The visible sheet, the product markup and the catalog feed must describe the same product at the same time. To avoid price or stock discrepancies, identify a truth source, synchronization times and controls to be executed after each update.
Map the path of a data item
Simply draw the price path from commercial administration to the site and exports. Add caches, scheduled tasks and possible transformations. Do the same exercise for availability. A discrepancy can come from a variant calculation, an old export or a cached page. Without this card, there is a risk of manually correcting a display that will become false again at the next import.
Compare representative references
Select a simple product, a variant, a promotional item and an unavailable item. Note the visible information and structured values, then compare it to the export. Control Currency, ID, URL, Image, and Availability status. Google presents structured data and Merchant Center as two additional ways to share the catalog. The consistency between these media deserves a dedicated verification.
Follow a reference from one medium to another
Give each tested reference a stable identifier in your acceptance test sheet. This is not necessarily the wording displayed to the customer: a commercial title can change without the product changing. For each medium, note the time of the reading and the origin of the data. You will then be able to distinguish a different commercial rule from a simple copy that has not been updated.
| Support | Values to be noted | point of alert |
|---|---|---|
| Reference Catalog | ID, price, currency, availability | Approved version and date of change |
| Visible sheet | Displayed values and selected variant | Cache or wrong product open |
| Structured data | product, offer and values actually issued | transformation different from visible content |
| Exported flow | ID, URL and transmitted values | Latest generation and successful import |
| Demo basket | Conditions calculated for this case | Destination and combination of products |
It is not a question of making all the calculations identical in all contexts. A destination, quantity or option can legitimately modify a condition. You have to be able to explain this difference and check that it is announced in the right place. An unexplained divergence on the same reference, in the same context and at the same time, on the other hand deserves an investigation.
Plan for updates and failures
Define what happens if an import fails or a task fails. A tracking table should indicate the last successful synchronization and the person to be notified. Avoid considering the absence of visible error as proof of freshness. Plan a check after massive price changes and after product template changes.
Fictitious example: a promotion completed
The normal price returns to the shop, but an old promotional value remains in the export. The team goes back to the task that has not been relaunched and corrects the process. It then checks some products concerned in all media. Modifying only the text of the form would have left the problem intact and created a new divergence in the next cycle.
Locate the cause of a deviation before retouching
- Reproduce the gap on a reference and keep the observed values. Check that you are comparing the same variant and the same conditions.
- Go back up the data path: Record, Transform, Publish, Cache, Export and then Recipient Channel. Note the last step where the value is correct.
- Check logs or confirmations of scheduled tasks. A configured task is not necessarily a successfully executed task.
- Correct the identified cause, then check a relevant reference and a control reference. Also check the next planned cycle, in order to detect a manual correction that would be overwritten.
In a fictional simulation, a promotion ends in the catalog, but an export retains the old price. If the commercial source is already correct, changing the price in the administration again may be useless. The right decision can be about the regeneration of the flow, its transmission or the management of the cache. Document what has been changed and how the team will be informed if the next cycle fails.
Google explains the Structured product data and Catalog Sharing. These mechanisms do not replace your consistency check. For a merchant site, plan this acceptance test in the handover to the team, with the Delivery and return information and update responsibilities.
Document a short acceptance test
Keep a list of test references, the fields to compare and the right frequency for your catalog. After a correction, check the actual broadcast data rather than the only configuration. Do not invent stock to get a rich display. This work completes the Clarity of sheets for purchase assistants And also reduces unpleasant surprises for customers.