In search you buy keywords. In Shopping you don’t buy anything of the kind. You submit product data, and Google decides which queries your products are eligible for by reading it.

There’s no keyword list. There’s no ad copy. Negative keywords are close to the only steering wheel you’re given, and in Performance Max you barely get that. Which leaves the feed doing the work that a media buyer used to do, and explains why so many Shopping accounts plateau while somebody adjusts bids on them every Tuesday.

We’ve managed more than $13 million in ad spend for Australian brands, and across those accounts I can’t think of a Shopping campaign where the largest available gain was in the bidding.

Merchant Centre, the Australian parts

Before any of the interesting work, the account has to be clean, and a few things here are local.

Prices go in as Australian dollars including GST, and they have to match the landing page exactly, including when a sale price drops. A mismatch between feed and page is one of the fastest routes to a suspended item, and it happens most often to retailers who run frequent promotions without a feed that updates in step.

Shipping needs to be configured properly rather than approximated. Australian delivery costs are high enough to change a purchase decision, and a shopper who sees one number in the listing and a different one at checkout leaves. Returns policy belongs in the account too.

Then the basics that trip up newer accounts: a claimed and verified website, business details that match what’s publicly registered against your business, and the target country and language set correctly on the feed rather than left on a default from whichever tutorial you followed.

The feed fields that decide most of it

Product titles carry more weight than anything else you’ll touch. Front-load what a person would actually type: brand, then product, then the attribute that distinguishes it, whether that’s size, colour, capacity or model year. Your internal product name is almost never what anyone searches, and only the first sixty or seventy characters typically survive truncation on the shopper’s screen.

Fill in both product_type and google_product_category. The Google taxonomy is Google’s classification, useful for eligibility. Product type is your own hierarchy, and it’s the field your campaign structure will hang off later. Most feeds we inherit leave it empty, and then segmentation is impossible without a rebuild.

Identifiers matter more than they look. GTIN, brand and MPN connect your item to every other listing of the same product, which is what makes you eligible for comparison placements and what lets Google understand it at all. A feed with missing GTINs is a feed competing with one hand behind its back.

Availability and price accuracy come down to update frequency. Advertising a product you sold out of yesterday is a straightforward way to buy clicks you can’t fulfil, and it costs you twice: the wasted spend, and the shopper who won’t come back.

Images have rules and they’re enforced. Plain background, the product filling the frame, no promotional overlays, no watermarks, no borders. Retailers routinely upload the hero image from a campaign with a sale flash burned into the corner and then wonder why the item is disapproved.

Descriptions are read for matching, so write them for a shopper and make sure the attributes are actually in there. Anything you’d expect someone to type should appear somewhere in the data.

Disapprovals cost more than they look

Open the diagnostics tab before you open the bid settings. Every disapproved item is a product you’re not selling through the channel at all, and the reasons are almost always dull: a price mismatch, a missing identifier, an image that breaks a rule, a policy category nobody knew applied.

The reason this is a commercial problem rather than an admin one is that disapprovals cluster in the same categories, so what looks like a handful of items is often an entire product line that has been invisible for a quarter.

Stop running one campaign

A single catch-all campaign spends on your worst products at the same priority as your best. That’s a decision, even if nobody made it.

Split by what actually matters to your margin. Custom labels are the mechanism: margin band, season, bestseller, clearance, whatever your business runs on. Then fund them differently, because a 40% margin product and a 6% margin product should not be competing for the same dollar on equal terms.

Separate brand from non-brand as well. Someone searching your brand plus a product name is a different customer to someone searching a category, and mixing them gives you an average return on ad spend that describes neither. Averages are how retailers talk themselves into scaling the wrong thing.

Performance Max doesn’t change the answer

Most Shopping inventory now runs through Performance Max, and the honest summary is that it moves the controls further away while leaving the feed exactly where it was.

You still shape it: listing groups, asset groups by product line, brand exclusions so you’re not paying premium rates for traffic you’d have had anyway, search themes to point it somewhere. But the feed still determines what can be shown and what it’s matched to. Thinner reporting makes feed hygiene more important, not less, because you have fewer places left to look when performance drifts.

The listings you’re not claiming

Unpaid Shopping listings run from the same product data. If your feed is in good shape and your Merchant Centre account is compliant, that surface is available to you for nothing beyond the work you’ve already done.

It won’t replace paid traffic. It’s also the closest thing to free money in the channel, and I’d rather a client had it than not.

Measuring it like a retailer

Send conversion value, not just conversion count. If you optimise toward counts, the system will efficiently buy you a pile of cheap, low-value orders, and it will be right to, because that’s what you asked for.

Be consistent about what the value means. Order value with or without GST, freight in or out, pick one and never quietly change it. If margin varies sharply across your catalogue, consider passing a margin-adjusted value instead of revenue, since revenue-optimised bidding will happily chase your worst-earning categories.

And decide in advance which number runs the business, because Google Ads, Meta and your platform’s own analytics will each report a different figure for the same week. That argument is worth having once, properly, and attribution models is where we’ve set out how we resolve it.

Where to start this week

Open Merchant Centre diagnostics and count your disapprovals by product line, not by item.

Take your twenty best-selling products and read their titles the way a stranger would, out loud if necessary.

Put a sale price live on the site and check how long it takes to appear in the feed. If the answer is more than a day, that’s your next project.

Then look at the bids, which by that point will matter less than you expected.

When m|devices moved into selling direct to consumers, most of the work sat in exactly this territory: getting the product experience and the data underneath it right before spending harder on the media. If you’d like us to look at your feed, that’s what our paid search team does, and the plumbing side of it sits with our technology team when the fix belongs in the platform rather than the ad account.

Google Shopping for Australian retailers, start to finish