A buyer sends a colleague a link to a sand-coloured apron in size large. The colleague opens it and sees a navy apron in medium. Both people reached a working page, but the link lost the choice that mattered. Before accepting a new store, ask the developer to demonstrate this journey with a product from your own catalogue.
Choose a product that can expose the mistake
Choose one family with at least two colours and two sizes. Ask the catalogue owner for the actual combinations offered, the identifier of each sellable item, the approved photos and the price and availability at a recorded time. Include a combination the business never makes or one that is temporarily unavailable. Those cases tell you whether an attractive colour selector can lead someone to order the wrong item.
Fictional example. A kitchen supplier sells medium navy, large navy and large sand aprons. There is no medium sand apron. Large sand is available for this test and large navy is temporarily sold out with backorders disabled. Give that description to the reviewer before the demo. Keep the expected outcome independent of what the new store happens to display.
Start at the shop entrance, then follow a shared link
- From the store menu, browse to the workwear category and open the apron. Record the category and product links. Google's ecommerce navigation guidance recommends linked paths to products. A search box alone can leave items undiscovered by a crawler.
- Select large sand. Copy the variant link supplied by the store and give it to a colleague using a fresh browser session. Ask what colour, size, price and availability they see before they touch any controls. Compare their answer and the main photo against the catalogue.
- Have the colleague add one to a test basket and read out the item identifier or variant description. Then repeat with sold-out large navy and the nonexistent medium sand combination. Check each against the agreed selling policy. The test ends at the basket and requires no payment.
A variant needs an address that can identify it independently of a previous visitor's choices. In Google's URL guidance, a path or a query parameter can distinguish a choice. A change confined to the part after the hash sign cannot distinguish indexed content for Google. In the apron example, opening a link with a colour but no size is insufficient if it silently selects a different sellable item.
Agree which URL represents the product family
Ask the developer to name the store's design. If all choices belong to a single product page, the family has a preferred address, also called its canonical URL. A link may still open a particular colour and size on that page. A store that distributes variants across several separate pages has a different design. Record which URLs the team intends to make independently indexable and how they relate to each other. Google's product variant examples cover both designs.
Where the scope includes product structured data, request one sample from the rendered page. Find the family identifier and the identifiers for the options described there. Match the offer, currency and availability for the selected item to the visible page and catalogue. Ask the developer to show the result of Google's Rich Results Test and any unresolved issues. A passing markup test alone cannot tell you that the basket contains the correct apron. See Google's introduction to structured data testing.
Decide whether the catalogue is ready
For each failure, keep the starting URL, expected item, actual item, device, language and observation time. If only one translated colour name is wrong, the catalogue editor can investigate. If every shared link resets to medium, send the examples to the developer responsible for the selector. After correction, repeat the shared-link test and check a second product family using the same template.
For a bilingual shop, add these cases to the website acceptance checklist. To include product discovery in an ecommerce website brief, send Lava one representative product family and the decisions your shopper needs to make.
