Google shopping ads: the complete guide

Your product data is the targeting. When someone searches, Google looks at the words in the query and matches them against the text in your feed, mostly the title, then the product type and the description, and decides which of your items are relevant candidates. Bids and history decide which of those candidates actually shows and where.

This is why two stores selling the identical product can get completely different results at the same bid. One writes a title a shopper would type. The other writes a title a merchandiser would write. The first gets shown for the searches that convert; the second gets shown for whatever Google can salvage.

You cannot add keywords to a Shopping campaign. You can only exclude them, which makes negatives the main steering tool you have on the query side. Everything else you do to change what you appear for happens in the feed.

Relevance is not the whole story. Google is trying to earn from the placement, so an item with a low click-through rate is worth less to it than an item at the same bid that people actually click. That means the image and the price are part of your effective bid. A product that is well photographed and competitively priced buys impressions more cheaply than one that is not, for reasons that never appear in the bidding interface.

The practical consequence is an order of operations that most accounts get backwards. First make the data describe the product the way a buyer describes it. Second decide which products deserve budget. Third set bids. Fourth exclude the searches you do not want. Only after all of that does anything else you can change in the interface start to matter.

One more thing about intent. Shopping placements catch people at very different stages, from someone typing a category name with no idea what they want to someone typing an exact model number with a card in hand. Those are worth wildly different amounts to you, and they arrive through the same campaign unless you deliberately separate them.

Do not launch until five things are true, because a campaign started on a broken foundation teaches the algorithm the wrong lessons and you pay for that education twice.

First, the products you want to sell are approved and active in Merchant Center, and the account is linked to Google Ads. Not most of them. The specific items you intend to put budget behind.

Second, conversion tracking records purchases with the actual order value attached, not a fixed number per sale. Value-based bidding is the point of Shopping, and it cannot work if every order looks the same size. Check that the value passing through matches your order totals, and decide once whether that number includes tax and shipping so every comparison afterwards is consistent.

Third, you know your margin by product, at least approximately. You do not need cost accounting to the cent. You need to know which products can afford customer acquisition and which cannot, because that decision drives everything about structure and bidding.

Fourth, the landing pages work on a phone. Shopping traffic skews mobile and it arrives with intent that decays in seconds. If a product page takes an age to become usable, you are buying clicks that bounce.

Fifth, you have decided what a successful order looks like in money. Write down the return you need for the campaign to be worth running, calculated from your own margin rather than a figure you saw somewhere. Everything after launch gets judged against that number.

Also worth doing before launch: set up the negative keyword list you already know you need. Every store has terms it should never pay for, and the obvious ones can go in on day one rather than being discovered expensively over a month. Free, cheap, repair, spare parts, second hand, job listings, and the names of things you do not sell but your product words overlap with.

Last, decide the geography and the schedule honestly. Do not advertise into places you cannot ship to profitably because the ad account makes it easy to include them.

Pick your campaign shape based on how much control you need and how much data you have, not on which option is newer. Google offers two ways to spend on Shopping inventory, and they suit different situations.

Standard Shopping campaigns give you direct control. You subdivide the catalogue into product groups, set bids at whatever level you choose, apply negative keywords, see the search terms and decide the priority of one campaign against another. You can tell exactly why something happened. The cost is your time, and the ceiling is that you are steering manually in an auction that moves faster than you do.

Performance Max campaigns hand more of the decision to Google in exchange for reach beyond the Shopping placement itself. You supply products, assets and a goal, and the system decides placement and bid. You get less visibility and fewer levers, and in return you get access to inventory that standard Shopping alone does not reach.

Most stores end up using both, and the useful question is not which one is better but which products belong where. A reasonable split is to keep your highest-margin and highest-volume products where you can see and control what happens, and let a broader campaign carry the long tail where individual attention is not worth the hours.

If you are starting from nothing, there is a strong argument for beginning with the controlled option even though it is more work. You learn which searches your catalogue actually attracts, which products convert, and what a click is worth in your niche. That knowledge makes every later decision better, including the decision to hand more over to automation. Starting with the automated option gets you spending faster but leaves you with results you cannot explain.

Whatever you choose, do not run two campaigns fighting over the same products without deciding which should win. Two campaigns containing the same item will compete, and the one that serves is decided by rules you may not have thought about. Either separate the products between campaigns, or use priority settings deliberately so the routing is your choice.

Build your product groups around what each item is worth to you, because the default structure treats a low-margin accessory and your best product exactly the same. This is the single highest-value structural decision in a Shopping account.

Google lets you subdivide by category, brand, product type, item id, condition and custom labels. The first four describe what a product is. Custom labels describe what a product means to your business, and that is where the useful grouping happens. There are five custom label fields available in the feed, and you fill them from your own data before the feed reaches Google.

A structure that works for most stores looks like this. One label carries margin band: high, medium, thin. One carries a performance tier based on your own sales history: bestseller, steady, slow. One carries seasonality or price band, whichever matters more in your category. That gives you the ability to say, in one move, that thin-margin slow sellers get a low ceiling and high-margin bestsellers get room to spend.

The reason this beats grouping by category is simple. Categories contain both your best and your worst products. Set one bid for the category and you overpay for the weak items and underpay for the strong ones, every day, forever.

Break out individual products only where they earn it. A handful of items usually accounts for most of the revenue in any catalogue, and those deserve their own group and their own attention. Everything else can sit in a sensible bucket. Splitting a thousand products into a thousand groups produces a structure nobody maintains and data too thin to read.

Rebuild the labels on a schedule, because they describe a moving target. A bestseller in spring is a slow seller in autumn. Regenerate them from your sales data monthly and let the campaign structure follow reality rather than a snapshot from last year.

Finally, use inventory filters or product group exclusions to keep products out of campaigns entirely when they should not be advertised at all: items with no stock depth, items you are discontinuing, items you sell at cost as a favour to a supplier.

Choose a bid strategy that matches how much conversion data the campaign actually has, because automated bidding is only as good as the signal underneath it. A strategy that needs volume, applied to a campaign with a handful of orders a month, produces expensive noise.

When a campaign is new or thin, manual bidding gives you a controlled way to learn. You set a cost per click you are willing to pay, based on your own arithmetic: what an order is worth, how often clicks turn into orders in your category, and what that implies per click. You will be wrong at first. The point is that you can see what happens and adjust with your eyes open.

When there is enough conversion history, value-based automated bidding usually beats what a person can do by hand, because it adjusts for signals you cannot see at the moment of each auction. The strategy to reach for is one that bids toward a return on ad spend target, since it accounts for the fact that some orders are worth much more than others.

Set that target from your own numbers. Work out the return at which the campaign breaks even after cost of goods, shipping, payment fees and any fulfilment cost, then set your target above it by however much profit you need. A target pulled from a general benchmark is a guess dressed up as a decision.

Move targets gently and give each change room. Every meaningful change resets the learning, and an account that gets adjusted twice a week never leaves the unstable period. Make one change, wait for enough conversions to judge it, then make the next.

Watch what a tight target does to volume. Raising the target return does not make the campaign more profitable in absolute terms; it makes it smaller and more efficient. There is a point where efficiency costs you more gross profit than it saves. If total profit is what you care about, be willing to accept a lower reported return on the products where the extra volume pays.

Budget is a separate lever from bidding. A campaign capped well below what it could spend at your target will show you an artificially good return and quietly hide the growth you are refusing.

Read the search terms report every week and cut what does not belong, because Shopping campaigns will happily spend on queries that were never going to buy. Since you cannot choose your keywords, negatives are the main way you shape demand.

There are four kinds of waste to look for. Wrong intent: people looking for repairs, manuals, spare parts, second-hand versions, or jobs. Wrong product: queries that overlap with your words but describe something you do not sell. Wrong stage: very broad category searches that generate clicks and almost never orders at your price point. Wrong buyer: bulk, wholesale or trade queries when you sell retail, or the reverse.

Apply negatives at the level where they belong. Terms that are always wrong for the business go in a shared list applied to every campaign. Terms that are wrong for one campaign but right for another stay at campaign level, which is how you keep a brand campaign and a category campaign from stealing from each other.

The more advanced version of this is routing rather than blocking. If broad category searches convert poorly at full price but you still want that traffic cheaply, you can send it to a separate campaign with a lower bid by using negatives and campaign priority together. High-priority campaigns get first claim on a query, so a low-bid campaign set to high priority with the specific terms excluded acts as a catch-all, while your specific terms fall through to the campaign that values them. This is more machinery than a small catalogue needs, but for stores with wide ranges it separates cheap browsing from expensive intent.

Do not over-prune. Aggressive negative lists starve the campaign of the discovery that finds you new demand, and every term you exclude is excluded until someone remembers to review it. Cut what clearly wastes money and leave the ambiguous ones alone until they have spent enough to judge.

Keep a record of why each negative was added. A list nobody understands becomes a list nobody will touch, and six months later it is quietly blocking a product line you have since started selling.

Point every Shopping click at the exact product page for that item, in the right variant, with the price visible without scrolling. A click on a specific product that lands on a category page is money spent to make a shopper start their search again.

That sounds obvious and it goes wrong constantly. Variant links that land on the default option rather than the one advertised. Out-of-stock products that redirect to a collection. Feed links carrying tracking parameters that break on the store's router. Links that go through a redirect chain and lose the parameters on the way. Check a sample of live feed links by clicking them from a phone, not by reading them in a spreadsheet.

Once the link is right, the page has to do its job. Shopping traffic has already seen your price and your photo, so the page is answering the next set of questions: is this the right one, when will it arrive, what happens if it is wrong. Delivery cost and timing near the price, returns terms in plain words, the sizing or compatibility detail that decides the purchase, and enough images to answer what the description cannot.

Speed matters more here than on most traffic, because the shopper has alternatives one tap away and no loyalty to you yet. The measure worth watching is how long the page takes to become useful on a mid-range phone on a normal connection, not what a desktop test reports.

For out-of-stock items, decide a policy rather than letting the theme decide. The cleanest approach is to stop advertising the item, keep the page live, and offer a notify option. Redirecting the click elsewhere converts poorly and irritates people.

One test worth running on your top products: strip everything above the fold down to image, title, price, delivery promise and the buy button, and see what happens. Shopping traffic arrives ready. A page that makes them scroll past a banner to find the button is charging them for the privilege.

The return figure in the ad platform is a measurement of the platform's own contribution as the platform sees it, and it is not the same as your profit. Use it for steering campaigns and use your own accounts for deciding whether the channel works.

Three adjustments turn reported return into something you can bank. Take out the cost of goods, because revenue is not margin and a strong return on a thin-margin product can still lose money. Take out returns, since a category with a high return rate reports revenue it later gives back. Take account of the fact that some of those orders would have happened anyway, particularly on searches for your own brand name.

That last one deserves separate treatment. Bidding on your own brand can look like the best-performing thing in the account while being the least incremental, because those shoppers already knew you. It is often still worth doing, for reasons like keeping the placement and controlling the message, but it should be measured and decided separately rather than inflating the average.

The number worth putting on the wall is contribution after ad spend: revenue, minus cost of goods, minus returns, minus the ad spend that produced it. Track it monthly by product group. It answers the only question that matters, which is whether more spend produces more profit.

Also watch the blended view. Divide total store revenue by total marketing spend across every channel, and track it over time. It cannot tell you what to change, but it catches the situation where every channel reports a healthy return while the business as a whole is not growing. That pattern almost always means channels are claiming the same orders.

Give numbers time. Daily return figures for a store with modest order volume are mostly noise, and reacting to them produces the pattern of changes that keeps a campaign permanently unsettled. Look weekly for steering and monthly for decisions.

Grow Shopping by widening coverage before you raise bids, because more spend on the same narrow set of products buys progressively worse traffic. There is almost always cheap volume left in a catalogue that is not being advertised properly.

Start with the products that get no impressions at all. Every account has them, and the cause is usually a title that does not match any real search, a missing attribute, or a product sitting in a group with a bid too low to enter the auction. Fixing those adds demand at your existing efficiency rather than buying worse demand at a higher price.

Next, look at where you are losing impressions on products that are already working. If a product is capable of appearing more often but budget or bid is holding it back, that is the cheapest growth in the account. Raising a cap on something already profitable is a different decision from raising a bid on something marginal.

Then widen the range of searches. New product types, sensible variants, complementary items that people already buy alongside what you sell. This is a merchandising decision more than an advertising one, and it usually starts in the feed by describing existing products more completely.

Only then consider paying more per click. When you do, do it where the margin can carry it and accept that a higher bid means a lower reported return by design. The question is not whether the return falls, it is whether total profit rises.

Seasonality deserves planning rather than reaction. Move budgets and targets ahead of your peak, not during it, because the learning period costs you the first part of the season. Get promotional prices and any offer data submitted early. And decide before the rush which products you actually want to sell hard, so that the busiest weeks are spent on the items with margin instead of whatever the algorithm found easiest.

Finally, treat a plateau as a signal to go back to the feed. When bids and budgets stop producing gains, the constraint has moved back to the data, and the next lift comes from titles, images and coverage rather than from anything in the campaign settings.