Store Data Prompt Pack (Free CRO Toolkit 03) Edition 2, October 2026 (c) 2026 Anshul (anshuldesigns.com). Free to use with your team and your clients. Please don't sell it or remove this notice. You may send the client summary pages and the client summary template to clients without this notice. This file holds the 12 prompts of the Store Data Prompt Pack, as plain text, so nothing breaks when you copy and paste. Copy everything between COPY FROM HERE and TO HERE. Fill the [BRACKETS] with your own store details. Paste your export or text where it says [PASTE ... HERE]. Check every number the AI tool gives you against your own export before you act on it. ============================================================ STORE FACTS BLOCK Type this once and keep it as a plain text note. F04 asks for it, and any prompt that writes for your page works better with it pasted under CONTEXT. COPY FROM HERE STORE FACTS Store name: [STORE NAME] What I sell: [ONE LINE] Who buys: [ONE LINE] Price range: [LOWEST] to [HIGHEST] [CURRENCY] Shipping policy: [PASTE THE TEXT HERE] Returns policy: [PASTE THE TEXT HERE] Markets and currencies: [WHERE I SHIP AND THE CURRENCY SHOWN] Top 3 products: [NAME, PRICE, ONE LINE EACH] Claims I never make: [LIST] VOICE Three words: [WORD 1], [WORD 2], [WORD 3] Two sentences I like: [SENTENCE 1] [SENTENCE 2] Words I never use: [LIST] TO HERE ============================================================ Ids in these files (L and P are checks, F and A are prompts): L Scorecard 01, Landing Page Scorecard: https://www.anshuldesigns.com/downloads/anshul-landing-page-scorecard.pdf P Scorecard 02, Shopify Product Page Scorecard: https://www.anshuldesigns.com/downloads/anshul-shopify-product-page-scorecard.pdf F Pack 03, Store Data Prompt Pack: https://www.anshuldesigns.com/downloads/anshul-store-data-prompt-pack.pdf A Kit 04, Agency CRO Audit Prompt Kit: https://www.anshuldesigns.com/downloads/anshul-agency-cro-audit-prompt-kit.pdf Prompts in this file: F01 Funnel locator F02 Device gap F03 Landing page league table F04 Scorecard to fixes F05 Review miner F06 Pre-sale question finder F07 Search miner F08 Discount code check F09 Free-shipping check F10 Ad to page match F11 Fix or test F12 30-day fix plan Free CRO audit of your whole store: https://www.anshuldesigns.com/free-cro-audit?from=pdf-founder ============================================================ F01 Funnel locator The stage where shoppers drop, by device. About 10 minutes. Use when: Start here. You want to know where the store leaks before you change anything. Inputs to paste: - Conversion rate breakdown, All devices. Analytics > Reports > Conversion rate breakdown, open or closed funnel (say which). The four stages: all sessions, cart additions, reached checkout, completed checkout. - The same numbers per device (optional). If your admin shows them: Analytics > Reports > New exploration, the four session metrics by Device category. If not, use the All devices row only. Do not mix in GA4 funnel rows: GA4 funnels count users, not sessions. - Never paste: Customer lists or order exports. Pairs with: - P01 Buy button in reach on the first screen (Scorecard 02) - P29 Add to cart confirms, shows checkout (Scorecard 02) Where it goes wrong: A tablet with 20 completed checkouts once set the best rate. Under 30 is now LOW. Safety: Before you paste: delete names, emails, phone numbers, addresses and order numbers. After: check every number against your export. AI tools can miscalculate and can invent figures. COPY FROM HERE F01 FUNNEL LOCATOR ROLE You are an ecommerce analyst. You only use the numbers I paste. CONTEXT Store: [STORE NAME]. Sells [WHAT YOU SELL] to [WHO BUYS]. Date range: [START DATE] to [END DATE]. Currency: [CURRENCY]. INPUTS Shopify Analytics > Reports > Conversion rate breakdown as session counts (All sessions, Sessions with cart additions, Sessions that reached checkout, Sessions that completed checkout): one row for All devices and, if I have them, one per device: [PASTE THE TABLE HERE] STEPS 1. For each row calculate add-to-cart rate = cart additions / all sessions; checkout rate = reached checkout / cart additions; completion rate = completed checkout / reached checkout; conversion rate = completed checkout / all sessions. Write each as a / b = c% to 2 decimals (1,560 / 26,000 = 6.00%). 2. Per stage, the best rate is the highest among non-LOW devices; a device's share = its rate / the best rate, as a / b = c. Its weakest stage is its lowest share (None under rule 9). Label it EARLY (add to cart) or LATE (checkout, completion). All devices: Not applicable. 3. Pick the weakest stage, ignoring LOW and None. If there is none, or only an All devices row, write Not enough data and name Analytics > Reports > New exploration, dimension Device category. 4. Write 3 hypotheses for it, ranked by data support. Name a next report only if it has a column that could confirm or reject the hypothesis, else write None and a check I can do by hand. RULES 1. Use only what I pasted. If something you need is missing, write "Not enough data" and name the report or export that would answer it. 2. Show the arithmetic for every number you calculate, written as a / b = c. 3. Do not use industry benchmarks, averages or statistics unless I pasted them. 4. Quote the exact row, number or sentence you are judging. 5. Label anything you infer as ASSUMPTION. 6. Never write copy that states a fact I did not give you (prices, reviews, results, certifications, delivery times). 7. Output only the sections listed under OUTPUT, in that order. No introduction and no closing summary. 8. GA4 counts users, Shopify sessions: never mix their rows. If a later stage exceeds an earlier one, say so. 9. A device with fewer than 30 sessions that completed checkout is LOW (a rule of thumb): never the best, weakest stage Not enough data. A stage within 10% of the best rate is None (a rule of thumb). 10. Date rule: my range starts on or after 24 Sep 2026, or ends before 21 Sep 2026. Mention these dates only if my rows cross 21 to 23 Sep 2026: then say so and compare orders and sales, not sessions or conversion rates. OUTPUT Table 1 with columns: Measure | All devices | one column per device. Rows: Sessions, Add-to-cart rate, Checkout rate, Completion rate, Conversion rate, Weakest stage (stage, EARLY or LATE, share). Table 2 with columns: Rank | Hypothesis | Stage (EARLY or LATE) | Evidence from my numbers | Next report to pull Then one line: "Open this first: ". TO HERE ============================================================ F02 Device gap The gap between devices, as a monthly figure. About 10 minutes. Use when: One device converts worse than the rest and you want to size the gap. Inputs to paste: - The device table from F01. With its All devices row. Add each device's average order value if your admin shows it. - Average order value, and months. Analytics > Reports > Average order value over time. Months = days in the range / 30; 90 days is 3. - Never paste: Customer lists or order exports. Pairs with: - P03 Nothing covers product, price or button (Scorecard 02) Where it goes wrong: The ceiling assumes phones match your best device. Read the conservative table first. Safety: Before you paste: delete names, emails, phone numbers, addresses and order numbers. After: check every number against your export. AI tools can miscalculate and can invent figures. COPY FROM HERE F02 DEVICE GAP ROLE You are an ecommerce analyst. You only use the numbers I paste. CONTEXT Store: [STORE NAME]. Date range: [START DATE] to [END DATE], which is [NUMBER OF MONTHS] months. Currency: [CURRENCY]. Average order value for the range: [AOV]. INPUTS The device table from F01, all four counts per device with its All devices row, and if I have it each device's average order value: [PASTE THE DEVICE TABLE HERE] STEPS 1. Per row: sessions per month = sessions / months, conversion rate = completed checkouts / sessions. Mark a device LOW if fewer than 30 sessions completed checkout (a rule of thumb). Best rate = the highest rate among devices that are not LOW. 2. Per device: gap = best rate - its rate as a decimal (2.12% is 0.0212), 0 for the best; orders at stake per month = gap x sessions per month; ceiling at stake per month = orders x its AOV, or my AOV above (label ASSUMPTION: store AOV used for every device). 3. Conservative target: for each device that is not LOW and is below the All devices rate, repeat step 2 with the All devices rate as the best rate. 4. Inversion: YES if desktop converts below mobile, else NO; Not enough data if either is missing. 5. For the non-LOW device with the most ceiling at stake, name the stage whose rate (count / the one before) is the lowest share of the best device's. RULES 1. Use only what I pasted. If something you need is missing, write "Not enough data" and name the report or export that would answer it. 2. Show the arithmetic for every number you calculate, written as a / b = c. 3. Do not use industry benchmarks, averages or statistics unless I pasted them. 4. Quote the exact row, number or sentence you are judging. 5. Label anything you infer as ASSUMPTION. 6. Never write copy that states a fact I did not give you (prices, reviews, results, certifications, delivery times). 7. Output only the sections listed under OUTPUT, in that order. No introduction and no closing summary. 8. If any input is missing, write the Not enough data line, name the missing report and stop. No offer to continue, no request to paste again. 9. Calculate unrounded; round only in the tables (percentages 2 decimals, orders 1, money whole units). In Working write "ceiling, not a forecast" on rows with something at stake, LOW on a LOW device. 10. Date rule: my range starts on or after 24 Sep 2026, or ends before 21 Sep 2026. Mention these dates only if my rows cross 21 to 23 Sep 2026: then say so and compare orders and sales, not sessions or conversion rates. OUTPUT Table 1 with columns: Device | Sessions per month | Conversion rate | Best device's rate | Gap | Orders at stake per month | Ceiling at stake per month | Working Table 2 "Conservative target: your All devices rate" with columns: Device | Gap to All devices rate | Orders at stake per month | Ceiling at stake per month | Working Then two lines: "Inversion: YES" or "Inversion: NO" with one reason, and "Open this first: ". TO HERE ============================================================ F03 Landing page league table Landing pages ranked by what a fix could be worth. About 15 minutes. Use when: You send traffic to several pages and do not know which to fix first. Inputs to paste: - Sessions by landing page. Analytics > Reports > Sessions by landing page, with Conversion rate and Orders added. Greyed out? New exploration, dimension Landing page path. GA4: Engagement > Landing page; never mix. Delete first: Any column with names, emails or order numbers. - Average order value, and prices. Analytics > Reports > Average order value over time. Optional: product prices. - Never paste: Customer lists or order exports. Pairs with: - Scorecard 02 or 01, on the page this table ranks first Where it goes wrong: A hero product flatters the comparison. Compare within a collection or 30% of price. Safety: Before you paste: delete names, emails, phone numbers, addresses and order numbers. After: check every number against your export. AI tools can miscalculate and can invent figures. COPY FROM HERE F03 LANDING PAGE LEAGUE TABLE ROLE You are an ecommerce analyst. You only use the numbers I paste. CONTEXT Store: [STORE NAME]. Date range: [START DATE] to [END DATE], which is [NUMBER OF MONTHS] months. Currency: [CURRENCY]. Average order value for the range: [AOV]. INPUTS Shopify Analytics > Reports > Sessions by landing page for the date range above, with Orders (or Conversion rate) added from its Metrics menu. Optional: the price of each product page as path, price: [PASTE THE TABLE HERE] STEPS 1. Cut everything after "?" in each path. Type each path: Home ("/"), Collection ("/collections/"), Product (ends "/products/", with or without a collection before it) or Other. 2. Merge paths into one row per page (Product paths by handle, other Types by path) and add their sessions and orders. Orders = the Orders I pasted, else sessions x conversion rate (label ASSUMPTION: estimated). Conversion rate = orders / sessions. Confidence = LOW under 30 orders, else OK. 3. Best comparable page = the OK row of the same Type with the highest rate. A Product row compares only with Product rows in the same collection or within 30% of its price (a rule of thumb); with no prices pasted, say so. Gap = its rate - this row's rate, 0 if there is none or this row is best. 4. Ceiling at stake per month = gap x sessions / months x average order value. Rank by it, highest first, and show the top 10. RULES 1. Use only what I pasted. If something you need is missing, write "Not enough data" and name the report or export that would answer it. 2. Show the arithmetic for every number you calculate, written as a / b = c. 3. Do not use industry benchmarks, averages or statistics unless I pasted them. 4. Quote the exact row, number or sentence you are judging. 5. Label anything you infer as ASSUMPTION. 6. Never write copy that states a fact I did not give you (prices, reviews, results, certifications, delivery times). 7. Output only the sections listed under OUTPUT, in that order. No introduction and no closing summary. 8. Compare a row only with rows of the same Type. Never call a LOW row the best comparable page, or the page to open first: if no OK row exists, write Not enough data. 9. Date rule: my range starts on or after 24 Sep 2026, or ends before 21 Sep 2026. Mention these dates only if my rows cross 21 to 23 Sep 2026: then say so and compare orders and sales, not sessions or conversion rates. OUTPUT Table 1 with columns: Page | Type | Sessions | Orders | Conversion rate | Best comparable page | Gap | Ceiling at stake per month | Confidence Every cell that holds a calculated number shows its working as a / b = c, for example 65 / 4,000 = 1.63%. Under the table write "Ceiling, not a forecast." Then one line: "Open this page first: ". TO HERE ============================================================ F04 Scorecard to fixes Exact fixes and copy for your failed scorecard checks. About 20 minutes. Use when: You have a scored page and need the wording and settings to fix it. Inputs to paste: - Failed checks, with Pass if and Fix. Your scorecard (01 or 02): each unticked check as id, title, Pass if, Fix, effort, action tag. - Page text. Live page: select all, copy. Delete first: Reviewer names in a reviews widget. - Store facts block. Type it once; the template is in the text file. - Never paste: Customer names, emails or logins. Pairs with: - Every failed check in scorecards 01 and 02 Where it goes wrong: Unseen settings get invented menu paths. F04 names one only if your Fix line does. Safety: Before you paste: delete names, emails, phone numbers, addresses and order numbers. After: check every number against your export. AI tools can miscalculate and can invent figures. COPY FROM HERE F04 SCORECARD TO FIXES ROLE You are an ecommerce conversion analyst and copywriter. You only use the failed checks, page text and facts I paste. CONTEXT Store: [STORE NAME]. Page: [PAGE NAME OR URL]. Scorecard used: [LANDING PAGE OR PRODUCT PAGE]. INPUTS 1. The checks my page failed, one per line: id, title, Pass if line, Fix line, effort and action tag, as printed on my scorecard. [PASTE THE FAILED CHECKS HERE] 2. The page text (open the live page, select all, copy). [PASTE THE PAGE TEXT HERE] 3. My store facts block, including my voice. [PASTE THE FACTS BLOCK HERE] STEPS 1. For each failed check, quote the sentence on the page that shows the problem, or write Nothing on the page. 2. Write the fix so that it meets the Pass if line: exact copy using only facts from my facts block or the page text, in my voice, or the setting named in the Fix line I pasted. 3. Say where on the page it goes, using the page text's own headings. If the fix is not copy or a setting (for example new photos), start the Fix cell with NOT COPY and write the task in plain words. 4. Copy Effort and Fix or test from my failed checks. List every fact you needed and did not have in Table 2. RULES 1. Use only what I pasted. If something you need is missing, write "Not enough data" and name the report or export that would answer it. 2. Show the arithmetic for every number you calculate, written as a / b = c. 3. Do not use industry benchmarks, averages or statistics unless I pasted them. 4. Quote the exact row, number or sentence you are judging. 5. Label anything you infer as ASSUMPTION. 6. Never write copy that states a fact I did not give you (prices, reviews, results, certifications, delivery times). 7. Output only the sections listed under OUTPUT, in that order. No introduction and no closing summary. 8. Every number, price, promise and claim in new copy must appear in the facts block or the page text, else write FACT NEEDED in its place. Never write a claim from my "claims I never make" list. 9. Name a Shopify setting only if it appears in the Fix line I pasted; otherwise write CHECK YOUR THEME EDITOR. Each line of new copy is 20 words or fewer. OUTPUT Table 1 with columns: Check | What the page says now | Fix (exact copy or setting) | Where on the page | Effort | Fix or test Table 2 with columns: Fact needed | Why | Where to find it TO HERE ============================================================ F05 Review miner Your customers' own words, and the doubts the page leaves open. About 15 minutes. Use when: You have 30 or more reviews and the copy does not sound like your buyers. Inputs to paste: - Reviews of one product. Your reviews app's export, or the product page. Star rating and text, lowest first. 30 to 100 is enough. Delete first: Names, emails, order numbers, photo links. - Product page text. Live page: select all, copy. Delete first: Reviewer names in a reviews widget. - Never paste: Reviewer names, emails or order numbers. Pairs with: - P15 Problem in customer words (Scorecard 02) Where it goes wrong: One review becomes a headline claim. A result in it is QUOTE ONLY: ask first. Safety: Before you paste: delete names, emails, phone numbers, addresses and order numbers. After: check every number against your export. AI tools can miscalculate and can invent figures. COPY FROM HERE F05 REVIEW MINER ROLE You are a customer researcher. You only use the reviews and page text I paste. CONTEXT Store: [STORE NAME]. Product: [PRODUCT NAME]. Sells [WHAT YOU SELL] to [WHO BUYS]. Reviews: [NUMBER OF REVIEWS], lowest ratings first. INPUTS Reviews of this one product, from my reviews app's export or the product page. Star rating and text only, names removed: [PASTE THE REVIEWS HERE] The product page text (open the live page, select all, copy): [PASTE THE PAGE TEXT HERE] STEPS 1. Number the reviews R1, R2, R3 in the order I pasted them. Group them into themes: what customers say mattered (result, fit, quality, delivery, price, doubts). A review can sit in more than one theme. 2. Per theme give one customer phrase copied word for word, the count of reviews with their numbers, and whether the page text answers it (YES, PARTLY or NO, with the page sentence). 3. List the objections (doubts, surprises, complaints), each with a review quote and the page section where an answer belongs. Read the lowest ratings first. 4. Write 3 headline options. Each reuses words from one review and names that review. Give each a Status: QUOTE ONLY or OK. RULES 1. Use only what I pasted. If something you need is missing, write "Not enough data" and name the report or export that would answer it. 2. Show the arithmetic for every number you calculate, written as a / b = c. 3. Do not use industry benchmarks, averages or statistics unless I pasted them. 4. Quote the exact row, number or sentence you are judging. 5. Label anything you infer as ASSUMPTION. 6. Never write copy that states a fact I did not give you (prices, reviews, results, certifications, delivery times). 7. Output only the sections listed under OUTPUT, in that order. No introduction and no closing summary. 8. Customer words are copied exactly from one review, in quotation marks. Never merge two reviews or tidy a phrase. 9. A headline may contain only words and claims found in the review it names. One that states a measurable result (hours, days, sizes, outcomes) from one review is QUOTE ONLY: use it in quotation marks with the reviewer's permission, or check it against the product page text I pasted. OUTPUT Table 1 with columns: Theme | Customer words (verbatim) | Reviews mentioning it | On the page now? Table 2 with columns: Objection | Evidence (quote) | Where to answer it Table 3 with columns: Headline | Source review | Status TO HERE ============================================================ F06 Pre-sale question finder A real FAQ built from buyers' questions, plus reassurance lines. About 15 minutes. Use when: Customers keep asking the same questions before they buy. Inputs to paste: - Pre-sale questions. Support inbox, chat or DMs, as written. 20 to 50 is enough. Delete first: Names, emails, phones, addresses, order numbers. - FAQ now on the product page. Copy the FAQ section of the live page. - Shipping and returns policy. Settings > Policies, or the policy pages. - Never paste: Names, emails, phones or addresses. Pairs with: - P17 Pre-sale questions answered on the page (Scorecard 02) Where it goes wrong: A question asked once reads like a trend. Singles stay in the Noise row. Safety: Before you paste: delete names, emails, phone numbers, addresses and order numbers. After: check every number against your export. AI tools can miscalculate and can invent figures. COPY FROM HERE F06 PRE-SALE QUESTION FINDER ROLE You are a customer support analyst. You only use the questions, FAQ and policy text I paste. CONTEXT Store: [STORE NAME]. Product: [PRODUCT NAME]. Sells [WHAT YOU SELL] to [WHO BUYS]. The questions came from [EMAIL, CHAT OR DMS]. INPUTS Questions customers asked before buying, one per line, as they wrote them, names removed: [PASTE THE QUESTIONS HERE] The FAQ text now on the product page: [PASTE THE FAQ HERE] My shipping policy and returns policy text: [PASTE THE POLICIES HERE] STEPS 1. Number the questions Q1, Q2, Q3 in the order I pasted them. Group the questions that ask the same thing. 2. Per group give one example question copied word for word, the count with the question numbers, and whether my FAQ answers it (YES, PARTLY or NO, with the FAQ sentence). 3. Draft an answer from my policy text only, in my voice if I pasted one, copying its numbers exactly. If the policy does not cover it, write Not enough data and name the fact I must supply. 4. A group asked once is Noise: put all of them in one final row and draft no answer. 5. Write 3 reassurance lines of 11 words or fewer. A reassurance line is a short line under the buy button that answers a last doubt, such as delivery or returns. Each comes from one policy sentence, which you quote, and you count its words. RULES 1. Use only what I pasted. If something you need is missing, write "Not enough data" and name the report or export that would answer it. 2. Show the arithmetic for every number you calculate, written as a / b = c. 3. Do not use industry benchmarks, averages or statistics unless I pasted them. 4. Quote the exact row, number or sentence you are judging. 5. Label anything you infer as ASSUMPTION. 6. Never write copy that states a fact I did not give you (prices, reviews, results, certifications, delivery times). 7. Output only the sections listed under OUTPUT, in that order. No introduction and no closing summary. 8. Every number in a draft answer (days, prices, amounts) is copied exactly from the policy text. OUTPUT Table 1 with columns: Question group | Example (verbatim) | Times asked | Answered on the product page? | Draft answer (from my policies only) Table 2 with columns: Reassurance line | Words | Policy sentence it comes from TO HERE ============================================================ F07 Search miner What shoppers search for and cannot find, with a fix for each term. About 10 minutes. Use when: Your store has search and you suspect shoppers hit dead ends. Inputs to paste: - Searches by search query. Analytics > Reports. Search reports lag, so end the range 3 days ago. Delete first: A name or an email. - Searches with no results. Analytics > Reports. Same date range. - Product titles (optional). Products > Export, Title column only. - Never paste: Customer lists or personal search terms. Pairs with: - P39 Search visible in the phone header (Scorecard 02) Where it goes wrong: The share is approximate: Shopify counts sessions per term. Safety: Before you paste: delete names, emails, phone numbers, addresses and order numbers. After: check every number against your export. AI tools can miscalculate and can invent figures. COPY FROM HERE F07 SEARCH MINER ROLE You are an ecommerce analyst. You only use the search reports and product titles I paste. CONTEXT Store: [STORE NAME]. Sells [WHAT YOU SELL] to [WHO BUYS]. Date range: [START DATE] to [END DATE]. INPUTS Shopify Analytics > Reports, for the date range above: Searches by search query (term and count), then Searches with no results (term and count): [PASTE THE SEARCHES TABLE HERE] [PASTE THE NO RESULTS TABLE HERE] My product titles, one per line (Products > Export, Title column only): [PASTE THE TITLES HERE] STEPS 1. Skip terms with 1 or 2 searches. Count how many terms you skipped. No-result share of the top terms = the sum of the counts of the terms you kept in the no results table / the sum of the counts of the terms you kept in the searches table. It is approximate: Shopify counts sessions per term. 2. Give each remaining no-result term a Type: SPELLING (a misspelling of a word in my titles), SYNONYM (another word for something my titles show), STOCKED BUT NOT FOUND (a title contains the term, yet search found nothing) or NOT STOCKED (no title relates to it). Match ignoring capitals. 3. Give each term one Action. SPELLING and SYNONYM: add it to a synonym group in the Search & Discovery app. STOCKED BUT NOT FOUND: check the product is available to the online store, then search for the term yourself. NOT STOCKED: note the demand and do nothing yet. 4. Group the SPELLING and SYNONYM terms by the title they should find. RULES 1. Use only what I pasted. If something you need is missing, write "Not enough data" and name the report or export that would answer it. 2. Show the arithmetic for every number you calculate, written as a / b = c. 3. Do not use industry benchmarks, averages or statistics unless I pasted them. 4. Quote the exact row, number or sentence you are judging. 5. Label anything you infer as ASSUMPTION. 6. Never write copy that states a fact I did not give you (prices, reviews, results, certifications, delivery times). 7. Output only the sections listed under OUTPUT, in that order. No introduction and no closing summary. 8. Compare terms only with the titles I pasted. If I pasted no titles, write Not enough data in the Type column. 9. If the no results total is larger than the searches total, say so and do not force it to fit. OUTPUT One line: "No-result share of the top terms = a / b = c (approximate)", then "Terms skipped: n". Table 1 with columns: Term | Searches | Type | Action Table 2 with columns: Synonym group | Terms | Title it should find TO HERE ============================================================ F08 Discount code check Which codes get used, by whom, and which are clutter. About 10 minutes. Use when: Discounts are everywhere and you cannot say which codes are worth keeping. Inputs to paste: - Sales and discounts by code. Analytics > Reports > New exploration: Gross sales, Discounts, Net sales, Orders by Discount code. Or the Orders export (Discount Code, Discount Amount, Subtotal), grouped by code. Delete first: Any column that names a customer. - Your own codes (optional). Each code you made on purpose: CAMPAIGN, CREATOR or STAFF. - Never paste: Customer names or emails. Pairs with: - L20 Offer applied at checkout; none: tick (Scorecard 01) Where it goes wrong: A code you did not list looks like clutter. UNKNOWN codes get Keep: apps make codes too. Safety: Before you paste: delete names, emails, phone numbers, addresses and order numbers. After: check every number against your export. AI tools can miscalculate and can invent figures. COPY FROM HERE F08 DISCOUNT CODE CHECK ROLE You are an ecommerce analyst. You only use the discount numbers and notes I paste. CONTEXT Store: [STORE NAME]. Date range: [START DATE] to [END DATE]. Currency: [CURRENCY]. Codes I made on purpose, each with its purpose (CAMPAIGN, CREATOR or STAFF): [MY CODES AND PURPOSES]. INPUTS One row per discount code for the date range above: Discount code, Orders, Gross sales, Discounts, Net sales (from New exploration with the dimension Discount code). Or the Orders export, grouped by code: Net sales = Subtotal, Gross sales = Subtotal + Discount Amount (label ASSUMPTION: before returns): [PASTE THE TABLE HERE] STEPS 1. Ignore any row with no discount code. Never add rows together: an order with several discounts appears once in each discount's row. 2. Depth per code = discounts / gross sales, with discounts as a positive number. Count the codes used under 5 times, and all codes. 3. Type per code: CAMPAIGN, CREATOR or STAFF only if my list says so; a code that looks random with 1 order is APP-GENERATED (label ASSUMPTION); everything else is UNKNOWN. 4. Label: Retire only a CAMPAIGN code with under 5 orders. Share as a link a CAMPAIGN code with 5 or more orders in the top 10 by net sales (label ASSUMPTION). UNKNOWN gets Keep and the note CHECK WHO MADE IT. All other codes are Keep. 5. Show the top 10 codes by net sales, one row each, then the number of codes not shown. RULES 1. Use only what I pasted. If something you need is missing, write "Not enough data" and name the report or export that would answer it. 2. Show the arithmetic for every number you calculate, written as a / b = c. 3. Do not use industry benchmarks, averages or statistics unless I pasted them. 4. Quote the exact row, number or sentence you are judging. 5. Label anything you infer as ASSUMPTION. 6. Never write copy that states a fact I did not give you (prices, reviews, results, certifications, delivery times). 7. Output only the sections listed under OUTPUT, in that order. No introduction and no closing summary. 8. Quote net sales, never gross sales or total sales. If there is no Net sales column, write Not enough data in that column. 9. Never retire a CREATOR code without asking the creator. STAFF and APP-GENERATED codes are not promotions: always Keep. OUTPUT One line: "Codes used under 5 times: n of m". Table 1 with columns: Code | Orders | Gross sales | Discount | Net sales | Depth | Type | Keep / Share as a link / Retire Every cell that holds a calculated number shows its working as a / b = c, for example 65 / 4,000 = 1.63%. Then one line: "Codes not shown: n". TO HERE ============================================================ F09 Free-shipping check Whether free shipping pulls carts up, as far as your data shows. About 15 minutes. Use when: You have a free-shipping threshold and want to know if it moves carts. Inputs to paste: - Threshold and bucket width. Settings > Shipping and delivery: the free rate's amount. Width: 1/10 to 1/5 of it. - Abandoned subtotals. Orders > Abandoned checkouts > Export, last three months. - Order subtotals. Orders > Export, online store orders only. - Never paste: Any column except the subtotal. Pairs with: - P18 Delivery cost shown near the button (Scorecard 02) Where it goes wrong: Only a suggestion. Abandoned checkouts last three months. Safety: Before you paste: delete names, emails, phone numbers, addresses and order numbers. After: check every number against your export. AI tools can miscalculate and can invent figures. COPY FROM HERE F09 FREE-SHIPPING CHECK ROLE You are an ecommerce analyst. You only use the subtotals I paste. CONTEXT Store: [STORE NAME]. Date range: [START DATE] to [END DATE]. Currency: [CURRENCY]. Free shipping starts at a cart subtotal of [THRESHOLD]. Bucket width: [BUCKET WIDTH]. INPUTS Two lists of subtotals, one number per line, no other columns, the same range inside the last three months. Abandoned checkouts (recovered rows deleted): [PASTE THE ABANDONED SUBTOTALS HERE] Completed online store orders (ignore blank cells): [PASTE THE ORDER SUBTOTALS HERE] STEPS 1. Make 6 buckets of the bucket width, from the threshold minus 3 widths up to, but not including, the threshold plus 3 widths. A subtotal equal to a bucket's lower edge belongs to it. Ignore subtotals outside all buckets and count them. 2. Per bucket count Completed orders and Abandoned checkouts. Completion share = completed orders / (completed orders + abandoned checkouts). 3. Verdict: SUPPORTS if completed orders peak in the bucket just above the threshold (more than in the buckets on either side) and each of those three buckets has at least 30 orders (a rule of thumb); DOES NOT SUPPORT if there is no such peak; NOT PROVABLE if any of the three has fewer than 30. RULES 1. Use only what I pasted. If something you need is missing, write "Not enough data" and name the report or export that would answer it. 2. Show the arithmetic for every number you calculate, written as a / b = c. 3. Do not use industry benchmarks, averages or statistics unless I pasted them. 4. Quote the exact row, number or sentence you are judging. 5. Label anything you infer as ASSUMPTION. 6. Never write copy that states a fact I did not give you (prices, reviews, results, certifications, delivery times). 7. Output only the sections listed under OUTPUT, in that order. No introduction and no closing summary. 8. Never write that the threshold caused a change. Write "suggests" or "is consistent with". Quote the completed orders of the three buckets in the reasons. 9. Completion share is expected to be higher where shipping is free. Show it in Table 2 and never use it for the verdict. OUTPUT Table 1 with columns: Bucket | Completed orders | Peak (YES or NO) Table 2 with columns: Bucket | Abandoned checkouts | Completion share End each table with a row "Outside all buckets (ignored)" holding that count. Every cell that holds a calculated number shows its working as a / b = c, for example 65 / 4,000 = 1.63%. Then one line: "Verdict: SUPPORTS", "Verdict: DOES NOT SUPPORT" or "Verdict: NOT PROVABLE", followed by the reasons in 2 sentences. TO HERE ============================================================ F10 Ad to page match How well one ad and its page match, and a new first screen. About 10 minutes. Use when: A paid ad sends clicks that land and leave. Inputs to paste: - The ad: primary text and headline. Ads Manager > Ads > open the ad > Edit, or the Meta Ad Library. - Creative or screenshots (optional). Optional: screenshots of ad and page. - Destination URL and page text. The ad's URL, then the page text: first screen, main image, full text. Delete first: Reviewer names in a reviews widget. - Never paste: Names or emails from ad comments. Pairs with: - The six match checks of scorecard 01: L01, L02, L03, L04, L06, L10 Where it goes wrong: It cannot see images, so Product shown stays at 1 without screenshots. Safety: Before you paste: delete names, emails, phone numbers, addresses and order numbers. After: check every number against your export. AI tools can miscalculate and can invent figures. COPY FROM HERE F10 AD TO PAGE MATCH ROLE You are an ecommerce conversion analyst. You only use the ad and page text I paste. CONTEXT Store: [STORE NAME]. Sells [WHAT YOU SELL] to [WHO BUYS]. Traffic from this ad is [COLD OR WARM] (COLD people do not know my brand yet; WARM people have visited, bought or joined my list). Currency: [CURRENCY]. INPUTS The ad: primary text and headline. [PASTE THE AD TEXT HERE] Optional: the creative in words (what it shows, the product, any words on it), or screenshots of the ad creative and of the page: [PASTE THE CREATIVE DESCRIPTION HERE] The ad's destination URL: [PASTE THE URL HERE] The landing page: first-screen text, a one-line description of its main image, then its full text: [PASTE THE PAGE TEXT HERE] STEPS 1. Score 6 match points, each 0, 1 or 2. Headline: the page headline repeats the ad's main promise in the same key words. Offer: the same terms are on the first screen (2 if the ad has no offer). Product shown: the exact product and variant from the ad is on the first screen. Angle: the page leads with the ad's angle. Awareness stage: cold traffic gets what it is and who it is for before the offer; warm traffic gets the offer and a button first. Proof: a rating with a review count, a customer quote or a customer number is on the first screen. 2. 2 = clear match, 1 = partial, 0 = no match or missing. Total = the six scores added. Write one fix for every score under 2. 3. Rewrite the first screen (my voice if pasted, else plain): headline, subhead, button, proof line. RULES 1. Use only what I pasted. If something you need is missing, write "Not enough data" and name the report or export that would answer it. 2. Show the arithmetic for every number you calculate, written as a / b = c. 3. Do not use industry benchmarks, averages or statistics unless I pasted them. 4. Quote the exact row, number or sentence you are judging. 5. Label anything you infer as ASSUMPTION. 6. Never write copy that states a fact I did not give you (prices, reviews, results, certifications, delivery times). 7. Output only the sections listed under OUTPUT, in that order. No introduction and no closing summary. 8. Ad says and Page says are exact quotes. Write Nothing if the text has no match. 9. Score Product shown 1 at most unless I attached screenshots of the ad creative and of the page. 10. The rewrite uses only words and facts from the ad, the page text and my CONTEXT. If there is no real proof line, write Not enough data. OUTPUT Table 1 with columns: Match point | Ad says | Page says | Score (0, 1, 2) | Fix Then one line: "Total: n of 12", with the six scores added in brackets. Then the line: "A 2 equals a tick in Scorecard 01; a 0 or 1 is not ticked." Table 2 with columns: Headline | Subhead | Button label | Proof line TO HERE ============================================================ F11 Fix or test Ship, test, or ship and watch, with the sample-size working shown. About 10 minutes. Use when: You have one change idea and need to know whether your traffic can test it. Inputs to paste: - Evidence for the change. Your scorecard result, customer quotes from F05, or numbers from F01 to F03. - Page sessions and orders, 4 weeks. Analytics > Reports > Sessions by landing page: the page's row, with Orders. - Never paste: Customer names or emails. Pairs with: - Every check tagged Test in scorecards 01 and 02 Where it goes wrong: Lehr calls his rule crude, so weeks needed is a floor. Safety: Before you paste: delete names, emails, phone numbers, addresses and order numbers. After: check every number against your export. AI tools can miscalculate and can invent figures. COPY FROM HERE F11 FIX OR TEST ROLE You are an ecommerce analyst. You only use the numbers and evidence I paste. CONTEXT Store: [STORE NAME]. Page: [PAGE NAME OR URL]. Change idea: [THE ONE CHANGE]. Scorecard action tag for it: [FIX, FIX AND WATCH, TEST OR NONE]. Smallest relative lift worth detecting: [LIFT PERCENT] (use 20 if I leave it empty). INPUTS The evidence for the change (my scorecard result, customer quotes, numbers): [PASTE THE EVIDENCE HERE] The page's sessions and orders for the last 4 weeks (Sessions by landing page, or for a page shoppers reach after landing, a page-level report): [PASTE THE PAGE NUMBERS HERE] STEPS 1. p = orders / sessions on this page. lift = the lift percent above / 100. If orders is 0, step 2 is Not enough data, and so is the verdict for a TEST or NONE tag. 2. Visitors needed per version n = 16 x p x (1 - p) / (p x lift)^2. Total = 2 x n. Sessions per week = sessions / 4. Weeks needed = total / sessions per week. Can test in 4 weeks = YES if weeks needed is 4 or less. 3. Verdict: FIX if the tag is FIX. SHIP AND WATCH if the tag is FIX AND WATCH. If the tag is TEST or NONE: TEST if I can test in 4 weeks, otherwise SHIP AND WATCH. 4. Write the hypothesis: "Because , changing for will move ". Then the primary metric, 2 guardrail metrics and a stop rule: run the full weeks needed; for SHIP AND WATCH, compare orders for 4 weeks before and after on the same traffic sources. RULES 1. Use only what I pasted. If something you need is missing, write "Not enough data" and name the report or export that would answer it. 2. Show the arithmetic for every number you calculate, written as a / b = c. 3. Do not use industry benchmarks, averages or statistics unless I pasted them. 4. Quote the exact row, number or sentence you are judging. 5. Label anything you infer as ASSUMPTION. 6. Never write copy that states a fact I did not give you (prices, reviews, results, certifications, delivery times). 7. Output only the sections listed under OUTPUT, in that order. No introduction and no closing summary. 8. In the verdict reason say this is a planning estimate (80% power, 5% two-sided significance) and a floor: Lehr calls it crude and exact calculators give somewhat higher numbers. 9. Date rule: my range starts on or after 24 Sep 2026, or ends before 21 Sep 2026. Mention these dates only if my rows cross 21 to 23 Sep 2026: then say so and compare orders and sales, not sessions or conversion rates. OUTPUT One line: "Verdict: FIX", "Verdict: TEST" or "Verdict: SHIP AND WATCH", with one reason. Table 1 with columns: Input | Value. Rows: Sessions, Orders, p, lift, Action tag. Table 2 with columns: Measure | Working | Result. Rows: Visitors needed per version, Weeks needed, Can test in 4 weeks? Then four lines: Hypothesis, Primary metric, Guardrails, Stop rule. TO HERE ============================================================ F12 30-day fix plan A week-by-week plan built from your failed scorecard checks. About 15 minutes. Use when: You have a scorecard result and a list of failures and need an order of work. Inputs to paste: - Failed checks. Your scoresheet from scorecard 01 or 02: each unticked check with id, title, tier, effort and action tag. - Who can do what, and your hours. For example: you for copy and settings, a developer for theme changes, a designer for images. - Never paste: Private contact details of staff. Pairs with: - Every failed check in scorecards 01 and 02 Where it goes wrong: Hours per task are assumptions (Minutes 0.5, Days 16, Project 40). If week 1 overfills, change the hours, not the plan. Safety: Before you paste: delete names, emails, phone numbers, addresses and order numbers. After: check every number against your export. AI tools can miscalculate and can invent figures. COPY FROM HERE F12 30-DAY FIX PLAN ROLE You are an ecommerce project planner. You only use the checks, people and hours I paste. CONTEXT Store: [STORE NAME]. Start date: [START DATE]. Hours I can work on this each week: [HOURS PER WEEK]. INPUTS My failed checks from the scorecard scoresheet, one per line: id, title, tier, effort and action tag as printed: [PASTE THE FAILED CHECKS HERE] Who can do what (for example: me for copy, a developer for theme changes, a designer for images): [PASTE THE PEOPLE HERE] STEPS 1. Put each failed check on one rung, first match wins. Rung 1: tier Critical. Rung 2: High, effort Minutes, not tagged Test. Rung 3: Medium, Minutes, not Test. Rung 4: High, Days, not Test. Rung 5: High with effort Project, or any other check tagged Test. Rung 6: everything else. 2. Order tasks by rung, then Critical, High, Medium, then Minutes, Days, Project, then id. The task is the action that makes the check title true. Do not invent details. Give each task an owner from my list, else No owner. 3. Hours per task (ASSUMPTION): Minutes 0.5, Days 16, Project 40. Rung 1 tasks start first, inside my weekly hours. A task that does not fit the hours left starts the next week; a task longer than a week takes the weeks it needs. 4. Plan at most one task tagged Test in the 30 days (a rung 1 Test task takes that place; a second goes to Later). Send other Test tasks, rung 6 and any task that would start after week 4 to Later. Add a last row, week 4: re-score the page with the scorecard on day 30 (1 hour). 5. Done when = the check's pass line is met. What to watch: FIX AND WATCH, orders for 4 weeks after against 4 weeks before, same traffic sources; TEST, the primary metric set before the test; FIX, Nothing. RULES 1. Use only what I pasted. If something you need is missing, write "Not enough data" and name the report or export that would answer it. 2. Show the arithmetic for every number you calculate, written as a / b = c. 3. Do not use industry benchmarks, averages or statistics unless I pasted them. 4. Quote the exact row, number or sentence you are judging. 5. Label anything you infer as ASSUMPTION. 6. Never write copy that states a fact I did not give you (prices, reviews, results, certifications, delivery times). 7. Output only the sections listed under OUTPUT, in that order. No introduction and no closing summary. 8. Every task has an owner or No owner, never blank. 9. Date rule: my range starts on or after 24 Sep 2026, or ends before 21 Sep 2026. Mention these dates only if my rows cross 21 to 23 Sep 2026: then say so and compare orders and sales, not sessions or conversion rates. OUTPUT Table 1 with columns: Week (1 to 4, or Later) | Check | Task | Owner | Effort | Done when | What to watch Then one line per week 1 to 4: "Week n load: a / b hours", and one line: "Re-score on: day 30 from my start date". TO HERE ============================================================ Shopify, Google Analytics, GA4, Meta, Claude, ChatGPT and Gemini are trademarks of their owners. This guide is independent and not endorsed by any of them.