Use a flat file export as your source of truth, a spreadsheet to find the errors, and a person to decide what each value should be. Tools surface duplicates and missing attributes. They cannot tell you which variation belongs as the parent. That judgment is the actual work.
The short version
- Export first, edit second. A full flat file of every active and inactive item is the only reliable starting picture.
- Write the rule set before anyone touches a field. Ambiguity is how catalogs get broken.
- Keep a change log. When ranking moves, the log is how you learn which change did it.
- Fix in order of revenue impact. Variations and duplicates before attribute tidiness.
- Decide what to keep using data, not sentiment. Review count and sales volume are the two weakest signals for that call.
Do this first: pull the export and freeze it
Go into Seller Central, request a full inventory report including inactive and suppressed items, and save an untouched copy with today's date. That copy is your rollback. Every seller who has destroyed a catalog with a bad bulk upload did it without one.
Then build a working sheet with one row per child item and columns for parent SKU, variation theme, category, browse node, title, bullet count, image count, A plus status, price, trailing sales, conversion rate, and return rate. Everything below runs off that sheet.
The cleanup checklist, and what done properly means
- Duplicates and orphans. Done properly means every product appears exactly once, and any listing carrying sales history has been merged rather than deleted. Deleting a listing with reviews throws away an asset you cannot rebuy.
- Variation families. Done properly means every child that a customer would consider an alternative sits under one parent, with the highest-converting child as the default. This is the highest-value item on the list and the one most often skipped.
- Category and browse node. Done properly means the item sits where its actual competitors sit, not where a hurried upload put it. Wrong nodes cost you filter eligibility and put you beside the wrong price points.
- Titles and keyword coverage. Done properly means the top search terms for the category appear once each, in readable language, with no duplication across title, bullets, and backend fields.
- Attribute completeness. Done properly means every field Amazon offers for the category is filled, including the unglamorous ones that drive filters. Empty attributes make you invisible to a customer narrowing a search.
- Images and A plus content. Done properly means a primary image that has been tested rather than assumed, six or more supporting images, and A plus content on anything with meaningful traffic.
- Suppressed and stranded inventory. Done properly means zero open errors and a named person who checks weekly, because this list regenerates itself constantly.
- Retirement. Done properly means items you will never restock are closed deliberately, with their sales history noted, rather than left to clutter reporting.
How to decide what deserves the work
Cleaning everything equally is the slow way to do this. Rank the catalog first, and rank it on more than the obvious two signals.
Most sellers sort by review count and sales volume. Those tell you what already worked, not what could work. When we research a market we look at 90 plus data points, including market size, growth trajectory, return rate, segment dynamics, and the rating gap between the leaders and the field. The same discipline applies inside your own catalog.
| Signal | What it tells you | Why volume alone misses it |
|---|---|---|
| Conversion rate vs category | Whether the listing, not the traffic, is the problem | A high-volume item can still convert badly |
| Return rate | Whether the listing is promising the wrong thing | Returns hide inside gross sales |
| Rating gap | Where a competitor is beatable on quality | Review count says popular, not good |
| Search impression share | Whether you are even eligible for the demand | Sales history is backward-looking |
| Margin after fees | Whether fixing it is worth anything | Revenue flatters low-margin items |
Fix the items where a poor conversion rate sits next to real demand. That is where cleanup turns into money.
What most agencies will not tell you
Half of what gets sold as catalog cleanup is a report. Someone runs an audit tool, hands you a list of errors, and invoices for the diagnosis. The list is not the work. The work is deciding what each field should say and then living with the consequences when ranking moves.
The other omission is timing. Any structural change resets some of what Amazon knows about a listing, so a merge or a category move can look worse for a fortnight before it looks better. A provider who does not warn you about that window either has not done this often or is hoping you will not notice. Agree the measurement window before the work starts, not after the dip.
Related answers
- Bundle creation and catalog cleanup services for Amazon
- What does a good Amazon account audit include
- Amazon A/B testing tools vs consultant
- How to choose between Amazon agency and automation software
- Amazon agency pricing and economics: the complete guide
Send us the export and we will tell you where the money is sitting, free, at Flapen.
