What is a critical error?
A critical error is a mistake an evaluation reports on its own line because one instance matters, such as a wrong allergen or a false merge.
In plain words: Some mistakes matter more than their share of a total. A wrong allergen or a false merge can cause harm in one instance. An evaluation lists these critical errors on their own line, with their own count, instead of hiding them in an average.
Published by Youzu · Editorial policy
- Published
Why it matters
An overall accuracy figure can look good while one rare mistake type is costly. Reporting critical errors separately shows the buyer the cases that need a person's check, and sets the standard that a method is judged on the worst mistakes as well as the typical ones.
How it works
Define
Before testing, list the mistakes that count as critical in your catalog.
Count
Score every record against your own labels and count critical errors apart, with their denominator.
Report
Show the critical line beside the overall figure, and state how many records were checked.
| Line | Records checked | Errors |
|---|---|---|
| All fields | 100 | A count of mistakes |
| Allergen fields | 100 | A count of critical mistakes |
| False merges | 100 | A count of critical mistakes |
Examples
A size read wrongly
A size M read as L is a mistake.
A false merge
Merging the 2-seat and 3-seat sofas is a critical error in this example, because it hides a real difference.
A missed prohibited listing
A listing that should have been escalated and was approved is counted on its own line, not inside the approval rate.
Common mistakes
Critical errors in the average. One blended figure hides the rare mistakes that matter most.
No denominator. A count of errors without the number of records checked cannot be compared across tests.
Related terms
- False merge
- A false merge joins 2 records that describe different products; a missed match leaves 2 records of the same product apart.
- Product data enrichment
- Adding missing information or improving existing information on a product record so products can be found, filtered and compared.
- Moderation ruleset
- A moderation ruleset is a versioned set of rules, each with a severity and a scope, that a team activates as one revision.
In Youzu
Youzu helps marketplaces, classifieds and retailers turn raw seller and supplier product data into structured, reviewable catalog records.
Sources
ISO/IEC 25012:2008 Data quality model, ISO. iso.org/standard/35736.html. Accessed .
Data Matching: Concepts and Techniques for Record Linkage, Entity Resolution, and Duplicate Detection, Springer. link.springer.com/book/10.1007/978-3-642-31164-2. Accessed .
FAQ
Questions buyers ask
Because one instance can matter more than its share of a total. A wrong allergen or a false merge can look small in an average and still be the case that costs most, so the evaluation shows it on its own line with its own denominator.
Evaluations start with a sample of your own catalog and expand to more workflows once the results are reviewed. You choose the records and decide which mistakes count as critical, so the result answers your own question and not a general one.
We start with a sample of your own catalog, and we publish customer results only with the customer's approval. Your evaluation results stay with you unless you agree otherwise, so you can test freely and share only what you choose to share.
Last reviewed
Related
Keep reading
Product
Catalog Intelligence for marketplaces and retailers
Catalog Intelligence maps products to your category tree, extracts category attributes, keeps variants and offers apart and flags issues for review.
Glossary
What is a false merge?
False merge defined in plain words, with a worked example, common mistakes and links to the guides and Youzu products that apply it.
Guide
Build or buy catalog AI: compare on the full workflow
In-house builds and model APIs are credible. Compare them with a vendor on the full workflow: identity, variants, offers, moderation, review, maintenance.