Skip to content
Work

Amazon

Improving shopper experience across 2M+ product listings on Amazon

Used AI to find and fix 2M+ Home & Kitchen listings that were hard to search or didn’t match the product, so shoppers could find the right item and get what they expected.

Senior Vendor Manager · Home & Kitchen2020 – 2021

ProductStrategy & Ops

Outcomes

item data quality score
69% → 98%
Seasonal Home Decor and Event & Party Supplies
product listings fixed with AI
2M+
Seasonal Home Decor and Event & Party Supplies
Walk the Store
Award
Home & Kitchen, for “relentlessly see through customer problems”

How it fit together

  1. 01 · Flag

    Three ways to spot a bad page

    AI flags on top-revenue products, customer returns and complaints, and walking the store myself

  2. 02 · Diagnose

    Find the gap

    Can shoppers find it? Does the page match the product that arrives?

  3. 03 · Fix

    Rewrite the listing

    Titles, images, attributes, bullets, and keywords, with the vendor or directly

  4. 04 · Scale

    Hunt the pattern

    Search the catalog for the same defect and fix every listing that has it

Result → 2M+ listings fixed with AI · item data quality 69% → 98% across 2 categories · Walk the Store award, Home & Kitchen
How I found and fixed product pages, from three kinds of signal to fixes made across the catalog.

Context

On Amazon, a product page has two jobs. It has to show up when a shopper searches, and it has to describe the product that actually arrives. As a Senior Vendor Manager in Home & Kitchen, I owned two categories, Seasonal Home Decor and Event & Party Supplies, and I treated the pages of their top-selling products as part of my P&L, because that is where most customers land.

Challenge

Many top-revenue products had thin or inaccurate pages. Short titles left out the words shoppers search for, attributes were missing, and some pages promised something the product wasn’t. The result was lost traffic for products customers would have bought, and returns from customers who got something other than what they expected.

Approach & key decisions

  1. Let AI pick where to look

    Step 1

    I used AI to flag listings in my categories with poor searchability or a weak page, starting with top-revenue products, and fixed 2M+ product listings this way.

  2. Rewrite pages for search and for clarity

    Step 2

    I rebuilt titles, images, attributes, bullets, and keywords. For a set of Philips Christmas lights, the title went from “Faceted Mini Christmas Lights” to one that names the green wire, indoor/outdoor UL listing, total length, and bulb spacing: the details shoppers search and filter on.

  3. Read returns for broken promises

    Step 3

    I read customer complaints and return comments to find gaps between what a page promised and what arrived. A Hallmark gift box listed as “large” and “white” was in fact 7.25 inches wide and ivory. I worked with the vendor to name the real color, put exact dimensions in the title, bullets, and main image, then found 12 more gift boxes and bags with the same problem.

  4. Walk the store every week

    Step 4

    I regularly shopped my categories the way a customer would. In browse, I reviewed the top products in each category and moved the miscategorized ones. On detail pages, I checked titles, images, attributes, bullets, variations, and duplicates. On deals, I checked copy, images, landing pages, and categories.

Work sample

Before and after: two product pages and a checklist

Two pages I fixed, and the Walk the Store checklist I used every week. The slides are my summaries, built from my original work. Amazon-internal data and tools are left out.

  • Philips Christmas lights product page before and after: a short title becomes a searchable one, with more images, attributes, and benefit-led bullets
    Searchability: a top seller’s page, rebuilt
  • Customer return comments saying the gift box was smaller and not white, next to the listing that said Large and White
    Returns showed the page broke a promise
  • Hallmark gift box listing after the fix: Ivory Pearl in the title, dimensions on the main image, off-white in the bullets
    The fix, then 12 more listings like it
  • Walk the Store checklist for browse, detail pages, and deals, and the Walk the Store award
    What I checked when I walked the store

Execution & collaboration

  • Worked with vendors to update titles, descriptions, and images, and made catalog fixes directly where I could.
  • Started with top sellers in every review, so each fix reached as many customers as possible.
  • Set up tracking of reviews and returns for the months after each fix, to check that complaints fell.
  • Planned an analysis of how many Home & Kitchen listings had the same color or size mismatch, to decide whether to ask the product and computer vision teams for automatic detection.
Visit Amazon

You might also like