What shopping feeds measurement means for multichannel reporting

What to take away

Why shopping feeds measurement changed recently

Google Ads reports web and app conversions together in a single value model. Its guidance on getting the full value from your web and app channels explains this; when reporting shopping feeds, state which Google Ads conversion actions and values you include.

A feed's contribution can no longer be read from a website-only conversion count. If your app converts a customer after a feed click, that value sits elsewhere in the account.

Meanwhile the ICO's online tracking guidance covers cookies and analytics. Under the Privacy and Electronic Communications Regulations 2003 (PECR), non-essential storage and access technologies generally require consent. Google Consent Mode v2 can pass consent choices to Google tags, but it does not obtain consent or make tracking compliant by itself. Feed measurement built on server-side events is still tracking: the rules apply to the data you collect, not the method you use to collect it.

On the finance side, Making Tax Digital for VAT requires digital record-keeping that links to returns. A feed report that cannot be reconciled to those records creates a gap between marketing and finance that someone will eventually have to explain.

Finally, the Consumer Contracts (Information, Cancellation and Additional Charges) Regulations 2013 require information to be provided to consumers and give cancellation rights for many distance contracts. Reflect cancellations and refunds in your order and return definitions; the regulations do not themselves redefine a platform's conversion event.

What should a shopping feed report actually contain?

A feed report needs four layers: spend, feed quality, order outcomes, and margin. For Google Shopping, Google Merchant Center is the place to inspect item-level status: track disapprovals, warnings, and price or availability mismatches, as well as whether products have the identifiers and attributes the feed requires, such as GTIN where applicable. Most teams stop at feed quality because it is the easiest thing to export. That is why reports can look healthy while the business loses money.

If you only report the first two layers, you are describing activity rather than results. A practical fix is to define a small set of report fields and refuse to add more until each one has an owner.

Map each report field to its source: use Google Ads for campaign spend and platform-attributed conversions, Google Merchant Center for item status and feed diagnostics, GA4 for configured web and app events and analytics attribution, and finance or order records for sales, refunds, costs and margin. Keep each source's definition visible rather than treating its figures as interchangeable.

Which metrics belong in the weekly report?

Weekly reporting should answer one question: did the feed produce profitable orders this week? That means revenue, order count, average order value, return rate, and contribution margin. Contribution margin is revenue minus product cost, payment fees, delivery and feed spend, divided by revenue. Feed health metrics belong in the same report but as context, not as the headline.

Monthly reporting can carry the slower measures: impression share trends, category-level performance, and the cost of feed errors. Quarterly reporting is where you reassess attribution and channel mix.

One caution. A metric that nobody acts on is a cost, not a control. If your weekly report has twenty columns and no decisions attached, cut it to six.

How should you attribute orders across channels?

Attribution is where feed measurement gets contested. The platform will claim a conversion. Your analytics will claim a different one. Your warehouse will claim a third. All three can be defensible if you state the model.

The most common models are last click, first click, and data-driven. Last click over-credits the final touchpoint, which is often a brand search rather than the feed. First click over-credits discovery. Data-driven models need enough volume to be stable, which small catalogues rarely have.

Most UK retailers can report two models side by side: platform-reported and a blended model. The blended model allocates a share of revenue to feeds based on assisted conversions. Blended attribution share is feed-attributed revenue divided by total revenue across all channels.

The gap between them is a useful number, showing how much of your revenue depends on which model you pick.

The trade-offs are: last click suits short consideration, first click suits discovery, and data-driven attribution needs volume. For each platform report, record the conversion action, attribution model, configured lookback window and reporting date range; shopping feeds attribution should not be compared across platforms without checking those definitions.

What does the platform change mean for your numbers?

When Google Ads reports web and app together, its platform number is broader than a website-only figure. In GA4, define the purchase event consistently across the web and app data you collect, then compare GA4's attributed conversions with Google Ads' platform-reported conversions without treating the two as interchangeable. If your internal report still uses website conversions alone, you will see a gap that may be a definition mismatch rather than an error. Fix the definition before investigating the gap.

Write down the model you use, the date you adopted it, and the reason. Attribution changes without documentation become arguments.

What compliance rules affect feed measurement?

Three sets of rules touch feed measurement directly. The first is tracking consent. The second is VAT record-keeping. The third is advertising claims.

Server-side tracking does not remove the consent obligation, because the obligation attaches to the personal data, not the transport.

If your feed report and your accounting system disagree on revenue by more than rounding, you have a reconciliation task. A reconciliation tolerance is the maximum difference you accept between the feed report and the finance ledger. A typical tolerance is 1% of monthly feed revenue or £1,000, whichever is greater. Build the bridge once rather than every quarter.

On claims, the ASA rulings database is where you check whether a comparative claim in a feed title or description has been ruled on before. Due diligence here is cheap. A ruling against you is not.

Compliance failures rarely show up as a single lost order. They show up as a reporting restatement, which is worse.

Which measurement approach should you choose?

Use the table below to match your situation to an approach. The avoid column matters as much as the choose column, because most reporting failures come from applying a method designed for a different scale.

How do you know your measurement is good enough?

Good enough means three tests pass. First, the numbers reconcile to finance within an agreed tolerance. Second, the attribution model is written down and unchanged for at least a quarter. Third, a consent refusal does not break the report, it only reduces its coverage.

If all three pass, more measurement will not help. If any fail, adding dashboards will not fix it.

How do you report feed performance to different audiences?

Different audiences need different cuts of the same data. The finance director wants revenue that ties to the accounts. The merchandising lead wants to know which categories are being disapproved. The trading director wants contribution margin by channel.

Build one dataset and three views. Do not build three datasets, because they will drift. The most common drift is between the platform export and the finance ledger, and it is usually caused by refunds being timed differently.

Benchmarks are useful here, but only if they come from a comparable catalogue. A benchmark from a fashion retailer with 50,000 SKUs will mislead a specialist retailer with 800. Choose a comparison set with similar SKU count, category and order value, and label it as a comparison rather than a universal target.

What does a good monthly pack look like?

Anything longer than five pages will not be read by the people who make decisions.

Sample monthly pack (illustrative figures only)

Tolerance: 1% of monthly feed revenue or £1,000, whichever is greater. Here 1% is £1,200, so the tolerance is £1,200. The £900 gap is within it.

What are the most common measurement mistakes?

Mistakes cluster around definitions, not tools. Teams argue about numbers when they are really arguing about what a conversion means.

The first mistake is counting a cancelled order as a sale. The Consumer Contracts Regulations 2013 give consumers cancellation rights for many distance contracts, so distinguish the initial order from the sale after cancellations and refunds are accounted for. If your report counts every order as retained revenue, your return rate looks like a surprise later.

The second mistake is treating platform-attributed revenue as net revenue. Check whether the figure includes cancelled orders or refunds and whether it is stated before or after VAT; it is not automatically the amount that reconciles to finance.

The third is changing attribution mid-quarter without flagging it, which makes trend lines meaningless. The fourth is measuring feed quality without measuring feed cost, which hides the labour spent on manual fixes. The fifth is ignoring app traffic when the platform has already merged it.

The fixes are: distinguish orders from retained sales, report revenue on a clearly stated returns and VAT basis, flag attribution changes, measure feed cost, and include app traffic where the platform reports it.

How do you fix a broken report quickly?

Start with the definition of a sale. Then rebuild the report backwards from that definition. Most broken reports are fixed by this single step, because the definitions were never agreed in the first place.

If you operate across more than one UK nation, do not assume market conditions are identical. Consent rules under UK data-protection and electronic-communications law apply across the UK; segment performance by delivery destination where the underlying data supports it, and keep market comparisons distinct from consent compliance.

For a UK-wide report, label geography consistently, such as by delivery destination, and compare like-for-like periods and product groups. That helps distinguish local delivery or returns patterns from changes in feed performance.

How often should you review your measurement setup?

Use this checklist.

Measurement that is nobody's job becomes everybody's excuse.

Common questions

How do I measure shopping feeds if I sell through an app as well as a website?

Report web and app activity using a clearly defined purchase event: use Google Ads' web-and-app value reporting for its platform conversions and GA4 for the web and app events configured in Analytics. State the attribution model and compare each figure with finance records; do not add platform-attributed totals together as though they were separate sales.

Do I need consent for analytics that measure feed performance?

The ICO's online tracking guidance treats non-essential analytics cookies as requiring consent unless they are strictly necessary. Google Consent Mode v2 can communicate consent choices to Google tags, but it does not gather consent or remove the need to comply with PECR and data-protection requirements. Server-side measurement does not remove that requirement. Design your reporting so that a refusal reduces coverage rather than breaking the report entirely.

What is a realistic attribution model for a small catalogue?

Use platform-reported conversions alongside a simple blended model, and state both in the report. Data-driven models need more volume than a small catalogue usually generates, so they can produce unstable allocations. Two simple models reported honestly beat one complex model reported as fact.

How do feed metrics connect to VAT reporting?

They connect through the underlying order and VAT records. Making Tax Digital for VAT requires businesses within its scope to keep digital records and use compatible software for their VAT return. Reconcile the feed report using order or invoice references, transaction dates, taxable values, VAT amounts and refunds, while keeping platform-attributed revenue distinct from the VAT figures. Build the bridge between marketing and finance data once, then automate it, rather than rebuilding it at each quarter end.