Google Shopping Agency Audit Checklist

What a useful agency audit should answer

A Google Shopping agency audit should answer three questions: can the performance data be trusted, is the full commercially important catalogue eligible and accurately represented, and does the account allocate spend according to business value? A deck of screenshots does not answer them.

Ask for an evidence register and a prioritised action plan. Every finding should identify the affected item IDs or setting, quantify the impact where possible, name the owner and state how completion will be verified. “Improve the feed” is not an audit finding.

Scope and access before the audit

The auditor needs read access to Google Ads, Merchant Center, analytics and tag management, plus catalogue context from the ecommerce team. Backend net sales, gross margin and returns data are highly useful even if supplied as an anonymised export. Record countries, marketing methods, campaign objectives, attribution settings and any restrictions on changing the feed.

Take a dated snapshot of campaign settings, product status and conversion actions. Product approval and campaign delivery can move during the audit; without a snapshot, reviewers end up debating two different account states.

Gate 1: measurement integrity

Stop the commercial performance audit if purchase values are materially wrong. You can still inspect eligibility and implementation, but recommendations based on ROAS must be labelled provisional.

Gate 2: Merchant Center eligibility

In Products → Needs attention, review account issues and product issues by country and marketing method. Google’s issues documentation distinguishes warnings, product disapprovals and account-level enforcement. Export affected products and join them to revenue, margin and strategic priority.

The audit should report weighted coverage. “96% approved” is misleading if the missing 4% contains the summer launch range. Include:

• eligible products as a percentage of active catalogue and recent revenue;
• issues by root cause, country and source;
• time since the last successful source update;
• price and availability consistency across feed, page, structured data and checkout;
• shipping, returns, business-information and policy setup.

Gate 3: product-data fitness

Sample the highest-spend items, best sellers, zero-impression items and one variant family from each major category. Compare the submission with the official product data specification. Requirements vary by product and country, so an audit must distinguish “required”, “required in this circumstance” and “recommended for better data”.

Check stable IDs; accurate titles and descriptions; primary and additional images; price and sale timing; availability; brand, GTIN and MPN; Google product category; product type; variant grouping and variant attributes; shipping; and custom labels. Do not score fields equally. A false price is an eligibility and trust risk; a weak product type is usually a reporting and relevance limitation.

A finding should look like this: “312 women’s footwear variants use a parent title without colour; 84% of their impressions are concentrated in 27 IDs. Add gender, product type, material/feature, colour and size in that order, beginning with those 27 IDs. Measure item CTR and qualified sessions for 28 days.”

Gate 4: inventory allocation

Map every active product to the campaigns and listing or product groups that can serve it. Identify unintended overlap, excluded “Everything else” groups, products in no active group, and products whose economics differ from the target governing them. Google’s guide to Shopping product groups is the implementation reference for Standard Shopping.

Then quantify concentration: percentage of spend and revenue in the top 10, 50 and 100 items; spend on items without a purchase beyond the chosen decision threshold; and products receiving no impressions. A highly concentrated account is not automatically broken, but the agency should explain whether concentration reflects demand, deliberate priority or uncontrolled automation.

Gate 5: commercial quality

Replace blended ROAS with product-group economics. For each meaningful group, calculate net sales after returns, contribution before ads, ad cost and contribution after ads. If complete cost data is unavailable, state the limitation and use margin bands rather than pretending revenue is profit.

Hypothetical example: Campaign A returns 500% ROAS on £10,000 revenue and £2,000 spend. After VAT, product cost, fulfilment and expected returns, it contributes £1,200 after ads. Campaign B returns only 350% ROAS on £14,000 revenue and £4,000 spend, but stronger margins leave £2,300 contribution. The audit should not label A the better campaign without explaining the business objective.

Audit the post-click path

Use the product report to select pages with high cost, high click volume and abnormal funnel rates. Test the exact advertised variant on mobile. Check image and title consistency, visible price, stock, delivery date and cost, return terms, variant controls, reviews, add-to-basket behaviour, checkout errors and page speed.

Google’s landing-page requirements are the compliance baseline. Conversion review goes further by asking whether the page resolves the hesitation created by the product and query. Separate policy failure, merchandising weakness and checkout friction in the report.

Score findings by impact, confidence and effort

Use a transparent scoring model. One simple method is impact from 1–5, evidence confidence from 1–5 and effort from 1–5; rank by impact × confidence ÷ effort, then override only where policy or measurement risk demands it. Keep the raw components visible so stakeholders can challenge the assumptions.

The deliverables to expect

A credible audit produces a one-page diagnosis, evidence workbook, issue register, product list, measurement caveats, and 30-day implementation plan. It also names what should not be changed yet. Recommendations should link to the relevant Google requirement where compliance is involved and identify any decision that needs finance, merchandising or development input.

At the end of 30 days, rerun the baseline: eligible revenue coverage, product-level waste, conversion integrity, category contribution and the metrics attached to each implemented fix. An audit is complete when decisions can be verified, not when the presentation is delivered.

Sample evidence register

The register is the bridge between findings and implementation. A practical row contains: finding ID, source screen or export, date range, affected item IDs, commercial exposure, confidence, recommended action, owner, due date and verification metric.

Attach an affected-product file rather than pasting ten examples into slides. The implementation team needs the complete set and the logic used to generate it.

Audit failure cases

A settings-only audit misses catalogue and website causes. An issue-count audit treats a warning on a dormant accessory as equal to a disapproval across a launch range. A best-practice score assigns points without showing whether the practice matters to this account. A 90-day average can hide a feed failure that began three days ago.

Another failure is recommending simultaneous feed, structure, target and landing-page changes. The account may improve, but nobody learns why. Sequence fixes so measurement and eligibility are stable before optimisation tests. Keep unresolved dependencies visible: an agency cannot repair an ERP stock delay solely from Google Ads.

Acceptance criteria for the first 30 days

Convert the audit into observable completion rules. “Fix tracking” becomes “one test order creates one primary purchase with correct value, currency and unique transaction ID”. “Improve Merchant Center” becomes “all priority item IDs are eligible in the intended country and destination, with no recurrence after the next source update”.

For a feed cohort, retain old and new processed values and define the metric window. For a structure change, preserve original cohort labels so pre/post contribution remains comparable. For a CRO issue, specify the funnel event, device and product template. The 30-day review should mark each action accepted, rejected or blocked with evidence.

The deeper procedures in the Shopping feed checklist and product-page CRO guide are useful when those areas become implementation workstreams.

An independent reviewer should be able to recreate the priority order from the evidence register. If priorities depend on verbal context that never reached the deliverable, the audit is not ready to hand to an implementation team.

Sources

• Google Merchant Center: Issues in Merchant Center
• Google Merchant Center: Product data specification
• Google Merchant Center: Landing-page requirements
• Google Ads: Product reporting