Marketplace product-card editing
The process was built around shipping a product card: brief, machine draft, editorial review, publication, and weekly tracking of search positions.
- Role
- Experiment setup and process support
- Scale
- 72 marketplace product cards
- Method
- Machine draft, human edit
- Metric
- +351 average search positions after 4 months
Search positions by week
Two product-card groups from the experiment. The chart shows marketplace search position by week: lower is better.
- Keywords
- 7
- Draft + editor
- +146 / position 63
- Copywriter
- +244 / position 32
Category constraints
Descriptions affected marketplace search, but the category did not allow loose product claims. Each text had to combine keywords, product facts, platform restrictions, and readable language.
Risk control
Editorial review was mandatory: forbidden wording, facts, length, tone, and repetition were checked before publication. A fast draft could not be allowed to create legal or catalog-quality risk.
Experiment design
The experiment used 72 cards and two production tracks: copywriter from scratch and machine draft followed by editorial review. The comparison was based on search position, time, and budget.
Brief preparation
Before drafting, old descriptions were reviewed, weak wording was marked, and search queries were grouped by frequency. Each brief included product facts, keywords, tone, forbidden phrases, and target length.
Approach
Editorial review
The first drafts quickly showed typical issues: missed stop words, repeated phrasing, and overconfident product details. The checklist was tightened before scaling the process to the rest of the cards.
Draft settings
Prompt rules, length limits, and tone requirements changed after the first runs. Word limits worked better than character limits, and stricter source facts kept the text away from generic promises.
Search tracking
The copywriter track started faster. The draft-and-editor track needed setup time, then caught up after four months: optimized cards gained 351 positions on average and moved closer to the first result pages.
Commercial result
Production time fell by 55%, the budget fell by 32.5%, and search movement stayed comparable to writing every card from scratch. For the team, this was a way to publish more cards without dropping control.