Case study · Scout
Scout: refreshing supplier prices and stock without overwriting enriched product pages
On Scout, property organisations order from suppliers whose catalogues change with each of their exports. Each new file has to update prices and availability without erasing what has been completed on the platform. Building on Medusa and Mercur, we created an update workflow that compares incoming data with existing product pages. It lets users choose which changes to keep and limits what each supplier feed can modify.
Scout is a B2B procurement platform for property organisations that manage properties made up of flats and units. These organisations find product suppliers and service providers there, and order from them in a single space. For their part, suppliers have a vendor space where they maintain their catalogue, prices and stock. This data rarely comes from entering product pages one by one. It comes from each supplier’s exports, and these exports differ in their columns, formats and naming habits. Nor is a product page published on Scout a mere copy of the supplier’s line: a description, images or a category may have been added to it. Applying a new file indiscriminately would have overwritten that work. It therefore had to be possible to take in what the file brings that is new while keeping what needed to stay.
A mapping saved once for each supplier format
The import guides the supplier’s team through successive steps. They first upload their file, then map its columns to the information Scout expects and convert the values that need converting. Then come row validation, a product categorisation step, and matching against the catalogue already in place. Breaking the process down this way makes it possible to understand what is read, what is transformed and what will be matched, instead of discovering the result once the file has been fully applied.
A supplier generally sends exports built on the same template. It would be tedious to redo the whole column mapping with each delivery. The supplier can therefore save this configuration in their account and reuse it for the next import. If their own export changes, for example after a column is added or renamed, they correct the existing configuration without starting from scratch. When a configuration no longer matches any real file, they can delete it so that it is not reused by mistake.
Receiving irregular exports and putting them back into shape
Some files do not lend themselves directly to import. When an attachment received from a supplier is malformed, Scout’s administration can clean it up to make it usable. The corrected file keeps a record of its origin: it is known which submission it came from.
The project documentation also describes mechanisms developed for particular suppliers. In one of them, catalogue, stock and price exports sent by email were routed to Scout’s import. Another converted a supplier’s irregular format before it was read, so that its product references would not be altered. A third turned another supplier’s text file into a spreadsheet, then sent it to the right import pipeline. These mechanisms are documented along with the results obtained when they were developed. They show how a stubborn format can be absorbed upstream. They do not, however, guarantee that such a connection is active today for every supplier.
Seeing what the file would change before accepting it
Before confirming an update, the supplier can compare an imported line with the existing product page. The comparison covers the product information, its price and its images. They therefore do not have to approve a file as a whole without knowing what it contains: they can see, product by product, what would be replaced.
This comparison makes it possible to spot expected and unexpected differences. A planned price increase is expected. A description shortened by the export, or an image replaced with one from another catalogue, is less so. The decision is made at that point, before publication, rather than after the fact, when the previous version of a product page would have to be tracked down.
Keeping certain changes and preserving the rest
The supplier can then confirm the update for all lines or for a selection of them. Differences they do not keep are not applied: the data in place on Scout is preserved. The same file can thus be used to refresh prices while leaving untouched the product pages that should not be changed.
By way of illustration, consider a fictitious example. A tap and valve supplier sends its new export. Prices have changed on part of the range, but some lines in the file carry more basic descriptions than those already written on the platform. After comparison, the supplier confirms the lines whose changes they approve and sets the others aside, and those product pages stay as they are. If the administration has restricted this supplier’s updates to prices, the descriptions in the file are not taken in anyway.
Once the choices have been made, the supplier closes the review and obtains final counts of what has been processed. The file is then marked as completed and does not remain in an intermediate state.
Setting, supplier by supplier, the information a feed can modify
Line-by-line selection addresses a specific file. A more durable rule is added to it. Scout’s administration can define, for each supplier, the information their updates are allowed to modify. On confirmation, only that information is taken from the file. A feed intended to refresh prices therefore cannot rewrite the rest of the product page.
The two levels complement each other. The rule set for the supplier applies from one import to the next: information it does not open to the feed is not taken from the file, even if the file contains unexpected columns. The review of each import then makes it possible to decide, within that framework, which changes to publish. The platform thus retains control over the product pages presented to property organisations, and suppliers can update what is theirs to update.
New products and the right accounts
An update file may contain references that Scout does not yet know. These new products appear as pending import, and the administration can view them to check the arrival of unknown references.
Other checks concern which account the data is attached to. An import is rejected if the supplier does not have their own stock location: imported quantities cannot feed into another account’s. The uploaded file must also belong to the supplier who starts the processing. On a platform that brings together many vendors, these checks prevent a delivery from changing another supplier’s catalogue or availability.
Large imports that can be tracked and resumed
A full catalogue can contain a great many lines. Processing therefore runs in the background, without the supplier having to keep a page open until it finishes. They find their batches sorted by type and status, with their progress. Long-running jobs, including those that use AI, thus remain visible.
If an import is interrupted, the supplier resumes it from what has already been loaded, instead of starting all over again. They can also cancel a creation batch launched by mistake as long as it can still be cancelled. When some lines fail, they retrieve only the products concerned to correct and rerun them, without resending the entire file.
On Scout, a new supplier export is not handled on an all-or-nothing basis. The supplier reuses the configuration for their format, compares what the file would change and confirms only what they keep. The administration sets, supplier by supplier, the information a feed can modify. Prices and availability can thus be refreshed without erasing the work done on the platform’s product pages.