Commerce data operations — verify what changes and what must stay intact
Evidence151 products / 268 rows repaired; zero residual mismatches
Evidence151 products / 268 rows repaired; zero residual mismatches
For an office-furniture retailer, I owned product, option and price reconciliation alongside new storefront development. After the initial migration, I continued tracing and correcting inconsistencies in live sales data.
Values diverged across channels. Some products appeared in autocomplete but disappeared from search results: the two features read different copies of the data, so checking only the master missed the fault.
I confirmed that autocomplete and search read different tables, then compared code, name, model, visibility and inventory classification across the catalogue. The repair scope was 268 rows covering 151 products.
I prepared a change plan and backup, with updates conditional on both the original and master values. On MyISAM, I retained backup hashes and conditional recovery SQL instead of assuming transaction rollback.
An independent database session confirmed zero remaining mismatches in the five audited fields on 7 September. I also checked protected price, option and mapping hashes, the actual price-sync reader and nine public searches.
On 5 September, I audited 603 products in each of two Cafe24 stores. Core price, code, source-provided option, shipping and sales-state checks matched. Three whitespace-only name differences were recorded separately.
151 products / 268 rows repaired; zero residual mismatches
Figures describe the audits on 5 and 7 September 2026. The search repair did not change prices or send new prices to external channels. Client identities, product data and internal admin screens are withheld.
Describe the thing you're currently stuck on. I'll reply by first separating what I can take on from what I can't.