Shopping Ads Agency for Ecommerce Brands: The Operating Model That Works

Last updated: 11 July 2026

Shopping is a shared operating system

The strongest Shopping agency relationship is not “client sends budget, agency sends ROAS.” It is a working system between trading, merchandising, feed or development, finance, creative and paid media. Each team holds information Google needs indirectly: product accuracy, stock, promotions, margin, returns and customer value.

The agency’s role is to turn those commercial inputs into product-data, campaign and budget decisions, then show the business what it learned. It should not pretend to own prices, inventory or claims that belong elsewhere.

Inputs the agency needs

Segment products for decisions

Campaign segmentation should correspond to a decision the team might make. Grouping by category is useful when categories need separate budgets or targets. Margin labels are useful when reliable economics differ. Stock labels help when scarcity or overstock should alter spend. A “best sellers” label is useful only if its definition and refresh cadence are known.

Avoid building dozens of campaigns merely for reporting. Google’s product reports can provide SKU and category evidence across campaign types. Structure earns its complexity when it changes control, budget, creative, geography or goal.

An example: revenue winner, profit loser

Imagine a discounted category reporting £50,000 revenue from £10,000 media. A 500% ROAS looks strong. After product cost, returns, fulfilment and the discount, contribution may be too low to support that spend. Another category at 350% ROAS can produce more contribution and healthier repeat customers.

The agency should flag this conflict, but finance must provide usable values and trading must approve reallocation. Sending margin-adjusted values to Google can help when data quality and volume support it; otherwise margin can still guide budgets and analysis outside bidding.

The weekly operating rhythm

A useful weekly review is brief and exception-led: tracking or spend anomalies, Merchant Center issues, products losing visibility, stock or promotion changes, tests reaching a decision and actions required from client teams. It should not be a narrated dashboard.

Monthly or quarterly reviews handle larger choices: category investment, campaign architecture, value logic, creative themes and landing-page priorities. Record decisions and owners. Without a change log, teams repeatedly debate the same issues.

Feed changes need governance

Title, description, product type, category, image and label rules should have an owner, version and rollback route. If the agency uses a supplemental feed or tool, document precedence over source data. If source data changes, consider SEO, marketplaces and site merchandising.

Test patterns across a defined product set. Do not mix title, image, price and campaign changes in the same week and then claim to know which caused movement.

Merchant Center incident response

Agree who receives account and item alerts, who can edit the site or feed, who gathers policy evidence and who requests review. Google distinguishes warnings, limited products, disapprovals and account-level issues; their urgency and impact differ.

Do not request reviews repeatedly before fixes are complete. Cooldown periods can delay recovery. The agency should maintain an incident log that links the issue, affected products, root cause, fix and result.

What the agency report should answer

• Which products or categories should receive more budget?
• Which high-spend products need feed or page work?
• Which products are ineligible or limited?
• Where is brand demand inflating acquisition performance?
• How do stock and margin change the recommendation?
• Which test produced a decision?
• What must ecommerce, finance or development do next?

Hiring test

Ask finalists to run a working session on one category. The proposed operator should examine product data, Merchant Center status, economics, campaign overlap and page quality, then prioritise actions by expected impact and implementation effort.

The best answer may include “do not change this yet.” Restraint indicates the team can protect a useful baseline instead of rebuilding for visibility.

Procurement evidence to request

First 90 days of the working relationship

Days one to 15 should establish administrative ownership, conversion accuracy, product status, feed method, campaign overlap and the retailer’s category economics. The team should identify what it cannot yet know and assign data owners.

Days 16 to 45 should address high-impact eligibility and tracking problems, create a small number of controlled feed or campaign tests and establish the weekly exception review. Avoid simultaneous bulk changes that erase the baseline.

Days 46 to 90 should produce decisions: products to scale, isolate, repair or stop; early test results; a feed-rule register; and the next category or creative priorities. The retailer should know which internal bottlenecks now limit growth.

Measurement boundaries the agency must state

Reported order revenue can overvalue discounted, returned or low-margin products. New-customer settings depend on customer identification quality. Attribution allocates credit but does not prove the sale would not have happened. Product reports explain where spend and conversions occurred, not the full effect of media.

Use each measure for a defined decision. Platform values guide bidding, finance contribution checks profitability, cohort data checks customer quality and controlled tests estimate incremental effect. Do not force one number to answer all four.

Decision scenarios for ecommerce leaders

Trading wants to clear overstock

Create a time-bound stock label and budget plan, but protect contribution and end the rule when inventory normalises.

Finance wants a higher ROAS target

Model likely volume loss and category effects. A higher target can improve reported efficiency while reducing profitable total contribution.

Brand wants new PMax assets

Define the audience problem, hook, proof and product set. Asset quantity without a creative hypothesis creates production activity, not learning.

What ecommerce leaders should retain internally

Keep a named commercial owner for the channel, even when the agency manages every campaign. That owner supplies margin and stock context, resolves cross-team blockers and decides whether the objective is revenue, contribution, customer acquisition or inventory movement. Finance should validate value logic; merchandising should validate product truth; development should own site reliability. The agency should make better media and feed decisions from those inputs, not become the unofficial owner of business facts it cannot verify.

The commercial owner should also control priority conflicts. A category can deserve less spend because stock is scarce even when ROAS is excellent, or more exposure because the business needs new-customer penetration despite a lower first-order return. Write these temporary objectives down with an expiry date so an exceptional trading decision does not become a permanent bidding rule.

Agency performance also depends on response speed from the brand. Track blocked actions and their commercial impact: delayed price fixes, missing creative, unapproved feed rules or unresolved checkout errors. This is not a device for shifting blame. It helps both teams distinguish media limitations from organisational delays and invest in the correct bottleneck.

Sources

• Google product reporting
• Google Merchant Center issues
• Google value-based bidding