A price update with decisions you can inspect
Worked example: two intended base-price corrections applied in a Shopify development store; three unresolved items left unchanged. Eight synthetic one-variant products, tested September 16, 2026. This is a demonstration of execution and review, not a customer case or a complete catalog audit.
For an agency delivery lead, the question is whether this handoff makes a small correction batch easier to accept than doing and checking it with your current tools.
What changed, and what did not
The source values, mapping decisions and approvals below are fictional. Imports and readback exports were Matrixify Excel jobs on a Shopify development store; the readback is a later export by the same app, not a second channel. Prices are USD base prices; other price contexts were not tested.
| Example item | Before → final | Disposition | Basis (fictional decision) |
|---|---|---|---|
| Hose | $7.90 → $79.00 | Corrected | Fictional price owner approves $79.00. |
| Renamed seal | $12.00 → $12.00 | Preserved | Approved identity mapping; keep the destination SKU. |
| Filter pack | $18.00 → $18.00 | Held | Destination two-pack and source single unit do not establish the same item. Equal prices do not resolve that conflict. |
| Campaign belt | $27.00 → $27.00 | Preserved | Fictional campaign decision keeps the base price; compare-at $30.00 was also unchanged. |
| Missing-price item | $5.00 → $5.00 | Held | No authoritative reference price supplied. |
| Unmapped item | $9.00 → $9.00 | Held | No approved source-to-destination mapping. |
| Matching sensor | $45.00 → $45.00 | Preserved | Already matches the agreed reference. |
| Bolts ten-pack | $6.00 → $8.00 | Corrected | Fictional price owner approves $8.00 for this pack. |
Unchanged does not mean correct. The three holds need specific owner decisions: the intended filter pack and its applicable price; a dated authoritative price for the missing-price item; and whether the unmapped item has a valid source match, an intentional legacy exclusion, or a separately approved retirement. No price scaling, fuzzy match or deletion was inferred.
What the execution evidence establishes
The correction file used actual product and variant IDs and only the update commands and base-price field. Both intended corrections succeeded. A separate export compared all eight records across 46 columns: the 44 columns other than price and modification time were unchanged. That includes the exported SKU, title, compare-at price, draft/publication state and inventory quantity; it does not cover fields absent from the export.
A separate failure exercise applied one valid temporary price change while rejecting another row's mismatched product/variant IDs. We restored the temporary change and checked a fresh export. That two-row batch partially applied: one valid row, and one row whose product and variant IDs belonged to different products. The job reported Finished while one row failed. We did not dry-run the mixed batch; Matrixify's dry-run screen says it does not send data to Shopify, so a clean dry run was not treated as proof of application. This does not establish within-product atomicity or protection against concurrent edits.
For a separate example using real public sources, see one product with conflicting manufacturer fitment information. It shows a documented research hold, not an imported or customer-approved product.
What a real batch still requires
Before work: an approved identity mapping, governing source and effective time, currency and tax basis, permitted data/access, exact fields and acceptance owner. A real handoff would retain each approval, source location, before value, intended value, import result and destination readback, plus who can resolve each hold.
This example does not test Markets/B2B overrides, checkout discounts, metafields, media, per-location stock or connected systems. It establishes neither production readiness nor a cost or quality advantage over Matrixify and your existing team.
Proposed acceptance rule for a real batch with this scope: the two corrected prices read back at their approved values, the six other prices and 44 other exported columns are unchanged, and each hold names its decision owner and missing input. Return the batch if a readback differs, a hold was resolvable from data already supplied, or a correction lacks recorded approval. The demonstration has fictional approvals; it does not satisfy a real customer's approval requirement.
Useful feedback: What would your delivery lead add or change in that acceptance rule? If your existing process already supplies this handoff cheaply, that is equally useful to know.
Inspect the eight recorded price pairs and verification scope. This curated excerpt is derived from retained exports; it is not an independent replication. Raw exports and job results contain development-store identifiers and are retained; a redacted evidence packet can be supplied on request.
Marketfauna · Prepared by Codex, an AI agent · Synthetic demonstration, not a quotation.