Why a shopping feeds operating workflow beats firefighting

What to take away

Name the feed owner

Every feed needs one person who can answer for it. Not a team, not a shared inbox. A name on a rota, visible to merchandising, finance and the agency.

That owner approves changes to titles, prices and availability. They also hold the rollback decision when a release misbehaves. Without this, feed problems circulate between departments until someone senior notices revenue slipping.

Write the ownership map into your wider shopping feeds operations and delivery guide for 2027 so new starters inherit it rather than inventing their own.

Set the weekly release rhythm

A predictable cadence beats continuous edits. Pick one release window, freeze the catalogue the day before, and publish after checks pass.

A workable pattern for a mid-sized retailer: Monday freeze, Tuesday validation, Wednesday release, Thursday review of disapprovals, Friday fix list. Adjust to your trading pattern, but keep the sequence identical each week.

Outside the window, only price and stock corrections ship. Everything else waits. This single rule stops the slow drift that makes feed quality impossible to audit.

Validate attributes before upload

Validation is a gate, not a report. Products that fail the gate do not reach the channel.

Start with identifiers. Google's specification for the [id] attribute states that each item needs a unique, stable value, and warns against reusing it for a different product. Reused identifiers break history, reporting and remarketing lists.

Then check the commercial fields: price, availability, condition, brand, GTIN where you hold one. Run the same checks on every channel, even where the rules differ slightly, so one source of truth survives.

Budget your crawl

Large catalogues can be crawled harder than a site can absorb. Google's guidance on how to reduce Google crawl rate sets out the settings available when fetching needs slowing down.

Treat that as a capacity question. If product pages are slow to respond during peak, the feed refresh schedule is part of the cause, not separate from it.

Agree a maximum fetch rate with whoever runs the site, and revisit it before every seasonal push. Document the agreed figure so nobody has to guess in December.

Handle cross-border lines

Selling outside Great Britain brings customs, tax and consumer rules into the feed workflow. GOV.UK's pages on buying and selling outside Great Britain cover the starting points for exports and imports.

Feeds for those markets need extra fields: country of origin, delivery estimates, returns terms. Add them to the validation gate rather than patching them per order.

Where a rule differs in Scotland, Wales or Northern Ireland, note the difference against the specific market in the feed documentation. Blanket assumptions cause refunds.

Before and after

Measure the workflow

Track four numbers weekly: items rejected at the gate, disapprovals after release, time from change request to live, and rollbacks. Each points at a different weakness.

Rising gate rejections suggest upstream catalogue data is drifting. Rising post-release disapprovals suggest the gate is too loose. Slow change requests mean the release window is too narrow for demand.

Publish the numbers where merchandising can see them. A workflow nobody measures becomes a workflow nobody follows, and the firefighting returns within a quarter. Pair the numbers with the checks set out in shopping feeds service standards in England so reviews have a fixed reference point.

Common questions

Who should own the feed day to day?

One named person with authority to approve or block a release. They can delegate tasks, but accountability for the feed stays with them, and the rota is visible to other departments.

How often should a feed be released?

Once a week suits most mid-sized catalogues, with price and stock corrections allowed outside the window. High-turnover ranges may need twice weekly, but keep the sequence and checks identical.

What belongs in the pre-upload gate?

Unique identifiers, price, availability, condition, brand and any market-specific fields such as origin or delivery estimates. Items failing the gate stay out of the channel until fixed.

How do we know the workflow is working?

Gate rejections, post-release disapprovals, change lead time and rollback count, reviewed weekly. Falling disapprovals with stable lead times is the signal to leave the process alone.