Use Case Intelligence RSS Search use cases Search
OpenAI Retail Japan

Mercari

Marketing Personalization · Sales Automation

Mercari needed to reduce the effort and guesswork sellers face when creating listings, because poor descriptions, missing attributes, and post-sale friction limited conversion and slowed marketplace growth.

Read the original story Explore all use cases
Over a few hundred AI-assisted listings are created per minute
a statistically significant increase in average sales per user
about two months from the initial concept to the final release

Reported impact

Business outcome

Mercari improved listing quality, increased conversion, saved sellers time, and created a scalable AI layer that now supports hundreds of AI-assisted listings per minute across its marketplace.

  • Customer Experience
  • Productivity Improvement
  • Revenue Growth

Solution

Mercari embedded OpenAI models into two listing workflows so sellers can create better marketplace content with less manual effort. The first feature, Mercari AI Assistant, combines offline GPT-4 analysis of top-performing category listings with online GPT-3.5 Turbo evaluation of live user listings to identify missing attributes and suggest improvements when items remain unsold. The second feature, AI Listing Support, takes uploaded product photos, uses GPT-4o mini to suggest categories, and then generates tailored titles and descriptions for each item. Mercari originally built that flow with GPT-4 and later moved it to GPT-4o mini for lower latency, better cost efficiency, and multi-image support, then continued refining outputs by comparing AI-generated descriptions to actual listings after launch. This architecture keeps AI inside the existing seller journey so the assistance feels native rather than like a separate tool.

Data flow

Seller photos and listing drafts enter Mercari's listing flows, where GPT-4 and GPT-3.5 Turbo power evaluation and optimization for existing listings, while GPT-4o mini classifies new photo-based listings and generates final titles and descriptions that are compared against real marketplace outcomes for ongoing refinement.

Solution architecture

5 components · 2 layers
  1. Compute
    • GPT-4 Analyzes top listings offline to identify the attributes that strong category listings should contain.
    • GPT-3.5 Turbo Evaluates active user listings online and returns real-time improvement suggestions.
    • GPT-4o mini Generates low-latency, cost-efficient listing content from multiple uploaded images.
  2. Serving
    • Mercari AI Assistant Reviews unsold listings against best practices and recommends changes that can improve sales outcomes.
    • AI Listing Support Lets sellers start from photos and receive categories, titles, and descriptions automatically.

Architecture clues

  • AI Listing Support moved from GPT-4 to GPT-4o mini for latency and cost reasons
  • Mercari compares AI-generated descriptions to actual listings to keep improving outputs
  • Mercari uses an offline analysis stage and an online suggestion stage for listing optimization
  • Multi-image support was enabled after the model transition to GPT-4o mini

Evidence from the source

GPT-3.5 Turbo evaluated active listings of users and provided real-time improvement tips
GPT-4 analyzed top listings in each category to identify key attributes
Originally developed with GPT-4, the feature later transitioned to GPT-4o mini for superior cost-efficiency and reduced latency
With AI Listing Support, users can list items effortlessly by simply uploading photos
Transparency Evidence & methodology 75/100 reference value · 90% confidence · 2 sources

Review how this reference was scored and trace each normalized claim back to its documented source.

01

Scoring methodology

How scores are calculated

Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.

Customer reference value75/100
  • Problem clarity20/20
    Measures whether the customer's operational problem is specific enough to compare.
  • Implementation depth25/25
    Rewards named components, data flow, integrations, and architecture detail.
  • Outcome specificity20/20
    Measures whether the reported benefit is concrete rather than a generic claim.
  • Quantified impact0/25
    0 metrics describe measurable impact.
  • Deployment scale10/10
    Rewards evidence of a live or scaled deployment and meaningful integration scope.
Evidence confidence90%
  • Source authority20/25
    The main evidence is an official vendor-published customer story.
  • Claim traceability23/25
    11 of 12 key fields link to a named source.
  • Direct evidence20/20
    4 direct source quotes retained.
  • Date provenance10/10
    The publication date comes from the source page or publisher.
  • Source transparency9/10
    2 source references document publisher, URL, authority, retrieval, and supported fields.
  • Implementation specificity8/10
    Rewards source-backed components, data flow, quantified impact, and a specific outcome.
02

Traceability

Sources & provenance

Field coverage shows which facts each source supports. Dates are shown only when documented by the source.

  1. OpenAI

    Mercari customer story

    First-party
    https://openai.com/index/mercari/
    Source type
    Vendor-published customer story
    Published
    Nov 13, 2024
    Supports
    • Customer
    • Business problem
    • Solution
    • Products
    • Solution components
    • Data flow
    • Integrations
    • technical area
    • use case category
    • Business outcome
    • outcome category
    • Metrics
    • Architecture
    • evidence quotes
    • Publication date
  2. Mercari Careers

    About Us

    First-party
    https://careers.mercari.com/en/about/

    The careers site lists "Number of Employees" as "2,075 (as of February 2024)."

    Source type
    Official company website
    Supports
    • Company size

Date evidence

  • Published on source page Nov 13, 2024 · OpenAI · First-party
    Open date source
Not separately sourced Industry, Region. Treat these normalized attributes as lower-confidence until a dedicated source is added.

Related

Similar use cases

See all like this
Bain Archived · Oct 2, 2026 Magalu Retail Brazil Magalu wanted to make product discovery, comparison, purchase, and post-sale support faster and more intuitive, replacing complex retail navigation flows with a conversational shopping experience that could still… 300,000 sellers Marketing PersonalizationSales Automation Lu · Lu's WhatsApp · WhatsApp In-house Jun 18, 2026 Target Retail United States As shoppers increasingly begin discovery and purchase journeys inside conversational AI tools, Target needed to make its assortment, account identity, loyalty benefits, and checkout available directly in those… AI-driven traffic to Target leapt 2,000% in the first quarter compared to last year. Marketing PersonalizationSales Automation ChatGPT · Copilot Checkout · Gemini app Google Cloud Archived · Feb 7, 2026 Verkkokauppa.com Retail Finland We've been a long-time user of Google Cloud, so it made sense to build on our existing data foundation in BigQuery. Google Cloud is a leader in AI innovation, and if anyone understands search, it's Google, so we were… 4% increase in products added to cart Marketing PersonalizationReal-time Analytics BigQuery · Vertex AI · Vertex AI Search