Put the plan into practice
Explore the decisions, examples and acceptance checks behind this work.
Preserve the failing file before editing it
A product import error is easiest to investigate when the original input, selected import behaviour and validation output are preserved together. Editing the spreadsheet repeatedly without keeping those versions can remove the example that explains the fault. Midoriweb starts with a small representative sample and the exact intended catalogue change.
Establish whether the import should create products, update existing values or replace a defined set of data. Those are materially different operations. A file that is valid for one purpose may have unintended consequences under another import behaviour, especially when blanks or omitted columns have special meaning.
Adobe’s import documentation describes the native process. A third-party importer or ERP connector may use different conventions, so identify the actual tool and version before applying native-import guidance.
Define the intended change and affected scope
Write down which fields should change and which should remain untouched. Include website assignments, store-view values, product types and attribute sets where relevant. This becomes the acceptance criterion for the import rather than treating “file accepted” as the goal.
Select a few products representing the important variations. A simple product, configurable parent and child variant may require different relationships. A test using only one uncomplicated row may miss the reason a production file fails or produces incorrect catalogue structure.
Confirm the source of truth for each field. If a scheduled integration owns stock or price, a manual import may be overwritten later. Coordinate the repair with that process so a successful immediate result does not disappear during the next feed.
Check identifiers and spreadsheet transformations
SKUs and other identifiers should survive the source export and editing process exactly as intended. Spreadsheet software can change leading zeros, interpret values as dates or alter numeric-looking identifiers. Compare the raw input with the original source rather than relying only on the formatted cell display.
Check encoding, delimiters, quoting and line structure. A malformed row can shift values into the wrong columns, making an apparently unrelated field fail validation. Inspect the machine-readable file around the reported record, especially where descriptions contain punctuation or line breaks.
Avoid broad search-and-replace corrections without understanding the field semantics. Replacing every blank or delimiter character can damage legitimate content. A controlled transformation should be documented and reproducible so the next export can be processed consistently.
Validate attributes and product relationships
Review whether the attribute codes and option values match the destination configuration. A label understood by a merchandising team may not be the identifier expected by the importer. Check required attributes for the selected product type and attribute set rather than applying one universal row template.
For configurable products, trace the parent-child relationship and the attributes used for variation. An imported parent without the intended children may exist in Admin but fail to produce a usable storefront selection. Test the resulting customer interaction, not only the record count.
Store-scoped values need deliberate handling. Confirm whether a row targets the default value or a specific store view and what inheritance should remain. An update that accidentally replaces inherited content across several storefronts can be more damaging than an import that stops with a clear validation error.
Treat media as a separate dependency
Image paths and remote sources need to be available to the importer in the expected location and format. A successful product-data import does not establish that images were downloaded, assigned to the right roles or generated for the frontend.
Our product image troubleshooting guide covers that path in detail. For the import investigation, keep the source media reference, resulting catalogue assignment and failed image response together. This helps distinguish a data-mapping problem from a media-delivery problem.
Do not embed long-lived secrets in image URLs or distribute private supplier links in general reports. Where access is required, agree a suitable transfer or retrieval process that fits the source and destination systems.
Distinguish validation failure from processing failure
Validation checks the input against a set of rules, but later processing can still fail because of runtime conditions, custom code or dependencies. Record whether the file is rejected before processing or stops after some records have changed. That distinction determines the recovery plan.
For a partial import, establish which records were applied. Do not assume a failed job left the catalogue untouched. Compare the affected sample and use the importer’s available result reporting. A full rerun may be safe for some updates and unsafe for other behaviours.
If runtime increases substantially, inspect file size, batch behaviour and background processing rather than simply raising every timeout. Our cron guide and indexer guide explain related stages that may delay the final storefront result.
Test recovery on a bounded sample
Correct the smallest reproducible input or implementation issue first. Run it in a representative environment with a known before-state, then compare the exact fields and relationships intended to change. Preserve the test file and outcome so the correction can be repeated reliably.
Include a product that should remain unchanged. This catches overly broad updates and incorrect blank-value handling. Check product visibility, website assignment, price, stock context, media and variant selection as appropriate to the import’s scope.
For destructive import behaviours, the backup and restoration plan must cover the affected data, not just the CSV file. Restoring an old catalogue snapshot over new orders or unrelated edits may introduce additional problems, so recovery should be scoped to the actual change.
Verify the storefront after processing completes
Wait for the relevant indexing or background work, then check the intended pages and purchase interactions. An Admin record can look correct while the product is absent from the expected website, has no selectable variant or displays stale information.
Compare a sample across the categories of data changed. For large imports, combine automated checks with targeted visual review. A summary count of successful rows is useful, but it cannot establish that product content, relationships and presentation are commercially correct.
Observe the next scheduled feed if another system owns the same fields. If that feed reverses the correction, the long-term fix belongs in the data ownership or transformation process rather than repeated manual imports.
How Midoriweb can help with catalogue imports
Our Magento development service and PIM integration service can connect catalogue requirements with reliable import behaviour. The investigation should identify the failing input or processing stage and leave a repeatable method for future updates.
For the initial enquiry, include the tool used, intended operation, error text, approximate file size and a redacted representative row. Explain whether any records changed before failure. That allows us to scope the work without requesting the entire commercial catalogue before it is needed.
Product import questions
Does a valid file guarantee a correct catalogue?
No. Validation is one stage. Verify the resulting attributes, relationships, scope and storefront behaviour against the intended change.
Can we rerun a failed import safely?
First determine whether it partly applied and how the selected import behaviour handles repeats. Preserve the original results before retrying.
Why do spreadsheet edits break SKUs?
Spreadsheet software may reinterpret identifiers as numbers or dates. Compare raw values and use a workflow that preserves identifiers exactly.