing
Demo flow (what the buyer sees)
- Buyer opens the landing page and clicks the sample PDF
- Buyer pays $99 through Stripe Checkout
- Buyer receives the intake form and submits URL, brand name, baseline conversion rate, and traffic estimate
- System captures the page, scores heuristics, and generates 12 test cards
- System computes sample size and duration per test
- Buyer receives an email with signed links to the branded PDF and CSV backlog
- Buyer forwards the PDF to the client or imports the CSV into Notion, Sheets, or Asana
Technical implementation plan
Stack
Architecture
A Next.js app serves the landing page, intake form, and Stripe Checkout redirect. A Stripe webhook writes a paid order to Supabase and enqueues a generation job. The worker captures the page through Firecrawl, caches extracted elements and the screenshot, runs the rule-based scorer, calls the model once with a strict JSON schema for 12 test cards, computes sample size and duration locally, renders PDF and CSV, stores both in object storage, and emails signed links through Resend. A usage ledger row per order caps retries and inference spend.
Build steps
- Create Vercel project, Supabase tables (orders, reports, usage), and environment secrets
- Build intake form and validated order creation
- Implement Firecrawl capture with rendered mode, screenshot storage, and content cache
- Write the rule-based heuristic scorer over captured elements
- Implement the OpenAI call with a strict JSON schema producing 12 test cards
- Add sample-size and duration calculator with explicit assumptions block
- Render branded PDF with React-PDF and generate the CSV backlog
- Add Stripe Checkout, webhook, and paid-order gating
- Add Resend delivery with signed links and resend-on-demand
- Add failure handling: capture-blocked flag, one automated re-run, refund path
- Run ten varied storefronts through the pipeline and fix the three worst outputs
- Deploy to production with Stripe live keys and TrustMRR-connected account
AI components
- Hypothesis and variant-copy generation (12 structured test cards) from captured page elements
- Variant copy drafting for headline, CTA, offer, and social proof modules
- Short page-specific evidence sentence per test card to prevent generic output
Pricing and monetization
Single Client Test Plan
Entry offerFor Agency owner, account manager, or solo CRO consultant who needs one client-ready test plan this week.
- 1 URL analysis with captured page evidence
- 12 prioritized A/B test hypotheses with variant copy
- Sample-size and duration recommendations
- Branded PDF report with agency logo and client name
- CSV backlog export for Notion, Sheets, or Asana
- One automated re-run if the page failed to capture
Model: One-time $99 per client test plan, no subscription, no seats, no tiers.
First revenue offer: I built a tool that turns one product page URL into a branded 10-page CRO test plan in about 3 minutes. Send me a client page and I'll run it free so you can see the output before paying anything.
Expected ACV: $99
Upsell path: Repeat purchase is the growth path: every new client page is another $99 order. After delivery, the buyer receives a one-line email asking for their next client URL at the same price; buyers with multiple onboardings become multi-order accounts with no added acquisition cost.
First-customer acquisition
Direct LinkedIn and email outreach to Shopify/e-commerce growth agency owners and solo CRO consultants, leading with a free watermarked sample test plan.
Where to find them
- LinkedIn people search: 'Shopify growth agency' owner/founder, 5–50 employees
- LinkedIn people search: 'CRO consultant' and 'conversion optimization consultant', US/UK/Canada
- Shopify Partner and Expert directory listings for small e-commerce agencies
- Clutch and DesignRush agency listings filtered to conversion optimization
- Twitter/X and LinkedIn posts asking for e-commerce CRO audits
- Founder and agency communities where e-commerce operators request CRO feedback
- Public e-commerce brand product pages as material for demo samples
Exact outreach motions
- Send 40 LinkedIn connection requests per day with a one-line note naming their client type, not a pitch
- Email 25 agency owners per day from the built list using the free-sample message and a plain-text signature
- Attach nothing on first contact; generate the free watermarked sample only after a reply
- Follow up once at 48 hours with the sample PDF attached and a single-line paid offer
- Post one watermarked sample per week in two communities where operators ask for CRO feedback, linking only to a free sample request form
- Ask every paying buyer for their next client URL three days after delivery
Message template
Hi [Name] — you run CRO for Shopify brands, so you probably rebuild conversion audits by hand. I built a tool that turns one product page URL into a branded 10-page test plan: 12 prioritized hypotheses, variant copy, and sample-size math. Want me to run it free on one client page so you can judge the output? No charge, no call.
30-day execution roadmap
Week 1 · Build & validate · 7 days
Accounts and infrastructure
- Register domain and create Vercel and Supabase projects
- Create Stripe account and connect TrustMRR before any payment
- Set up Resend, Firecrawl, and OpenAI keys with spend caps
- Start the agency contact list and log the first 20 rows
Deliverable: Live skeleton deployment, payment account connected, 20 contacts logged.
Page capture pipeline
- Build the intake form with URL, brand, baseline rate, and traffic fields
- Implement Firecrawl capture with rendered mode and screenshot storage
- Cache captured elements in Supabase to avoid re-scraping
- Test capture on 8 varied storefronts
Deliverable: Working capture that returns structured page elements and a screenshot.
Scorer and hypotheses
- Write the rule-based heuristic scorer across five checks
- Implement the OpenAI call with a strict 12-card JSON schema
- Add a page-specific evidence sentence to every card
- Review 3 outputs and tighten the prompt
Deliverable: Twelve structured test cards per URL with page-specific evidence.
Math and export
- Add sample-size and duration calculator with an assumptions block
- Render the branded PDF with React-PDF
- Generate the CSV backlog with Notion/Sheets/Asana columns
- Generate one full report end to end
Deliverable: Complete PDF and CSV from a single URL submission.
Payment and delivery
- Wire Stripe Checkout and the paid-order webhook
- Send PDF and CSV links by email through Resend
- Add the usage ledger and per-order cost cap
- Run a full test-mode purchase and delivery
Deliverable: MVP live with a complete paid flow.
Public samples
- Generate three watermarked samples from public brand pages
- Publish the landing page with sample preview
- Record a 60-second screen capture of the flow
- Add refund and delivery-time policy to the page
Deliverable: Landing page with three samples and a demo video.
First contact wave
- Send 25 LinkedIn connection requests with a one-line note
- Send 25 emails using the free-sample message
- Log every send and reply in the contact sheet
- Draft the 48-hour follow-up
Deliverable: 50 contacts reached, pipeline sheet started.
Week 2 · Launch & first sale · 7 days
Volume and first samples
- Send 30 more LinkedIn requests and 20 emails
- Generate and deliver the first free watermarked samples
- Ask three recipients whether the cards feel page-specific
- Log sample generation cost per report
Deliverable: 5 free samples delivered, first qualitative feedback captured.
Fix weak outputs
- Rerun the three weakest samples and compare
- Tighten variant copy length and lift-range framing
- Add capture-failure detection and retry path
- Update the sample on the landing page
Deliverable: Revised report template with measurable quality lift.
Communities and volume
- Send 40 more outreach messages
- Post one sample in two communities with a free-sample request form
- Reply to every community comment within a few hours
- Track which message angle gets replies
Deliverable: Community post live, reply-rate data by angle.
Second touch
- Send the 48-hour follow-up with the sample attached
- Add agency logo upload and brand color to the PDF
- Add the re-run-on-request flow
- Review reply-to-sample conversion
Deliverable: Logo-branded PDF and a live second-touch sequence.
Convert warm replies
- Send the paid checkout link to every warm replier
- Publish delivery-time and refund terms clearly
- Monitor the first live paid order end to end
- Confirm the order lands in the Stripe/TrustMRR account
Deliverable: First paid order delivered within ten minutes.
Reliability and trust
- Time five consecutive generations and record p95
- Add an FAQ covering capture failures and refunds
- Fix any email deliverability issue
- Ask the first buyer for one improvement
Deliverable: Documented p95 delivery time and published FAQ.
Week-two data
- Calculate reply rate, sample rate, and paid rate
- Cut the worst-performing outreach angle
- Rewrite the message with the best-performing angle
- Set week-three contact targets
Deliverable: Channel scorecard with one message variant retired.
Week 3 · Repeat & systematize · 7 days
New segments
- Send 60 messages to a fresh agency list
- Target UK and Canada agencies explicitly
- Include a one-line result from an existing sample
- Log all sends in the same sheet
Deliverable: 60 new contacts reached with a segmented message.
Specificity
- Feed captured page copy and offer detail into the hypothesis prompt
- Add one competitor-comparison sentence per card
- Test output on 5 new page types
- Compare new cards against the day-9 baseline
Deliverable: Sharper, more page-specific test cards.
Public teardowns
- Publish two anonymized page teardowns built from the tool
- Include the sample PDF as a download
- Share both posts in the two chosen communities
- Add a sample request form link to each post
Deliverable: Two public teardowns with a working sample request form.
Volume plus follow-up
- Send 50 more outreach messages
- Send 24-hour follow-ups to day-15 contacts
- Reply to all inbound sample requests
- Track cost per sample generated
Deliverable: Inbound sample requests handled, cost per sample recorded.
Paid push
- Offer the paid report to all sample recipients
- Remove any remaining friction in checkout
- Add a one-line testimonial request to the delivery email
- Confirm no delivery exceeds ten minutes
Deliverable: Paid offers sent to every sample recipient.
Second orders
- Email every buyer for their next client URL
- Show the same $99 price with no discount
- Track repeat-order rate
- Fix the top delivery complaint
Deliverable: Repeat-purchase email live, first repeat order attempted.
Week-three data
- Recompute reply, sample, and paid rates
- Decide whether paid spend is justified
- Adjust the contact list for quality
- Set final-week targets
Deliverable: Week-three scorecard and final-week plan.
Week 4 · Scale what works · 7 days
Scale the list
- Send 60 messages to a third list segment
- Refresh the hook with a delivery-time claim
- Follow up with every 48-hour non-responder
- Log replies by segment
Deliverable: 60 new contacts and refreshed outreach copy.
Backlog quality
- Reorder the CSV columns to match common Notion imports
- Add a priority score explanation row
- Test the CSV import in Notion and Sheets
- Fix any parse error
Deliverable: Verified CSV import in Notion and Sheets.
Branding
- Add client name and date fields to the PDF cover
- Add a footer with agency contact details
- Test with three different logos
- Fix layout issues on small logos
Deliverable: Client-ready branded PDF cover and footer.
Zero-cost distribution
- Ask every buyer for one agency introduction
- Add a plain referral line to the delivery email
- Post one more sample in a community
- Track inbound from referrals
Deliverable: Referral ask sent to all buyers.
Final volume
- Send 60 final-wave messages
- Post the strongest sample with a clear request form
- Reply to all open threads
- Prioritize warm replies over new contacts
Deliverable: Final outreach wave sent, warm pipeline ranked.
Reduce refund risk
- Review all refund or complaint reasons
- Add a pre-purchase sample link next to checkout
- Clarify that lift ranges are hypotheses, not guarantees
- Update the delivery email wording
Deliverable: Refund-risk wording fixed in checkout and email.
Warm pipeline
- Send the last-call offer to every warm reply
- Offer one free re-run to convert hesitant buyers
- Confirm all paid orders are delivered
- Verify Stripe payouts and TrustMRR records
Deliverable: Warm pipeline closed, all orders delivered.
Days 29-30 · Consolidate · 2 days
Remaining prospects
- Send a final one-line message to non-responders
- Handle any outstanding sample requests
- Confirm no undelivered paid order exists
- Log every transaction against the contest window
Deliverable: Complete transaction log reconciled to Stripe.
Reconciliation and next step
- Reconcile gross revenue, refunds, fees, and ad spend
- Archive every generated report and CSV
- Document the winning message and segment
- Decide whether to continue past the contest window
Deliverable: Final reconciliation sheet and channel learnings document.
Revenue and profit scenarios
How the numbers are defined
Hackathon profit — Collected revenue − refunds − chargebacks − payment processing fees − paid marketing spend.
Business profit — Hackathon profit − operating costs (hosting, domains, tooling, AI/API delivery costs).
The contest ranks entrants on hackathon profit only. Hosting, tooling and AI/API delivery costs are not deducted by the organisers, but each sale still has to cover them — so we show both. Definitions verified on 2026-11-15 against the official rules.
Stated assumptions
- All scenarios use the single $99 one-time price; no discounts and no subscription (all scenarios).
- Payment processing is modeled at roughly 3.9–4.2% of collected revenue, consistent with Stripe card pricing plus typical international card mix (all scenarios).
- No paid marketing in Conservative and Base; the Upside case includes $150 for one sponsored community post or paid newsletter slot (Upside only).
- Refunds are 0% in Conservative and Base because the buyer sees a watermarked sample before paying; Upside assumes about 2% refunds at higher volume (all scenarios).
- Operating costs cover Vercel, domain, Supabase, Resend, Firecrawl, and OpenAI delivery cost at under $0.50 per report; these are excluded from hackathon profit but included in business profit (all scenarios).
- Conservative assumes 3 paid orders from roughly 150 contacts; Base assumes 10 from roughly 300; Upside assumes 30 from roughly 500 (all scenarios).
- Revenue is counted only if collected inside the contest window, November 16 to December 16, 2026 (all scenarios).
Conservative
3 customers- Refunds
- $0
- Chargebacks
- $0
- Payment fees
- $12
- Paid marketing
- $0
- Hackathon profit
- $285
- Operating costs
- $25
- Business profit
- $260
Base case
10 customers- Refunds
- $0
- Chargebacks
- $0
- Payment fees
- $39
- Paid marketing
- $0
- Hackathon profit
- $951
- Operating costs
- $45
- Business profit
- $906
Optimistic
30 customers- Refunds
- $60
- Chargebacks
- $0
- Payment fees
- $118
- Paid marketing
- $150
- Hackathon profit
- $2,642
- Operating costs
- $110
- Business profit
- $2,532
Model your own numbers
Runs in your browser · nothing is saved- Collected revenue
- $990
- Payment fees (3.9%)
- $39
- Paid marketing
- $0
- Hackathon profit
- $951
- Business profit
- $906
Hypothetical arithmetic, not a forecast. The fee rate is a flat 3.9% assumption, refunds and chargebacks are left at zero here, and none of this is tracked data.
Scenario models from stated assumptions, not verified revenue. These are projections only; no customers, sales, or traction exist yet.
Main risks and validation experiments
Risk 1Cold outreach reply rate is far below the plan, producing too few sample requests and no paid orders.
Run three contact waves with distinct message angles and retire the weakest after day 14; prioritize warm repliers over new contacts.
Send 50 messages in 48 hours with the free-sample offer and track replies, sample requests, and paid orders separately.
Fewer than 3 replies from 50 messages, or zero sample requests, after two message variants.
Risk 2AI-generated hypotheses feel generic and buyers will not pay $99 for output they could approximate with a general assistant.
Force each card to cite a captured page element as evidence and reorder cards by page-specific priority; show the evidence in the PDF.
Show one sample to five agency owners and ask whether they would present any three cards to a client without rewriting them.
Fewer than 3 of 5 say they would present at least three cards unedited.
Risk 3Page capture fails on bot-protected or JavaScript-heavy storefronts, producing poor reports and refunds.
Detect empty captures before generation, retry once with rendered mode, request an alternative URL, and refund if both attempts fail.
Run ten varied storefronts through capture and count failures and partial renders.
More than 3 failures out of 10 with no fixable pattern within one build day.
Risk 4Agencies prefer human CRO audits and treat a software artifact as a threat to their service value.
Position the output as the agency's own branded deliverable, never as a replacement for their judgment, and never sell to their end clients.
In five outreach replies, ask directly whether they would hand a branded PDF like this to a client at kickoff.
Sample recipients consistently say they would rebuild the artifact themselves rather than forward it.
Hour-by-hour: the first 24 hours
Hours 1-2
- Create the Vercel project and connect the domain
- Create the Supabase project and the orders, reports, and usage tables
- Create the Stripe account and connect TrustMRR before any payment
- Write the contest start date and window into the project notes
Output: Deployed skeleton with a database and a TrustMRR-connected payment account.
Hours 3-5
- Build the intake form with URL, brand name, baseline conversion rate, and traffic fields
- Wire Firecrawl capture and store the above-fold screenshot
- Cache captured page elements to a Supabase row
- Test capture on three storefronts
Output: Working capture that returns structured page elements and a screenshot for three URLs.
Hours 6-8
- Create the agency contact sheet with name, company, segment, and profile URL columns
- Log the first 60 agencies from LinkedIn searches
- Draft the free-sample outreach message and one follow-up
- Set the daily send quota to 40 contacts
Output: 60-row contact list and a finished outreach message.
Hours 9-12
- Dump captured page elements as raw text for one URL
- Confirm the heuristic fields map to real page data
- Note any missing element the hypothesis generator will need
- Push the day-one commit and set tomorrow's build order
Output: End-to-end capture proof and a written build order for day two.
Full 30-day launch report
Read the full generated report
TestPlan Scribe — 30-Day Launch Report
Concept: TestPlan Scribe (v1). JEV weighted score 6.35, total 63.5/100, band C, confidence 0.76. Strongest criteria: speed to first revenue (8.41) and solo build feasibility (7.58). Weakest: monetization potential (4.88) and competitive defensibility (4.99). This plan keeps the scored concept exactly as generated and adds execution detail only.
1. Product opportunity and positioning
TestPlan Scribe sells one artifact: a branded, client-ready A/B test plan for a single e-commerce product page, delivered as PDF plus CSV within ten minutes of payment. The buyer is the owner, account manager, or senior growth marketer at a 5–20 person Shopify growth agency, or a solo CRO consultant in the US/UK/Canada, who has just won or onboarded a client and needs something credible before the kickoff call.
The pain is deadline-driven. To look competent in that meeting, the buyer has to review the page, invent test ideas, guess at sample sizes, and format a deck — roughly half a day per client, with output that looks like every other agency's audit. TestPlan Scribe collapses this into a four-field form: URL, brand name, baseline conversion rate, traffic estimate. It returns page-specific evidence: captured headline, CTAs, forms, offer, image alt text, and above-fold screenshot; a heuristic score across clarity, offer strength, friction, trust, and mobile risk; twelve prioritized hypotheses each with control, variant copy, primary metric, and a labeled lift range; then sample size, duration, a branded PDF, and a CSV backlog for Notion, Sheets, or Asana.
Positioning stays narrow: not an experimentation platform, not live A/B execution, not consulting. Optimizely's original problem — teams cannot run rigorous experimentation — is real, but the enterprise-platform answer is unavailable to a nine-person agency. The wedge is the pre-test planning artifact agencies already sell and currently fake in spreadsheets. Revenue comes from repeat purchase per client page, not from broadening the product. Every lift figure is framed as a hypothesis with stated assumptions, never a guarantee.
2. Smallest sellable MVP
Pay $99, submit one URL plus brand and baseline rate, receive a branded PDF test plan and CSV backlog by email within ten minutes, with one automated re-run if capture failed. Eight features ship by day 5: intake form; Firecrawl capture with screenshot; rule-based heuristic scorer; AI generation of exactly 12 structured test cards; sample-size and duration calculator; branded PDF; CSV export; Stripe Checkout with automatic delivery and a usage ledger. Out of scope: live test execution, Shopify/GA4/Klaviyo integrations, team accounts, SSO, consulting.
3. Technical implementation plan
Next.js on Vercel, Firecrawl for capture, OpenAI with a strict JSON schema for the 12 cards, React-PDF for export, Stripe Checkout, Resend for delivery, Supabase for the order and usage ledger. A Stripe webhook writes the paid order and enqueues generation; the worker captures the page, caches elements, scores heuristics, calls the model once, computes sample size from baseline rate, a default 15% relative detectable effect, and daily traffic, then renders PDF and CSV, stores them, and emails signed links. The hardest risk is capture failure on bot-protected storefronts; the fix is rendered-mode retry, an alternative-URL request, and a refund if both fail.
4. Pricing and monetization
One price: $99 one-time per client test plan. No subscription, no seats, no tiers. An agency owner can expense $99 against a client engagement without approval. Growth comes from repeat purchase: each new client page is another $99 order at zero added acquisition cost. Refunds are limited by showing a watermarked sample before purchase.
5. First-customer acquisition
One channel: direct LinkedIn and email outreach to Shopify/e-commerce growth agency owners and solo CRO consultants. Build a 300-contact list from LinkedIn searches, Shopify Partner listings, and agency directories. The first message offers a free watermarked test plan for one client page; the first reply asks for feedback, not money. Daily cadence from day 7: 40–60 new contacts, one 48-hour follow-up with the sample attached, one weekly community post linking only to a free sample request form. Week-one goal: 100 contacts, 5 samples delivered, 12 replies.
6. 30-day roadmap
Days 1–5 build: accounts and TrustMRR connection, capture pipeline, scorer and hypothesis generator, math and PDF/CSV export, checkout and delivery. Day 6 produces three watermarked samples from public brand pages plus a 60-second demo. Days 7–14 run the first 150 contacts, harden report quality, and target the first paid order on day 12. Days 15–21 run a second wave of 120 contacts, publish two anonymized teardowns, and push sample recipients to pay. Days 22–30 run a third wave, ask every buyer for a second client page, fix the top complaint, and close the pipeline before the December 16 cutoff.
7. Revenue and profit scenarios (projections only)
All figures are projections from stated assumptions, not verified revenue. Conservative: 3 orders at $99, $297 gross, $12 in payment fees, no ad spend, $25 operating cost. Base: 10 orders, $990 gross, $39 fees, $45 operating cost. Upside: 30 orders, $2,970 gross, $60 refunds, $118 fees, $150 paid spend, $110 operating cost. Hackathon profit subtracts refunds, chargebacks, fees, and ad spend only; business profit also subtracts operating costs. Payment fees are modeled at roughly 3.9–4.2%, consistent with Stripe card pricing plus international mix. Every paid order must still cover its own delivery cost, which stays under $0.50 per report.
8. Risks and validation experiments
Cold outreach is the top risk: test 50 messages in 48 hours and abandon a variant with fewer than 3 replies. Generic output is second: show one sample to five agency owners and kill the format if fewer than three would present three cards unedited. Capture failure is third: run ten varied storefronts and count failures. Willingness to pay is fourth: ask sample recipients to pay, not to praise. Each test is cheap, runs in days, and has an explicit stop condition.
9. First 24 hours
Hours 1–2 create the Vercel project, Supabase tables, and Stripe account connected to TrustMRR. Hours 3–5 build intake and Firecrawl capture with screenshot storage and test three storefronts. Hours 6–8 log the first 60 agencies and finish the outreach message. Hours 9–12 prove the capture pipeline end to end on one URL and write the day-two build order.
All financial scenarios are projections based on stated assumptions. No customers, revenue, partnerships, or traction exist at launch.
This brief was generated from public data about established startups, then narrowed and scored for a solo 30-day launch. Revenue and profit figures are scenario models, not verified results, guarantees or evidence of existing customers. Validate demand yourself before you spend a month on it — and check the contest rules for what counts toward your score.
Related ideas
All ideas →Foldline Report
Paste a landing page URL and get a device-by-device Fold Map PDF that shows what visitors see before scrolling, with a ranked fix list for designers to ship.
Skeleton Sheet
Skeleton Sheet turns three competitors' best-performing Facebook and Instagram posts into a white-labeled 30-day content pack — 30 caption drafts, the structural blueprint behind each one, and a posting calendar — for freelance social media managers, delivered in under 12 hours.
Cartlift Recovery Pack
Cartlift turns a store URL into three import-ready abandoned-cart emails rendered with the merchant's own products, prices and shipping policy, delivered as a ZIP the merchant drops into Klaviyo, Shopify Email or Mailchimp the same day.