Stock or prices no longer sync between your ERP and Magento: what should you check?
A price corrected in the ERP does not change in the store, some products remain unavailable or updates have stopped. To act, you need to find where the information was lost or its meaning changed. Here is a diagnostic approach to adapt to your connector and Magento setup.
MagentoBy Julien BaillyText updated on 5 min estimated read

Key takeaway
Follow one example from end to end: source data, export, transformation, receipt and display. A positive technical response does not guarantee that the right price or stock is visible in the right context. Recovery must be targeted, verifiable and organised to avoid duplicates.
Choose an example and date the problem
Record a product identifier, the expected value, the observed value and the relevant contexts: store, currency, customer group or stock source, as applicable. Note the time and time zone used. Comparing screens refreshed at different times may suggest an inconsistency that has already been resolved.
Then identify the last known successful operation and the incident's scope: all products, a family, one store or only new products. A recent update, permission change or format change provides a lead to check without automatically identifying the cause.
Follow the data at every step between systems
Draw the actual path: ERP, file or API, intermediate processing, connector, then Magento. Find the same identifier at each stage and compare what was sent, transformed and received. Run or message identifiers let you match the records for a single operation.
Check field meanings: decimal separators, units, identifiers with leading zeroes, empty values and null values may be handled differently. An export filter can also exclude a product before it reaches the connector. The aim is to locate the first discrepancy while preserving useful evidence for analysis.
Read the business outcome, not just the request status
Examine returned errors and batch details: an overall operation may contain rejected lines. An authorisation error, unknown identifier or invalid field calls for a different fix. Records shared for diagnosis must exclude tokens, passwords and unnecessary personal data.
If your data flow uses Commerce asynchronous APIs, accepting a request and executing it are two separate steps. The documentation describes queuing followed by processing by a consumer. Keep the identifier provided and check the operation's outcome; not every ERP connector uses this architecture.
Check the scheduled tasks and queues actually used
A process that no longer starts, an execution that fails and a waiting batch lead to different diagnoses. Look at the last successful run, unusual durations and items still to process. For a file-based exchange, also check file generation, delivery and processing.
When the data flow relies on Magento message queues, consumers must run through the mechanism provided for the installation, such as cron or a process manager. Adobe documents these options. Identify the actual setup before restarting anything, to avoid creating concurrent processing or hiding the original error.
Compare prices and stock in the right Magento context
An imported price and the price shown to a visitor may differ according to the store, catalogue rules, taxes or customer context. Compare values that answer the same question. Also check whether another synchronisation or local correction rewrites the data after import.
For stock, distinguish quantities held in sources from the salable quantity in the relevant channel. Inventory Management connects these concepts to stock and source configuration. A direct comparison with the ERP's physical stock may therefore miss part of the behaviour; checkout and availability rules must be examined.
Check indexing and display after receipt
If the correct data is in Magento but not on the website, the investigation moves to how it becomes available: store context, indexer status and relevant cache layers. Adobe documents how to view indexer status and mode; scheduled operation depends on the tasks configured for the installation.
The following table helps guide the analysis. These are diagnostic hypotheses, not causes automatically inferred from a symptom. Reindexing or purging everything does not replace identifying the failing stage.
| Observation | Items to compare | Lead to examine |
|---|---|---|
| The data is not exported | Value in the ERP and export contents | Product selection or batch generation |
| The request is accepted but has no effect | Operation identifier and final result | Queue or asynchronous processing, if used by the data flow |
| The data is correct in the admin panel | The same store, product and sales context | Rules, indexing or cache, as applicable |
| The correction later disappears | Write history and timeline | Another data flow or overwrite rule |
Prepare targeted recovery and verify its effect
Before replaying a batch, establish what has already been applied and whether processing can recognise a repeated operation. This is essential for data flows that also create orders or movements. Define the items to reprocess, the intervention window and expected checks with the owners of both systems.
Validate the correction on a representative sample in a suitable environment before widening recovery. The diagnosis should leave a useful record: established cause, corrected items, remaining uncertainties and how to detect another failure. Recovery is then assessed through data and actual use, not merely an error-free log.
- Tools and direction of the affected exchange
- An example identifier, expected value and observed value
- Last successful run and first known anomalies
- The team responsible for the connector and people able to grant access once the work is scoped
Sources and references
Documentation consulted when writing this guide. Advice and procedures should be adapted to your own context.
