Google Ads auto-applied recommendations: switch them all off or keep a few?

Auto-apply is where Google's recommendations get messy. Depending on the type, it can raise budgets, add keywords, add responsive ad assets, or change bidding targets, all without a preview. People tend to split into two camps: turn it off entirely and review the list by hand, or leave a narrow set on, like text asset additions, and assume the rest is harmless.

My view is off for everything. A budget raise or a target change can land while nobody's looking, and you don't control the audit trail. Going through the list by hand takes about ten minutes a week, and you still get the useful suggestions. Has anyone kept one specific category on without regretting it?

I'd split this up instead of turning everything off. Adding text assets or keywords doesn't move much money on its own, so letting those run is a different call from letting budgets or bidding targets change overnight. Kill those two first, since they shift the economics of the account while nobody's watching.

Ten minutes a week is also optimistic once you have several campaigns, especially PMax or Shopping. The list grows fast, and the recommendations have no idea about things like a margin drop or a product going out of stock. For a store, I'd review keywords by hand each week, keep the cheap asset suggestions on, and turn off anything that touches money.

Splitting it up makes sense, but for a store it also depends on what your conversion value actually means. If purchases report as plain revenue, a tROAS target that Google nudges down 10% looks like a small dashboard change, yet it can cut margin badly. A sale on a 15% margin product and one on a 45% margin product count the same. Auto-applied target changes are much less risky when the value already reflects margin, for example when you pass profit-based values through a feed or a custom value rule, because then the target is aimed at what you actually care about.

So the real question is whether your conversion values are revenue or margin. With revenue-only values, I'd keep target and budget recommendations off no matter how the account is performing. With margin-aware values you've verified, a small target adjustment is easier to live with. Without that, any target change is a guess about profit.

Agreed on margin, and I'd go a step further. Even with margin-aware values, a tROAS target applies to the whole campaign, so it just averages across every product in it. If a Shopping campaign mixes 15% and 45% margin items, one target still lets the low-margin products spend freely. Splitting into margin tiers with custom labels and setting a target per tier is what makes a small auto-applied nudge reasonable. Without that split, I'd keep auto-applied target changes off even if the values are correct.

Rowan, labels alone won't get you there. tROAS is set per campaign, so labels only help you group products for reporting or for product group splits. They can't give one campaign two targets. If you put a 15% margin tier and a 45% tier under labels in the same PMax or Shopping campaign, the target still applies to the whole lot. A real target per tier means separate campaigns, each with its own budget and target, and that splits your budget too, so the smaller tier can starve on low volume.

So splitting by margin is the right direction, but it has a structural cost. For a small catalog, you're usually better off with one blended target and auto-applied target changes switched off.

Agreed that labels can't carry separate targets, and the budget split is the cost people underestimate. One thing worth adding is that the change history does log auto-applied changes, so there is an audit trail, just after the fact. If you keep a category on, you can check that log weekly and revert a bad target change in one click. That makes a weekly review workable, though it still doesn't replace a preview.

"Revert in one click" oversells it. The log tells you what happened, but it doesn't undo the damage. A budget raise that ran for four days has already spent the money. A tROAS nudge that sat live for a week has already changed how bidding spent, and reverting it often restarts the learning period, so you pay twice. Checking weekly also means a bad change can run for up to seven days before anyone notices, which is a long time on a Shopping campaign with a moving catalog.

So I'd treat the log as a post-mortem tool, not a safety net. If you keep one category on, pick one where the worst case is cheap, like text assets. Target and budget changes need a preview before they touch money, not a rollback after.