Google has announced the Universal Cart, a shared shopping cart that allows orders from different retailers to be bundled together.Instead of clicking through product pages, shopping carts and checkout in several shops one after another, a cross-shop purchasing process is created in one central place.For retailers, this can become an additional sales channel – though it comes with the question of how much control over customer relationships and presentation they give up in the process.

The function gains particular weight in connection with so-called agentic shopping: this refers to software agents – AI systems that independently work through multi-step tasks – searching for products, comparing them and placing orders on behalf of users.How mature this model will become in everyday life is open.The direction is nevertheless relevant for shop operators, because the technical requirements are already shifting before such systems achieve a breakthrough.

From frontend to data source

For years, shop optimisation revolved around what people see: product images, layout, loading time, conversion paths.If orders increasingly run through agents or cross-shop carts, a second layer appears alongside this interface.There, it is not the design but the quality of the structured data that determines whether a product is even included in the selection.

Structured data is machine-readable information about a product – such as price, availability, variants, shipping costs, delivery time or return conditions – in a format that software can evaluate without interpreting the page presentation.Gaps or contradictions often go unnoticed by human visitors because the context compensates for them.An automated system does not have this context.

What shop operators should look at now

  • Completeness of product attributes: Is there clean information for every item on variant, size, material, condition and identifiers – or is decisive information only in the running text of the description?
  • Currency of price and availability: The further the ordering process moves away from your own frontend, the more expensive outdated stock becomes.Discrepancies between feed, shop and ERP lead to abandoned purchases or cancellations.
  • Consistency across all channels: Shop, product data feed and marketplace connections should be fed from the same source.Maintaining several data sets in parallel is hardly possible to keep synchronised in the long run.
  • Interfaces instead of forms: Processes such as order creation, status queries or return notifications should not be representable exclusively via forms in the browser, but also via an API – that is, a programmatic interface.
  • Robust rules for shipping and returns: Conditions that today appear as prose on an information page must be available as an evaluable rule if they are to flow into an automated purchase.

The price of reach

A cross-retailer shopping cart shifts the place where the purchase decision is made.Anyone listed there potentially reaches customers who would never have visited their own shop.At the same time, the decisive comparison takes place outside their own interface – where brand management, cross-selling and advisory elements have only limited effect.

This intensifies a dependency familiar from the marketplace business: reach versus control.The decision for or against it cannot be made across the board; it depends on product range, margin, brand strength and competitive environment.It makes sense to conduct this weighing deliberately, rather than deciding it through inaction or unreflective participation.

Data quality pays off regardless of the channel

The good news for operators: the work that agentic shopping requires is not a bet on a single platform.Clean product data, unambiguous attributes and reliable interfaces also improve classic channels – shop search, filters, price comparisons, marketplaces and the maintenance of the product range in day-to-day business.

In practice, this means: product information should be maintained centrally and distributed from there to all output channels, not the other way round.Where data currently sits side by side in spreadsheets, shop forms and export files, it is worth clarifying the source of truth first before connecting new channels.

A pragmatic first step

A sober stocktake is enough to begin with: which attributes are complete for how many items?How often do price or availability deviate between systems?Which business processes could be triggered without human input – and which not?This produces a list of concrete gaps that can be closed step by step, without requiring a decision on future platform strategies.

Anyone who postpones the topic for too long risks having to make improvements in several places at once under time pressure later – to the data model, feeds and interfaces, often in parallel with ongoing business.

Classification for web and software projects

For web projects, the focus shifts from the pure interface to the data architecture behind it: product information, interfaces and process logic become the actual product, the frontend one of several output forms.Anyone planning new shop or portal projects should include machine-readable output and API access as a requirement from the outset, not as a later addition.

Existing systems can usually be brought into line with this gradually – provided that data maintenance is bundled in one place and not spread across multiple isolated solutions.

Sources