Post-purchase and winback for online stores: four no-code AI steps

Act as an ecommerce retention analyst for [BRAND_NAME], which sells [PRODUCT_CATEGORY].
Below are recent return reasons, low reviews and customer questions (personal details removed):
[PASTE_NOTES]

1. Root causes: group these into the top 3 preventable misunderstandings. For each, give how many notes mention it and quote 2 short examples.
2. "While your order ships" email: write a friendly email, sent the day after purchase, that answers those 3 misunderstandings with practical tips. Use only facts from the notes and from this product info: [PRODUCT_FACTS].
3. Product page fixes: write 3 short bullets for the top of the product page that set the right expectations before checkout.
4. List any issue that looks like a product defect or a shipping problem rather than a misunderstanding, so I can handle it separately.

Here is our store policy, word for word:
[PASTE_POLICY_TEXT]

Write 5 short, friendly customer-support replies for [BRAND_NAME] that follow this policy exactly:
1. Return request: thank them, then offer two choices: (A) an exchange or store credit [CREDIT_OFFER], or (B) standard refund steps under our policy.
2. "Where is my order?" (stuck in transit): explain the tracking status in plain words and say what we will do and when, using only our policy: [LOST_PACKAGE_RULE].
3. Arrived damaged: apologize and explain the next step under our damage policy (photo request, replacement or refund).
4. Cancel or change an order: one version for orders not yet fulfilled, one for orders already shipped.
5. Subscription pause: offer a skip or a frequency change before cancelling, if our policy allows it.

Rules: never promise credit amounts, free replacements, deadlines or refunds the policy doesn't state. Where the policy is silent, write [POLICY GAP: decide] instead of inventing a rule. Use [CUSTOMER_NAME] and [ORDER_NUMBER] as placeholders.

Act as an ecommerce data analyst. I've attached an order export with personal details removed. Each row is an order line with: Customer code, Created at (order date), Lineitem name, Lineitem quantity, Lineitem price, Discount Code.
Entry products: [ENTRY_PRODUCTS]. Companion products: [COMPANION_PRODUCTS]. Today's date: [TODAY].

1. Repeat timing: for customers with 2 or more orders, give the median days between order 1 and order 2 and how many customers that is based on.
2. Build three segments. Show the exact rule and the customer count for each:
   - A, Lapsed regulars: [MIN_ORDERS]+ orders and no order in the last [LAPSE_DAYS] days.
   - B, One-time entry buyers: one order, containing an entry product, placed [MIN_DAYS]–[MAX_DAYS] days ago, and no companion product.
   - C, Discount-only buyers: every order used a discount code.
3. For each entry product, the product most often bought next, with counts. If fewer than [MIN_COUNT] customers support a finding, say the data is too thin.
4. Write each rule in plain words so I can rebuild it in [EMAIL_TOOL].
Show your working and don't invent numbers.

Write a 3-email plain-text winback sequence for customers who bought [HERO_PRODUCT] [X] days ago and haven't ordered again. These customers opted in to email marketing.

Rules:
- Write as short personal notes from [FOUNDER_NAME], founder of [BRAND_NAME]. No banners or heavy formatting.
- Email 1 (Day [X]): ask how [HERO_PRODUCT] is working for them, share one practical tip from this list: [TIPS], and mention [COMPANION_PRODUCT] as a natural next step.
- Email 2 (Day [X+7]): offer [OFFER] for past customers. Mention an end date only if it's real: [OFFER_END_DATE].
- Email 3 (Day [X+14]): four sentences asking whether they switched, don't need more yet, or had a problem we can fix, inviting a direct reply.
- Every email: an honest subject line, and end with the placeholders [UNSUBSCRIBE_LINK] and [POSTAL_ADDRESS].