Google Ads and Shopping: ManCave

The challenge

A competitive grooming market, a need for product visibility, and a Shopping account where spend was not readable by margin. The underlying issue was upstream of bidding: Shopping results are matched from product data, not from keywords. If a title is vague, a category is wrong or an attribute is missing, the product simply does not appear against the searches it should win, and no amount of bid management fixes that.

What we did

We started in the feed: titles rewritten to match how people actually search, categorisation corrected, missing attributes completed. Then the campaign structure was rebuilt around profit margin and purchase intent instead of loose product groupings, so that spend could be read against the thing the business cares about.

Data-quality checks were added because Shopping fails quietly, a disapproval removes a product from the auction without announcing itself, and an account can lose its best sellers for a fortnight before anyone notices.

What changed

The account moved from undifferentiated spend to groups that can be read and decided on, with conversion tracking pointed at eCommerce outcomes and a written plan for scaling the groups that justify their spend.

The project in detail

The summary above is what changed. What follows is how: the market this work had to compete in, the decisions taken in full rather than in outline, the order the work actually ran in, and what another retailer in the same position should take from it.

Why Shopping performance starts with the feed

Search campaigns can use marketer-selected keywords. Shopping begins with the product information submitted through Merchant Center. Titles, descriptions, price, availability, images, category, identifiers and variant attributes establish what the item is and whether it can serve.

Google’s current product data specification says it uses submitted data to match products to the right queries, and warns that missing or inaccurate information can cause disapproval, limited eligibility or incorrect display. That makes the feed part of advertising strategy, not a developer export completed once. Review the official Merchant Center product data specification.

A bigger budget can compete more forcefully only in opportunities the product enters. It cannot reliably buy relevance that the data failed to express. Feed work therefore came before bid work: describe products accurately, make variants and identifiers coherent, and ensure the landing page confirms what the listing promises.

What the old structure made impossible

Products with different margins, price points and intent levels shared budget and reporting context. Aggregate performance could describe total activity, but it could not answer which product family deserved more investment, which needed feed repair or which could not support the same acquisition cost.

Catalogue structure is designed for merchandising and navigation. It may group products by storage category, range or administrative convenience. Advertising decisions require economics and search behaviour. If high-margin repeat purchases and lower-margin introductory items compete inside one undifferentiated group, an account manager cannot interpret the same return equally.

The diagnosis was therefore structural. Before optimisation, the team needed stable product IDs, commercial grouping and conversion evidence linked to the items that generated it. Readability was a prerequisite for action.

Rebuilding the product feed first

Titles were rewritten to lead with accurate terms buyers use to identify the item, then add distinguishing attributes where relevant. The objective was clarity, not a string of repeated keywords. Product categories and client-defined product types were reviewed so reporting and platform interpretation had a useful foundation.

Required and relevant attributes were completed consistently: identifiers, brand, condition, price, availability, images and variant fields as applicable. Landing-page values were checked against submitted data because a technically complete feed can still fail when the page shows a different price or stock state.

Stable IDs preserved history. Variants were grouped and distinguished correctly rather than collapsed into ambiguous records. The feed also exposed business labels for margin or intent groups without asking customer-facing titles to carry internal reporting logic.

This sequence matters. Changing campaign structure before repairing the feed would have reorganised incomplete evidence. The rebuild made the catalogue eligible to participate on its merits and gave the later groups consistent inputs.

Data-quality checks that prevent silent disapprovals

Merchant Center diagnostics were monitored for account and item issues, but the process did not wait for a visible crisis. Each refresh checked record count, required-field presence, duplicated IDs, invalid values, broken destinations, price and availability consistency, image access and unexpected product loss.

Products can stop serving because a feed fails, an item is disapproved, a landing page diverges or an expiry and refresh process breaks. A successful upload does not prove complete active inventory. The operating view compared expected catalogue coverage with approved and eligible products, then assigned exceptions.

Changes were documented with source, correction and validation. That history reduced the chance of repairing the export today and allowing the same upstream issue to recreate it tomorrow.

Regrouping campaigns by margin and purchase intent

Products with materially different economics stopped competing for one undifferentiated budget. Groups reflected how the business evaluated products: margin capacity, role in the customer journey, product priority and observable search intent. Catalogue categories remained available for navigation; advertising structure served allocation.

This did not require a unique campaign for every SKU. Excessive fragmentation can leave each group without enough evidence. The team chose the smallest separation that changed a real decision. A group needed an owner, budget logic, product membership rule and a reason its performance would be acted on differently.

Negative relationships and query observations helped reduce obvious mismatch where the campaign type allowed control. Product labels connected feed membership to reporting so items did not drift between groups through manual lists.

The resulting view was decision-ready. It could show which groups were earning their spend, which had a product-data or landing-page problem and which needed more evidence before scaling. Reporting moved from a description of activity to a controlled allocation system.

Aligning tracking with ecommerce outcomes

Clicks show that an ad received attention; they do not show that the selected product justified the cost. Conversion tracking was aligned with ecommerce outcomes so review could connect spend with transactions and product value rather than celebrate traffic alone.

The implementation required consistent event definitions, order identifiers, values, currency and duplicate control. Test orders checked the path from advertisement and landing page through purchase confirmation and platform reporting. Attribution remained a model, not a perfect record of every influence.

Product-level and group reporting then asked commercial questions: which products generated accepted orders, what value was recorded, where did checkout or page experience interrupt intent, and did the economics support additional spend?

Customer service, returns and fulfilment still affect realised value after an advertising conversion. Teams that need operational coverage can use ecommerce customer support; campaign reporting should not pretend the media account controls the whole customer experience.

How it ran

1. Product feed optimization: rebuilt search-aligned titles, categorisation, attributes, identifiers and labels from the catalogue foundation.
2. Shopping campaign setup: grouped products by commercial economics and purchase intent so budgets and decisions had coherent units.
3. Inventory and data quality checks: monitored expected coverage, diagnostics, landing-page consistency and refresh health to find silent loss.
4. Performance review and scaling plan: compared ecommerce outcomes by readable group and documented the conditions under which proven groups could receive more investment.

The stages formed a dependency chain. A campaign could not be evaluated fairly while important products were missing or misdescribed. Scaling could not become a decision until structure and tracking made product economics visible.

What this project confirmed

Feed quality sets the ceiling

Bid management cannot rescue a title that fails to identify the product or a record that is not eligible to serve. The feed is the campaign’s product-language layer. It needs an owner, validation and a response when catalogue data changes.

Structure by margin and intent, not catalogue alone

A catalogue category can remain useful for customers while being too broad for budget control. Advertising groups should separate products only where economics or intent produce a different decision. Structure exists to improve action, not to make the account look sophisticated.

Check the landing experience on a phone before increasing spend

A relevant listing can still arrive at a slow, unclear or inconsistent product page. The team reviewed mobile product paths, variant selection, price, availability, cart and checkout. We make no unverified lost-revenue claim; the principle is simpler: paid acquisition should not outrun the destination’s ability to serve it.

Shopping and Search answer different purchase moments

Shopping presents product image, price and merchant information to product-oriented demand using feed data. Search advertisements can answer broader category, problem, brand and comparison queries through selected keyword and copy strategy. The channels can work together when their roles, landing pages and conversion evidence are explicit.

Paid social addresses another moment: introducing a proposition, earning interest and developing demand among audiences who may not be actively searching. The linked Meta advertising campaign case study explains that creative and audience problem for the same client. It should not be judged through the feed mechanics on this page.

Teams considering paid versus organic search can read SEO vs PPC return for small businesses. Search visibility without media spend requires useful content and technical authority over time; it is not a feed substitute.

Keeping the feed and account readable

A documented data dictionary identified each field, source, transformation and owner. Merchant labels used for campaign membership were separated from customer-facing descriptions. Product additions and discontinued items followed the same rules as the initial rebuild.

The reporting pack connected spend, ecommerce outcomes, diagnostic coverage and product-group decisions. Every proposed budget change recorded the group, available evidence, commercial constraint and review date. Low-volume groups were allowed to remain inconclusive instead of being forced into confident language.

Feed changes, campaign changes and site changes were annotated. When performance moved, the team could ask what else had changed rather than assigning every effect to the most recent bid decision. This operating discipline is what lets a repaired account remain understandable.

Catalogue teams that need structured entry can use product data entry services. Automated public-source collection is a different risk and workflow covered by web scraping services.

Google Shopping questions

Why does the product feed matter more than bids?

The feed identifies and describes what can enter relevant results; bids influence competition after eligibility and matching. More budget cannot correct missing required data or make a vague record accurately express the product.

How should Shopping campaigns be grouped?

Group products where a difference in margin, intent, priority or evidence would change budget or optimisation. Avoid both one blended catalogue and fragmentation so narrow that no group becomes readable.

How long before performance is readable?

Readability depends on search volume, product count, conversion frequency, tracking quality and the decision being tested. Low-volume products may require a longer period. A two-week demand for certainty does not create sufficient evidence.

Can Shopping run alongside Search campaigns?

Yes. Shopping serves product-led discovery through feed data, while Search can answer category, problem, comparison and brand intent. Shared measurement should preserve their distinct roles rather than forcing one to claim every sale.