A Shopping campaign can show plenty of impressions, attract clicks at an acceptable CPC, and still quietly lose money every day. That is why Google Shopping ads underperform for many established retailers: the account is often optimised around platform activity rather than the commercial reality behind each order.
Google does not understand your contribution margin, stock position, return rate or capacity to fulfil another hundred orders. It works from the data you provide and the constraints you set. If product data is weak, conversion tracking is unreliable, or every SKU is treated as equally valuable, automated bidding will scale the wrong outcomes with impressive efficiency.
The campaign is being judged on ROAS alone
ROAS is useful, but it is not a profit metric. A 500% ROAS may look excellent until you account for gross margin, VAT, shipping subsidies, returns, payment fees, discounts and agency or internal management costs. Equally, a 250% ROAS may be highly profitable for a high-margin own-brand range with strong repeat purchase behaviour.
The first job is to establish a breakeven ROAS or allowable cost of sale at product or category level. One blended target across the catalogue is usually too blunt. A retailer selling 20% margin branded electronics and 70% margin accessories should not ask Google to pursue the same return from both.
This is where many accounts become trapped. The campaign reports a respectable average, yet budget moves towards lower-margin products because they convert easily. Revenue rises. Profit does not keep pace.
Your product feed is limiting relevance before bidding begins
Shopping is feed-led advertising. There are no traditional keywords to rescue an unclear title, a thin description or the wrong product classification. Google uses the feed to decide which searches a product can appear for and how confidently it can match that product to buyer intent.
Generic titles such as “Men’s Trainers Blue” give the system very little to work with. A title that includes brand, product type, key attributes, size or capacity where relevant, colour and model identifier gives Google a much clearer commercial signal. The right format depends on the category, but the principle does not: write for how people shop, not for how a stock system labels inventory.
Feed quality also extends beyond titles. Missing GTINs, inconsistent variants, inaccurate availability, poor images and vague descriptions all reduce eligibility or relevance. Custom labels matter too. They allow profitable segmentation based on margin bands, best sellers, price points, seasonality, clearance status and stock depth.
A feed should be treated as a live sales asset, not an administrative export that was configured once and forgotten. For retailers with thousands of SKUs, it is often the highest-leverage area of optimisation.
Not every product deserves paid traffic
A full catalogue in Shopping is not automatically a sensible strategy. Some SKUs have margins too thin to support acquisition. Others are out of stock regularly, have poor landing pages, attract disproportionate returns or are priced uncompetitively against major retailers.
Start by separating products according to commercial potential. Protect proven profitable products, test promising items with sufficient stock, and restrict or exclude products that repeatedly consume spend without a realistic path to profitability. This is not about making the account look cleaner. It is about stopping your best products from subsidising the worst ones.
Performance Max is being allowed to hide the problem
Performance Max can scale Shopping revenue effectively, but it can also make diagnosis harder. Asset groups, audience signals and broad automation do not remove the need for product-level control. In many underperforming accounts, all inventory is placed into one campaign with a single target ROAS, then left to compete internally.
That structure gives Google permission to favour whichever products and placements generate the easiest recorded conversions. The result can be a concentration of spend on branded demand, low-value orders or products with superficial ROAS that fails the profitability test.
There is no universal campaign structure that fits every retailer. A smaller catalogue may perform well with a focused Performance Max setup and disciplined feed segmentation. A large catalogue with major differences in margin, stock and demand usually needs more deliberate separation. The point is control with a reason, not complexity for its own sake.
If you cannot explain where budget is going, which product groups are driving incremental profit, and why a target applies, the structure is too opaque.
Conversion tracking is feeding automation bad information
Google’s bidding systems are only as intelligent as the conversion value they receive. When tracking duplicates purchases, misses transactions, reports gross revenue without refunds, or attributes every sale back to paid media, the algorithm learns from distorted signals.
For eCommerce brands, purchase tracking should be checked against the actual platform order data. Differences will occur because of attribution windows, consent settings and reporting timing, but large or persistent discrepancies need investigation. Enhanced conversions and consent-aware measurement can improve signal quality, particularly where browser-based tracking is incomplete.
The more difficult question is whether the value being sent reflects the value of the sale. If your margins vary significantly, use custom labels and campaign segmentation to align bidding with economics. For more advanced setups, profit-based conversion values may be appropriate, but only if the underlying margin data is reliable. Bad profit data is worse than honest revenue data because it creates false confidence.
Bidding targets are too ambitious, or not ambitious enough
A target ROAS is not a performance wish. It is an instruction that affects how widely Google can enter auctions. Set it unrealistically high and the campaign may throttle, lose volume and stop learning. Set it too low and the system can buy growth that your finance team will later have to explain.
Retailers often change targets too quickly after a few days of weaker results. That is particularly damaging during promotional periods, seasonal shifts or periods of stock change. Smart bidding needs sufficient stable conversion data, but that does not mean handing over control. It means making measured changes and assessing them against a meaningful window.
Before raising a target, ask whether the campaign has enough profitable headroom. Before lowering it to chase growth, calculate the effect on contribution. A lower ROAS target can be the right commercial move when lifetime value is proven and cash flow supports acquisition. It is a poor move when the business is already losing money on first orders without a credible retention model.
The landing page is costing you the sale
Shopping ads bring high-intent traffic, but they do not compensate for a weak product page. If price, delivery information, returns policy, reviews, stock availability and product benefits are hard to find, paid traffic will expose the problem faster.
Check the experience on mobile first. Is the selected variant available? Does the image match the ad? Is the delivery promise clear before checkout? Are unexpected costs appearing late in the journey? A marginal increase in conversion rate can transform Shopping performance because it improves both profitability and Google’s ability to bid competitively.
Price competitiveness matters too. If identical products are widely available and you are consistently more expensive, no amount of bid refinement will create a durable advantage. You may need to focus spend on exclusive bundles, own-brand products, stronger-margin ranges or audiences where your proposition is genuinely differentiated.
Reports are describing activity, not exposing waste
The most dangerous Shopping accounts are not always the obviously bad ones. They are the accounts producing enough sales to avoid scrutiny while leakage builds underneath. A monthly report that leads with clicks, impressions and top-line revenue will not reveal that issue.
Review product-level spend, conversion value, margin band, return behaviour and stock status. Look for products that have accumulated meaningful spend without sales, ranges where performance deteriorated after stock changes, and apparent winners that depend heavily on brand searches. Segment by device, location and time only when there is enough data to make a decision. False precision wastes as much money as neglect.
A profitable Shopping programme is not built by chasing every available click. It is built by making product data, campaign architecture, measurement and commercial targets agree with one another. When those foundations are in place, scaling becomes a controlled decision rather than an expensive gamble.