The useful version of this question is not which buttons moved in the interface. It is where control sits now, because that is what decides whether your budget buys the searches you want. Shopping moves in one direction every year. More of the matching, the bidding and the placement is decided by Google's systems, and more of the outcome is decided by the quality of what you hand those systems.
So we read the question as one about inputs. Which of the things you control carries more weight than it used to, and which of the things stores still argue about no longer decides anything.
Our answer, for any store spending real money here: your product data now does the targeting, and your conversion data now does the bidding. Campaign structure matters, but it matters third. If the feed is wrong and the conversion values are wrong, no structure rescues the account. If both are right, several structures work well enough to be worth arguing about later.
The feed became the targeting
You do not choose the searches in Shopping. Google reads your product data and decides which queries your item is a candidate for. That has always been true, and each year it becomes more true as manual controls thin out. It means the title is the single most powerful field you own.
A title written for a shelf tag does not work here. A title written the way a stranger describes the product does. Brand, product type, then the attributes people actually type: size, colour, material, capacity, compatibility, pack quantity. Put the words that carry the search intent in the first part of the title, because that is the part that survives truncation and the part that reads as most relevant.
After the title, the attributes do the fine tuning. Product type is your own taxonomy and it is worth writing carefully, because it groups items for reporting and for bidding. Google product category tells the system what kind of thing this is. Brand, GTIN and condition settle identity. Colour, size, material, gender and age group matter far beyond apparel, because they let your item answer a narrower query where the competition is thinner.
If you want to change which searches you appear for, change the feed. That is the lever now.
Shopping did not take your controls away, it moved them into the feed and the conversion data.
Your conversion values steer the spend
Automated bidding buys what you tell it is valuable. Most stores tell it that all revenue is equally valuable, then wonder why the budget drifts to the cheapest, highest-volume, lowest-margin part of the catalogue.
Two fixes sit under your control. First, decide what counts as a conversion. If newsletter signups, add-to-carts and phone clicks are all counted in the same campaign as purchases, the system optimises for a blend of things, and the blend is not your bank account. Purchases should be the conversion the campaign is paid on.
Second, decide what a purchase is worth. Sending order revenue is the default. Sending a value that reflects gross margin is better, because it tells the system that two orders of the same size are not the same order. A worked example. Say you sell two products at $80. One costs you $20 to buy and ship, the other costs you $60. Revenue treats them as identical and the system will happily spend the same to win either. Margin-weighted values tell it the first order is worth three times the second, and the spend moves accordingly.
The same logic applies to returns. If a category comes back at a high rate, the value you report should account for it, or you are paying to acquire refunds.
Free listings run on the same data your ads do
The unpaid surfaces in Google's shopping results are fed by the same product data as your paid campaigns, and they are held to the same requirements. That has a practical consequence stores keep missing. When you fix a feed problem for the ads, you are also fixing what shows up in the free results, and when you neglect the feed, you lose in both places at once.
It also changes how we judge the cost of a feed problem. A disapproved item is not just an ad that stopped. It is an item that has left every surface Google can show it on. A price mismatch between your feed and your page does not only suspend the campaign, it removes the product from the results that cost you nothing.
This is why we treat Merchant Center as a weekly job rather than a setup task. The diagnostics page is a work queue, not a status light. Items with warnings today are the items that get disapproved next month when the tolerance tightens, and stores rarely notice because revenue drifts rather than drops.
Thin data hurts more than a small budget
The system is not biased against small stores. It is biased against uncertainty. A campaign with very few conversions gives the bidding almost nothing to learn from, so it behaves conservatively, spends unevenly, and takes a long time to settle. That is what most stores experience as automation not working.
The way out is to give the campaign fewer things to be uncertain about. Group products so that the items inside a group behave alike on price, margin and demand. A campaign that holds your ten strongest sellers alongside three hundred long-tail items will spend most of its learning on noise.
We usually split by role rather than by category. The products that already sell get their own campaign and their own target, because there is enough data there to bid properly. The rest sit in a second campaign at a lower target, doing discovery work on a smaller budget. Seasonal or clearance items get their own home so their strange economics do not contaminate everything else.
The budget question follows from that. Splitting a small budget across many campaigns makes every one of them data-poor. Two well-defined campaigns beat six tidy ones.
Telling a real change from a seasonal wobble
Every year stores tell us the platform changed when what actually happened is that a competitor started bidding, a feed attribute broke, or last month was a holiday and this month is not. Before you rebuild anything, run three checks.
Compare like with like. The same weeks against the same weeks last year, not against last month. Retail demand has a shape and it is not your fault.
Separate the price from the volume. If impressions held and clicks held but orders fell, that is a store problem, not an ads problem, and no bid change will fix it. If impressions fell while your budget was untouched, look at the feed and at price competitiveness before you look at bidding.
Check the catalogue against the account. Count active items in Merchant Center this week against the same count last month. A quiet drop of a hundred items explains most sudden performance changes we get called about, and it never shows up in the ads interface as a problem.
Only after those three would we start changing the campaign. Most of the time the change you are feeling is your own data moving.
What we would fix first
In order, because order matters more than effort here. Fix disapprovals and warnings until the item count in Merchant Center matches the item count you expect to sell. Nothing else you do reaches anyone until this is true.
Then rewrite titles for the products that already earn, not for the whole catalogue. Take the twenty items that produce most of your revenue and write their titles the way a stranger would search for them. This is a couple of hours of work with a bigger effect than a week of campaign restructuring.
Then fix what you report as a conversion and what value you attach to it, so the bidding is aimed at profit rather than at turnover.
Then, and only then, look at the campaign structure. By that point you will often find the structure was never the problem. Shopping did not take your controls away. It moved them into the two places most stores were not looking.