Why online sales change the setup
Shop campaigns invert the normal order of work. A search advertiser chooses queries and writes ads; a product advertiser hands Google a catalogue and lets the matching system decide which searches each item enters. That puts the feed in the place keywords occupy elsewhere, and it means that a missing attribute or a vague title quietly removes products from auctions no bid could reach. The Google Ads hub sets out the campaign types; this page covers what changes when the outcome is a purchase.
The second shift is that purchases carry their own value. Enquiries and installs are counted; orders are priced, and the price differs from one basket to the next. Value-based bidding exists for that reason, and it is the only objective where the account can ask the system for revenue rather than for a count. The catch is that revenue is not profit. A campaign maximising conversion value will happily sell low-margin stock all day, because the value it receives says nothing about cost of goods. Most of the decisions below are ways of putting margin back into the signal. Where conversion sits in the wider store, from product page to checkout, the ecommerce conversion hub covers the work that happens after the click. The online sales objective hub compares search with the other channels that can carry the objective, and the paid search channel page for online sales sets the engine-neutral tactics, KPIs and budget ceiling this setup implements.
Recommended setup for online sales
| Setting | Recommended for online sales | Why it differs from other objectives | Feature page |
|---|---|---|---|
| Feed | Titles written for search, full identifiers, Google category and product type, custom labels for margin and seasonality | Products are matched from attributes, so the feed does the job keywords do elsewhere | Product feeds |
| Campaign types | Standard Shopping where search-term visibility matters; Performance Max for groups with trusted values | Only shops can use feed-driven inventory; the choice is control against reach | Performance Max structure |
| Product groups | Listing groups split by a margin-band label, with clearance stock on its own | Orders differ in what they earn, not only in what they cost | Performance Max structure |
| Conversion value | Purchase action with the transaction value per order, adjusted for returns and cancellations where the account allows | Lead and install actions are counted; purchases are priced | Conversion tracking setup |
| Bidding | Maximise conversion value, then a target ROAS derived from break-even | Value bidding is meaningful only when conversions carry real values | Smart Bidding strategies |
| Brand | Brand searches in an exact-match Search campaign, brand excluded from Performance Max | Existing demand would otherwise flatter the automated campaign's return | Performance Max structure |
Write the feed as the targeting it replaces
There is no keyword list in a product campaign, so every phrase a shopper might type has to be present in the product data. Titles should lead with brand, product type and the attribute people search on, in the order the product feeds guide recommends per category. Product type and Google product category tell the system which shelf an item belongs to, and identifiers connect it to Google's own catalogue. Custom labels are the part most shops leave blank and the part this setup leans on hardest: they carry information the feed specification does not ask for, such as margin band, season or bestseller status, and they become the levers for splitting campaigns later. Treat the Merchant Center diagnostics view as the first report of the week, because a disapproved product is invisible regardless of budget.
Group products by what each sale earns
A single campaign over the whole catalogue asks one return target to fit products that earn very different amounts. Splitting by category feels natural but usually groups high-margin and low-margin items together. The better split is a margin-band custom label, with listing groups built on it so each band can have its own campaign or target. Clearance lines deserve a group of their own with a lower return target, since the goal there is to move stock rather than to earn the usual margin. Keep the number of groups small enough that each still gathers conversion data; the Performance Max structure guide covers when to split campaigns and when listing groups inside one campaign are enough. Products also change economics as they age, and the guide to new, seasonal and clearance products labels each item by lifecycle stage so that its listing group and target move with it.
Pass a value that resembles revenue kept
The purchase conversion should report the order value of each transaction, not a fixed number. Decide whether that value includes tax and delivery and keep it consistent, because the return target is expressed on the same basis. Where returns or cancellations are common, reduce the reported values for them wherever the account's tracking allows, because a category with a high return rate otherwise looks better to the bidding system than it is, and values net of returns produce a different return figure from gross ones. The conversion tracking tutorial covers the purchase action, the transaction identifier that prevents double counting, and the checks that the value arriving in the account matches the order system.
Keep brand demand out of prospecting
People who search for the shop by name were going to find it. If those searches flow into Performance Max, the campaign's reported return rises without it having created any sales. Run brand termsSearch queries that contain a company's or product's brand name. in their own Search campaign on exact and phrase match, and attach a brand exclusion list to Performance Max so that it spends on shoppers who have not yet chosen. Read the Shopping and Performance Max search terms for branded queries that slip through. The aim is not to stop paying for brand but to see its contribution separately, so that the prospecting return is honest.
Derive the return target from break-even
The target ROAS is an economic statement before it is a bid setting. Start from Break-even ROASMinimum ROAS to avoid a loss., which gross margin implies, add the profit the business requires, and use the ROAS calculator and break-even calculator to make the arithmetic explicit. Then compare the result with what the campaign has returned without a target. If the trailing return sits below the economic target, set the first target near the trailing figure and move towards the goal in steps, as the Smart Bidding guide advises, since a target the system cannot reach shrinks spend rather than lifting margin.
A worked online-sales setup
Tillyard Kitchen Supply, an invented online cookware shop
Tillyard sells pans, knives and kitchen tools online, with an average order value of £64. Gross margin is around 55% on its own-brand knives, 35% on branded pans and 15% on clearance lines. It runs one Performance Max campaign over all 900 products on maximise conversion value with a target ROAS of 400%, set two years ago. Over a month it spends £6,200 for £27,900 in revenue, a ROAS of 450%, but brand searches account for £9,800 of that revenue and a third of non-brand sales are clearance items earning almost nothing.
| Setting | Before | After |
|---|---|---|
| Feed | Page titles copied in; no custom labels | Titles by brand, type, material, size; custom_label_0 set to high, mid or clearance margin |
| Campaign types | One Performance Max over the full catalogue | Performance Max for high and mid bands; standard Shopping for clearance |
| Product groups | All products | Listing groups on the margin label |
| Conversion value | Gross order value including delivery | Order value excluding delivery, returns adjusted weekly |
| Bidding | Target ROAS 400% for everything | High band 230%, mid band 330%, clearance on maximise clicks capped at £0.40 |
| Brand | Served inside Performance Max | Exact-match brand Search campaign; brand excluded from Performance Max |
Break-even ROAS is about 182% on the high band and 286% on the mid band, so the new targets sit modestly above them. In the second month non-brand revenue in Performance Max is £21,400 on £5,300 of spend, a lower headline ROAS of 404%, but gross profit from those sales rises from £3,100 to £5,900 because the mix shifts towards knives. Brand search spends £300 and reports £10,200 of revenue, now visible on its own line.
The headline return fell and the business earned more, which is the point of the rebuild. Removing brand demand made the prospecting figure honest, the margin label let each band bid to its own economics, and moving clearance to a cheaper campaign type stopped it from absorbing the value-based budget.
Measuring online sales in Google Ads
Read the platform's figures on the purchase action alone. Conv. value / cost is the account's label for ROAS, and it should be read per product group or campaign, not only for the account, since a blended figure hides which bands earn their keep. Conv. value and Conversions on the purchase action show revenue and order count; dividing one by the other gives AOVAverage revenue per order. for paid traffic, worth comparing with the store average. The product and product-group reports show which items take the spend and which sell, and they are where a single product consuming the budget without orders becomes visible. Search impression share on Shopping, with its lost-to-budget and lost-to-rank splits, shows whether more profitable sales are available. Merchant Center diagnostics, listing disapproved and limited items, belongs in the same review, because missing products depress every other figure silently.
Beside those, the business needs gross profit after ad cost and the share of revenue that comes from new customers. The platform cannot report either directly unless the values sent are already margin-adjusted, so reconcile them from the order system. Check feed diagnostics and the brand search line weekly; review return by product group, conversion rate (CVRShare of visits or clicks that convert.) on product pages and impression share monthly; revisit margin labels and targets each quarter or when pricing changes. The RACE framework applied to sales places these paid figures in the wider reach-to-engagement chain, and the Shopping and catalogue ads guide compares how other platforms handle the same product data.
Where the setup goes wrong
Feed changes and target changes made in the same fortnight cannot be told apart, so stagger them. A margin label that is set once and never refreshed drifts as costs and promotions change, and the bidding follows a map that is out of date. Splitting the catalogue into many thin groups starves each one of conversions and makes targets unstable. And a sale period inflates conversion rates for a short time; the help centre documents seasonality adjustments for that case, which are safer than raising targets by hand and forgetting to lower them. The guide to bidding through sales and tracking outages sets out when an adjustment fits, how to set the budget beside it and which changes to hold back during the event. Most of these faults can be caught before a retail campaign starts: the Performance Max launch checklist checks goals, values, listing groups and the budget against the target before switch-on.
Common mistakes
- Optimising to gross revenue. Revenue bidding without margin information sells the cheapest-to-sell stock, not the most profitable.
- Letting brand inflate Performance Max. Without a brand exclusion the automated campaign takes credit for demand it did not create.
- Ignoring disapproved products. An item that is not approved cannot enter any auction, whatever the bid.
- Setting the target from habit. A return target copied from another account or set years ago bears no relation to current margins.
- Splitting too finely. Product groups too small to record conversions leave the bidding with nothing to learn from.
