neurocheckout-ai/synthetic-abandoned-cart-email-examples

Synthetic Abandoned Cart Email Examples

An entirely synthetic, bilingual collection of abandoned-cart email drafts with transparent checklist annotations. It is intended for education, prototyping, and evaluation, and contains no real recipients, customer messages, orders, merchant data, or campaign results.

Dataset Description

The dataset mirrors the five visible checks in NeuroCheckout's public Abandoned Cart Email Checker:

• message clarity;
• primary call to action;
• synthetic personalization placeholders;
• send timing;
• excessive discount risk.

The scores reproduce deterministic educational rules committed with the generator. They are not conversion predictions, deliverability scores, legal reviews, or evidence that an email will perform well.

Composition

Dimension	Count
Total examples	180
Train	108
Validation	36
Test	36
English	90
French	90
First reminder	60
Second reminder	60
Final reminder	60
Each readiness label	60

Ten fictional product categories are used. All languages, stages, and quality variants for one product category stay in the same split, preventing closely related product concepts from crossing split boundaries.

Readiness Labels

Label	Score range	Meaning
ready_for_final_review	85-100	No critical checklist issue remains; final campaign review is still required
needs_focused_edit	65-84	One or more checklist areas need a focused correction
revise_before_scheduling	0-64	Multiple or serious checklist issues should be revised before scheduling

Each of the five check scores ranges from zero to twenty. issue_tags lists checks below sixteen, while primary_issue identifies the lowest-scoring check.

Schema

Field	Description
id	Stable synthetic draft identifier
draft_variant	balanced, focused_edit, or high_risk generation variant
language	en or fr
stage	first, second, or final reminder
product_category	Fictional product concept used to vary the draft
subject, preview_text, body, cta	Synthetic email content
delay_hours	Illustrative delay after cart abandonment
configured_discount_percent	Discount entered in the illustrative campaign settings
detected_discount_percent	Highest percentage detected in the draft text
clarity_score, cta_score, personalization_score, timing_score, discount_score	Transparent checklist scores from zero to twenty
readiness_score	Sum of the five checklist scores
readiness_label	Readiness band derived from the total
issue_tags, primary_issue	Checklist areas requiring attention
synthetic	Always true

Usage

from datasets import load_dataset

dataset = load_dataset("neurocheckout-ai/synthetic-abandoned-cart-email-examples")
print(dataset["train"][0])

Potential educational uses include testing rule reproduction, evaluating classification of draft issues, demonstrating score explanations, and exploring controlled email generation. Real campaigns require consent, suppression handling, localization, deliverability testing, accurate promotional terms, brand review, margin validation, and applicable legal review.

Generation and Validation

The committed generate_dataset.py script creates all rows deterministically and independently recalculates every annotation. It verifies:

• exact split, language, stage, variant, and label counts;
• unique identifiers and normalized draft content;
• score totals, readiness bands, issue tags, and detected discounts;
• product-concept isolation between splits;
• use of only {{ first_name }}, {{ product_name }}, and {{ cart_url }} placeholders;
• absence of email addresses, phone-like strings, and long digit sequences.

Limitations and Responsible Use

• The drafts are template-generated and intentionally balanced, unlike real campaign traffic.
• English and French wording does not represent every locale, register, brand, or accessibility need.
• The timing windows and discount thresholds are educational starting points only.
• The dataset contains no opens, clicks, purchases, deliverability outcomes, consent status, or causal performance labels.
• A high readiness score does not establish safety, compliance, fairness, brand fit, or production quality.
• The public rules and examples do not reveal or claim to represent NeuroCheckout's internal systems or customer workflows.

Never replace placeholders with real personal data in a public dataset. Do not use these examples to target vulnerable people, create deceptive urgency, bypass consent, or guarantee promotional outcomes.

License

The dataset and generation script are released under the Creative Commons Attribution 4.0 International license (CC BY 4.0).

Learn More

Read NeuroCheckout's educational guide: AI Models for Ecommerce.