Prompt Details
Model
(claude-4-8-opus)
Token size
835
Example input
[PROMPT_OBJECTIVE]: Write cart abandonment email sequences that recover sales without discounting
[CURRENT_DRAFT]: Write an email to people who left stuff in their cart, make it convincing
[TARGET_AI_MODEL]: Claude
[DESIRED_OUTPUT]: A 3-email sequence with send timing, subject lines, preview text and one call to action per email, each under 150 words
[USE_CONTEXT]: Independent home fragrance brand, average order value 45 euros, repeat buyers aged 30-55, warm and unhurried tone, no countdown timers, no discount codes, no guilt-based copy
Example output
DIAGNOSIS
**Intent summary:** You asked for a cart abandonment email sequence. What you actually need is a prompt that forces a model to replace the missing discount with a substitute reason to buy β because "recover sales without discounting" is not a formatting constraint, it is a persuasion problem. Strip the discount and the model's default fallback is manufactured urgency, which your context also bans (no timers, no guilt). The implicit requirement nobody wrote down: define what fills the gap where the incentive used to be. Also unwritten: who the emails come from, what happens when the same person abandons twice, and how the model handles product specifics it does not know.
Role clarity: **Absent** β the draft assigns no role, expertise level, or point of view, so the model defaults to generic marketing-email voice.
Task specificity: **Weak** β "an email" and "convincing" are named; sequence length, spacing, escalation logic and what "convincing" means without an offer are not.
Context sufficiency: **Weak** β brand category, AOV, age range and tone are supplied, but there is no sender identity, no product names, no brand voice sample, no post-purchase context, and no guidance on what the model should do when it lacks product detail.
Output definition: **Strong** β three emails, send timing, subject line, preview text, one CTA each, under 150 words. Countable and verifiable.
Constraint coverage: **Weak** β the prohibitions are clear (no codes, no timers, no guilt) but they are all negative. Nothing states what replaces them, and nothing prevents the model from inventing scarcity, fabricating product claims, or inventing a shipping offer to substitute for the banned discount.
---
VERSION A β CONSERVATIVE
```
You are writing a cart abandonment email sequence for an independent home fragrance brand.
Audience: repeat buyers aged 30-55. Average order value is 45 euros.
Write a 3-email sequence. For each email, provide:
- Send timing (how long after abandonment)
- Subject line
- Preview text
- Body copy, under 150 words
- One call to action
Tone: warm and unhurried.
Do not use:
- Discount codes or any price reduction
- Countdown timers or expiry language
- Guilt-based copy
Use [PRODUCT] as a placeholder wherever a specific product name is needed.
Output the three emails in order, labelled Email 1, Email 2, Email 3.
```
---
VERSION B β OPTIMIZED
```
<role>
You are a lifecycle email copywriter for independent premium consumer brands. You specialise in
recovery sequences for brands that refuse to discount, because discounting trains repeat buyers to
wait for sales and erodes margin on a 45 euro basket. Your working assumption: an abandoned cart is
almost never a price objection. It is an interruption, a hesitation about fit, or a doubt about the
brand. You write to resolve those, not to apply pressure.
</role>
<brand_context>
Independent home fragrance brand. Average order value: 45 euros.
Audience: repeat buyers, aged 30-55. They have purchased before. They know the brand.
Voice: warm, unhurried, first person singular from the founder. Confident without being precious.
Writes to the reader as an adult with taste, not as a lead in a funnel.
</brand_context>
<objective>
Recover the sale by removing the reason the person stopped, not by making stopping expensive.
</objective>
<method>
Work through these steps before writing. Do not show your working.
1. For each of the three emails, choose ONE recovery angle from this list and do not repeat an angle:
- Reassurance: resolve a practical doubt (scent longevity, room size, returns, delivery, gifting)
- Craft: the making of the product β sourcing, blending, batch size, the person behind it
- Fit: help them choose between the item in cart and one alternative, so the decision gets easier
- Belonging: what the brand is for, and why this buyer already belongs to it
- Use: a specific, sensory moment the product is good for
2. Assign angles so the sequence escalates in intimacy, not in pressure.
3. Write each email at the length that serves the angle. Shorter is better. 150 words is a ceiling,
not a target.
4. Every email must be able to stand alone. Assume emails 1 and 2 were never opened.
</method>
<sequence_spec>
Email 1 β send 1 hour after abandonment. The interruption email. Assume they were distracted, not
undecided. Short. Practical. Low emotional weight.
Email 2 β send 24 hours after abandonment. The hesitation email. Address the most likely unspoken
doubt for a 45 euro home fragrance purchase.
Email 3 β send 72 hours after abandonment. The close. This is a graceful exit, not a last chance.
Give them a clean way to leave the cart behind without friction, and one reason to stay.
</sequence_spec>
<output_format>
For each of the three emails, output exactly this structure and nothing else:
EMAIL [n] β [send timing]
Angle: [the recovery angle chosen]
Subject: [under 45 characters]
Preview text: [under 90 characters, must add new information, must not repeat or complete the subject line]
Body:
[copy, under 150 words, plain text, no headers, no bullet lists unless the angle genuinely requires one]
CTA: [one call to action, under 5 words, as it appears on the button or link]
After the three emails, output:
SEQUENCE NOTE: [one sentence explaining how the three angles work together]
</output_format>
<constraints>
Absolute prohibitions:
- No discount codes, percentages off, free gifts, or any price reduction.
- No countdown timers, expiry dates, "cart expires", "reserved for", or any time-limited framing.
- No scarcity claims: no "low stock", "selling fast", "last few", "almost gone".
- No guilt: no "you forgot", "did you change your mind?", "we noticed you left", "still thinking?",
"don't miss out", and no implication that the reader owes the brand anything.
- No fabricated product attributes. If you do not know a scent note, ingredient, burn time, or
material, write around it or use a bracketed placeholder: [SCENT NOTES], [BURN TIME].
- Do not invent a shipping offer, gift wrap, sample, or any substitute incentive. If a shipping
policy is relevant, reference it as [SHIPPING POLICY] and let the brand fill it in.
- No exclamation marks. No emoji. No all caps.
- Do not repeat a subject line's promise inside the preview text.
Required:
- Use [PRODUCT] for the abandoned item and [FIRST NAME] for personalisation. Every email must read
naturally if [FIRST NAME] is empty.
- Exactly one call to action per email, linking to one destination.
- Each email must contain at least one concrete, sensory or specific detail. No email may consist
only of brand sentiment.
</constraints>
<self_check>
Before delivering, verify each email against these questions. Rewrite any email that fails.
- If I removed the brand name, would this read as generic ecommerce copy? If yes, rewrite.
- Have I created urgency through implication rather than stating it? If yes, remove it.
- Does the reader end this email knowing something they did not know before? If no, rewrite.
- Is the subject line under 45 characters and the preview text adding new information?
- Would a 45-year-old repeat customer find this respectful of their time and intelligence?
</self_check>
```
---
DECISIONS MADE FOR YOU
1. **Question: what replaces the discount as the reason to act?** Chose a fixed menu of five recovery angles (reassurance, craft, fit, belonging, use), one per email, no repeats. Reason: without a named substitute, models reliably fall back on urgency or flattery. A closed list makes the persuasion mechanism auditable rather than emergent.
2. **Question: what send timing?** Chose 1 hour / 24 hours / 72 hours. Reason: the 1-hour email catches genuine interruption while intent is live; 24 hours addresses considered hesitation; 72 hours closes before the cart goes cold. Standard for a considered-but-low-value basket. Changeable in one line.
3. **Question: who is the sender?** Chose first person singular from the founder. Reason: an independent brand's structural advantage over a large competitor is that a person can write the email. "We" wastes that. It also makes warmth credible without sentimentality.
4. **Question: does the sequence escalate?** Chose escalation in intimacy, explicitly not in pressure. Reason: your constraints ban every conventional escalation device, so the prompt must name what escalates instead or the third email will drift toward pressure by default.
5. **Question: what is email 3?** Chose a graceful exit rather than a final push. Reason: "last chance" is expiry framing wearing a disguise, and it is banned. Offering a clean way out preserves the relationship with a repeat buyer worth more than one 45 euro cart.
6. **Question: how does the model handle product details it does not have?** Chose mandatory bracketed placeholders and an explicit ban on invented attributes. Reason: fragrance copy invites hallucinated scent notes and burn times, which are product claims you would have to fix by hand or, worse, publish.
7. **Question: can free shipping substitute for the discount?** Chose no β referenced as [SHIPPING POLICY] only. Reason: free shipping on a 45 euro AOV is a margin concession by another name, and you did not authorise it. The decision stays yours.
8. **Question: how long should the emails actually be?** Chose "150 words is a ceiling, not a target". Reason: word limits stated as targets cause padding, and padding in a warm, unhurried voice reads as waffle.
9. **Question: how are subject line and preview text related?** Chose hard character caps plus a ban on the preview completing the subject. Reason: models default to treating preview text as a sentence continuation, which wastes an independent line of inbox real estate.
10. **Question: how is "no guilt" enforced?** Chose a banned-phrase list rather than a principle. Reason: "we noticed you left something behind" is guilt-adjacent copy that passes any abstract test. Named phrases are enforceable; principles are negotiable.
---
FAILURE MODE AND FIX
**Most likely failure:** urgency re-enters through the back door. With discounts and timers removed, the model substitutes implied scarcity and soft pressure β "before it's gone", "while it's still yours", "your cart is waiting" β which technically satisfies every stated prohibition while violating the intent, and reads as pressure to exactly the 30-55 repeat buyer you are trying to keep.
**Preventive line (already in Version B, isolate it here for reuse):**
`Have I created urgency through implication rather than stating it? If yes, remove it.`
---
VALIDATION CHECKLIST
Verify these by reading the prompt before you run it:
1. A specific expert role is defined, with a stated assumption about why carts are abandoned β not just a job title.
2. Every negative constraint has a positive counterpart: the angle menu tells the model what to do instead of discounting.
3. Send timing is specified per email, not left to the model.
4. Output format is literal and copy-paste shaped β a person can paste the result into an ESP without reformatting.
5. Character limits exist for subject line and preview text, not only a word limit for the body.
6. Placeholder handling is explicit, so no product attribute can be invented.
7. Banned phrases are named literally, not described as a principle.
8. Each email is required to stand alone, so the sequence survives unopened emails 1 and 2.
9. A self-check step exists inside the prompt, with rewrite instructions rather than warnings.
10. The prompt never asks the model to decide anything commercial (shipping, incentives) that belongs to you.
---
REFINEMENT LEVERS
Paste any of these at the end of Version B to change one dimension without rebuilding.
**Change the timing:**
`Override the send timing. Use 30 minutes, 12 hours, and 48 hours instead.`
**Make it shorter and colder:**
`Reduce the ceiling to 80 words per email. Remove the founder's first person voice and write in neutral brand voice. Keep all other constraints.`
**Make it warmer and more personal:**
`Raise the ceiling to 200 words for Email 2 only. Include one specific, unglamorous detail about the making or shipping of the product in that email.`
**Add a fourth email:**
`Add Email 4, sent 7 days after abandonment. Its only job is to keep the relationship, not the cart. It may not mention the abandoned item.`
**Test which angles work:**
`Produce two complete versions of the full sequence, using entirely different angle combinations. Label them Variant A and Variant B. Do not explain which you prefer.`
**Adapt for first-time buyers instead of repeat buyers:**
`Change the audience to first-time buyers who have never purchased. Remove all assumed familiarity with the brand. Prioritise the Reassurance and Craft angles.`
**Loosen the format for a plain-text send:**
`Output the body copy as it would appear in a plain-text email from a personal address, including a sign-off. Remove the CTA button framing and place the link inline in a sentence.`
**Tighten subject lines:**
`Generate three alternative subject lines per email instead of one. Cap all of them at 32 characters. Do not repeat a sentence structure across the nine.`
By purchasing this prompt, you agree to our terms of service
CLAUDE-4-8-OPUS
π€ Turn a rough idea into a ready-to-use AI instruction, diagnosed and rebuilt in a single run.
β
Scores your starting draft on 5 engineering axes
β
Returns 2 calibrated versions: conservative and optimized
β
Documents every decision made for you, with reasons
β
Adds a failure-mode warning and a validation checklist
π‘ Great for: marketers, consultants, developers, educators, agencies, solo founders, course creators, AI freelancers.
...more
Added 1 week ago
- Reviews for this prompt (1)
