A store owner asked me last month which keywords he should add to his Shopping campaigns. It is the most reasonable question in the world and it has no answer, because Shopping campaigns do not have keywords.
What they have is your product feed. Google reads the titles, descriptions, types and attributes in that file and decides from them which searches your products are eligible for. The feed is the targeting. His had never been touched since the day his platform generated it on install.
That is the normal state of affairs, and it is why so many Shopping accounts feel unfixable from inside Google Ads. You are trying to steer with a control that is not connected to anything.
The title is most of the job
If you change one thing, change titles. Google weights the product title heavily when matching a query, and most platform-generated titles are written for a shelf, not a search box.
A title like "Classic Tee — Blue" tells Google almost nothing. A searcher types "men's organic cotton t-shirt navy large". None of those words exist in the title.
The format that works for most catalogues puts the important words first, because titles get truncated:
Brand + Product type + key attribute + colour + size
So that tee becomes "Northwall Men's Organic Cotton T-Shirt — Navy, Large". Same product, now eligible for a far wider set of searches, several of which are closer to a purchase.
Two rules worth holding to. Put the words people actually type at the front, not your internal product names. And do not stuff: a title crammed with every term you can think of reads as spam to the system and as noise to the shopper who sees it.
The attributes that actually change delivery
Beyond titles, a small number of fields do most of the work.
GTIN and brand. Missing identifiers reduce eligibility, and products without them can be excluded from some surfaces entirely. If you make your own products, say so properly rather than leaving the field empty.
Product type and Google product category. Product type is your own taxonomy and you should use it deliberately, because it is one of the cleanest ways to segment bidding later. Google's category should be as specific as the tree allows — not "Apparel" when "Apparel & Accessories > Clothing > Shirts & Tops" exists.
Price and availability accuracy. A mismatch between feed and landing page is the single most common cause of disapprovals, and it usually happens after a sale price changes on the site but not in the feed. Check this before any promotion, not after.
Image quality. The image is the ad. White background, product filling most of the frame, no watermarks or promotional text overlaid — overlays are a policy violation and will get items disapproved at the worst possible moment.
Custom labels. Underused and genuinely powerful. Tag products by margin, by seasonality, by bestseller status, by price band. You cannot bid differently on your high-margin lines if nothing in the feed tells you which they are.
Start with the diagnostics, not the edits
Before rewriting anything, two reports tell you where the money is going.
Open the search terms report for Shopping. You will usually find a large share of spend on queries that were never going to convert — wrong size, wrong gender, a competitor's brand name, research queries with no purchase intent. Negative keywords still work in Shopping even though positive ones do not, and this is where most of the quick savings are.
Then open Merchant Center diagnostics. Disapproved items are obvious, but the ones worth attention are the warnings — missing attributes, limited performance flags, items with poor-quality images. Those products are usually still running and quietly underperforming.
Structure the campaign around profit
Once the feed carries custom labels, the structure can finally do something useful.
The common setup — one campaign, every product, one bid strategy — treats a product with a forty percent margin exactly like one with eight. Splitting by margin band or by bestseller status lets you push the products that actually make money and cap the ones that do not.
The same applies to brand. If Performance Max is running your Shopping inventory, it will happily absorb searches for your own brand name and report excellent returns on revenue you would have had anyway — I wrote about how to separate that and what the honest number looks like afterwards.
The Q4 version of this
Black Friday is on 27 November, which from today is a little over five weeks. Three feed-specific things matter in that window.
Fix disapprovals now. An item disapproved on the 26th cannot be rescued in time, and approvals take longer when the review queue is full.
Get sale pricing into the feed properly. Use the sale price fields rather than editing the price, so the promotion shows as a markdown. Make sure the feed updates on the same schedule as the site, or you will spend peak-rate traffic sending people to a price that does not match.
Do not restructure in the last fortnight. The same rule that applies to campaigns applies here — the five-week plan is the order to work in.
What to expect
Feed work is unglamorous and it compounds. A catalogue with rewritten titles, complete attributes and sensible custom labels does not produce a dramatic chart on the day it ships; it makes every later decision possible, and it removes a category of waste that no amount of bid tuning can reach.
If your Shopping performance has been flat for months while you have been adjusting target returns inside the ad account, the problem was probably never in the ad account.
Two related reads: the negative keyword lists that stop broad match wandering applies to Shopping queries too, and before you judge any of this, make sure the conversions you are optimising against reconcile with your store.
Feed work is part of our Google Ads and Shopify development engagements. The ad account audit includes a feed and search terms review, which is usually where the wasted spend turns out to be.